I just want to get my 10 or 11 slides presentation

Question:

Background:
I just want to get my 10 or 11 slides presentation on decision making under uncertainty. I will provide you with all the downloaded articles but some of the articles I am not able to download. you will have to chose any two guiding questions and explain them

Decision Making Under Uncertainty

Title of the presentation is decision making under uncertainty and you can chose two or three guiding questions from below. Moreover you need to provide me end notes at the end of every slide which I have to explain every slide what’s written on that. It should be 10 slides or 11 depending upon the data.

Guiding Questions

• How decisions are made in times of disasters?
• Can fully rational decision making process be achieved in disaster management context?
• What are the best criteria in developing a sound decision making in disaster management and emergency responses? Please identify the variables for sound decision making in disaster situations?

You need to use all this data for presentation I am providing you some downloaded articles from this as well.

• Comfort, et. al. 2020. Crisis Decision‐Making on a Global Scale: Transition from Cognition to Collective Action under Threat of COVID ‐19. Public Administration Review 80(4): 616-622.
• Caball and Malekpour 2019. Decision making under crisis: Lessons from the Millennium Drought in Australia. IJDRR 34:387-396
• Uitdewilligen and Waller 2018. Information sharing and decision‐makingin multidisciplinary crisis management teams. Journal of Organizational Behavior 39(6):731-748. PDF
• Kalkman et. al. 2018. Crisis response team decision-making as a bureau-political process. Journal of Contingencies and Crisis Management26(4):480-490
• Varma 2015. Understanding Decision Making During a Crisis: An Axiomatic Model of Cognitive Decision Choices. International Journal of Business Communication. 56(2) 233 –248
• Alison et. al. 2015. Decision inertia: Deciding between least worst outcomes in emergency responses to disasters. Journal of Occupational and Organizational Psychology88:295–321. Pdf Pdf – Alternative Formats
• Binder 2015. Rebuild or Relocate? Resilience and PostdisasterDecision-Making After Hurricane Sandy Am J Community Psychol (2015) 56:180–196. Binder et al 2015.pdf Binder et al 2015.pdf – Alternative Formats
• Daniels RS. 2013. The rise of politics and the decline of vulnerability as criteria in disaster decisions of the United States, 1953–2009 . Disasters 37(4): 669−694. Daniels-2013-Disasters.pdfDaniels-2013-Disasters.pdf – Alternative Formats
• Kolen Decision-making and evacuation planning for flood risk management in the Netherlands. Disasters, 2014, 38(3): 610−635. Kolen_et_al-2014-Disasters.pdfKolen_et_al-2014-Disasters.pdf – Alternative Formats
• John Cosgrave, (1996),”Decision making in emergencies”, Disaster Prevention and Management: An International Journal, 5 (4) 28 – 35. Cosgrave 1996.pdfCosgrave 1996.pdf – Alternative Formats
• J. Wheeler (2002) Decision-making Rules and Procedures for Humanitarian Intervention, The International Journal of Human Rights, 6:1, 127-138, DOI: 10.1080/714003749. Wheeler 2002.pdfWheeler 2002.pdf – Alternative Formats
• Dash and Gladwin Evacuation Decision Making and Behavioral Responses: Individual and Household. Natural Hazards Review 69-77. Dash and Gladwin 2007.pdfDash and Gladwin 2007.pdf – Alternative Formats
• Fiedlerand von Sydow Heuristics and biases: Beyond Tversky and Kahneman’s (1974) judgment under uncertainty. Fiedler and von Sydow 2015.pdf Fiedler and von Sydow 2015.pdf – Alternative Formats
• Tamuraal. 2000. Modeling and analysis of decision making problem for mitigating natural disaster risks. European Journal of Operational Research 122:461±46. Tamura et. al. 1999.pdf Tamura et. al. 1999.pdf – Alternative Formats
• Amos Tversky; Daniel Kahneman 1974. Judgment under Uncertainty: Heuristics and Biases. Science185(4157)1124-1131. pdfTversky_Kahneman_1974.pdf – Alternative Formats

Evacuation Decision Making and Behavioral Responses: Individual and Household

Nicole Dash¹ and Hugh Gladwin²

Abstract: Researchers have examined a wide range of factors that affect evacuation decisions after people hear hurricane forecasts and other information. This review of the literature focuses on three broad areas of research that often overlap: warning, risk perception, and evacuation research. Whereas it is challenging to demarcate the literature along these lines, we believe each of these areas represents important dimensions of evacuation decision making. The literature on warning focuses to varying degrees on warning as a social process, rather than a simple result of hearing official warnings. Warnings by themselves do not motivate evacuation—people must perceive risk. The extensive literature on objective and subjective processes in risk perception has to be evaluated. The review concludes with a focus on some important work in modeling evacuation and evacuation decision-making. Finally, we present recommendations for future research that draws on the strength of earlier work while focusing more directly on risk, the information included in hurricane forecasts, and the timing of those forecasts.

DOI: 10.1061/(ASCE)1527-6988(2007)8:3(69)

CE Database subject headings: Evacuation; Human factors; Hurricanes; Risk management; Decision making.

Introduction

Over the last few decades, considerable research has focused on a range of issues related to hurricane evacuation. These issues in- cluded understanding how people interpret warning messages, how they interpret risk, and what types of protective action they take as a result. The body of literature in the area is voluminous, with significant findings that have shaped the process of evacuat- ing at risk populations in the United States when tropical weather systems threaten coastal areas. In this paper we briefly review this elaborate and significant history while highlighting important di- mensions of evacuation decision making by discussing warning research, risk perception, and research focuses specifically on evacuation (see the following for more extensive reviews: Drabek1986;Tierney et al. 2001;Sorenson and Vogt-Sorenson 2006). Specifically, we discuss what is known about warnings, their characteristics and the importance of viewing them as a social process, the balance of objective and subjective processes in risk perception, and finally we examine actual evacuation behavior with a focus on evacuation modeling and social context. In the current age of increasing threats due to tropical weather, the paper concludes with suggestions for future research. While there has been some belief that we have exhausted research in this area, events such as Hurricane Katrina in New Orleans (2005) and

¹Assistant Professor, Dept. of Sociology, Univ. of North Texas,

P.O. Box 311157, Denton, TX 76205. E-mail: dash@unt.edu

²Director, Institute for Public Opinion Research, Florida International Univ., Room HM 246 BBC, 3000 NE 151 St., North Miami, FL 33181.

E-mail: gladwin@fiu.edu

Note. Discussion open until January 1, 2008. Separate discussions must be submitted for individual papers. To extend the closing date by one month, a written request must be filed with the ASCE Managing Editor. The manuscript for this paper was submitted for review and pos- sible publication on August 4, 2006; approved on March 14, 2007. This paper is part of the Natural Hazards Review, Vol. 8, No. 3, August 1, 2007. ©ASCE, ISSN 1527-6988/2007/3-69–77/$25.00.

Hurricane Rita in the Houston area (2005) suggest the need for a reexamination of paradigms in light of the evolving social world.

Warning

Before research centered specifically on understanding evacuation itself, the focus tended to be on warnings in general and the warning process more specifically. This section of the paper briefly summarizes some of the significant literature on warning as an integral component of evacuation decision making. Whether official forecasts from the National Hurricane Center, advisories from local and national news media, or information gathered from a variety of online sources, warnings trigger the evacuation decision-making process, and understanding how research in this area evolved helps conceptualize the direction future research should take.

Understanding who evacuates and who does not has been one of the cornerstones of research on the preimpact phase of both natural and technological hazards. Warning message characteris- tics, such as its content, source, and frequency, have been an important focus of research (Drabek 1986, p. 74). These early attempts at understanding evacuation (or other protective mea- sures) focused on the warning itself and its belief as key to un- derstanding protective action such as evacuation. The more specific, and less vague the warning, the more likely adaptive response occurs (Mileti et al. 1975). If warnings were heard and ultimately believed, then evacuation would be the end result.

An alternate approach focused on warning response seeks to understand what characteristics of a warning most effectively trig- ger action. Understanding this perspective would assist in rethink- ing evacuation issues and is a vital component in modeling how individuals process warnings in order to effectively predict behav- ior. Mileti and O’Brien (1992, p. 42) summarize public perception and response to warning as follows: “Public response to commu- nicated risk information is a direct consequence of perceived risk (understanding, belief and personalization), the warning informa-

NATURAL HAZARDS REVIEW © ASCE / AUGUST 2007 / 69

tion received (specificity, consistency, certainty, accuracy, clarity, channel, frequency, source), and personal characteristics of the warning recipient (demographics, knowledge, experience, re- sources, social network, cognition); and perceived risk is a direct function of both the warning information and personal character- istics of the warning recipient.”

Clearly, warning goes beyond communication of messages. Warning is a social process that involves a range of activities that precede issuances of warnings, actual dissemination of warnings, receptions of it, and responses to it (Mileti et al. 1975; Nigg 1993). Individuals may listen to the same warning, but not com- prehend the same message. People react to what they hear based on how they interpret messages. Individuals are stimulated differ- ently based upon who they are, whom they are with, who and what they see or do not see, and what they hear (Mileti et al. 1975, p. 43). In other words, even a cry of “FIRE” in a movie theater will be heard, understood, interpreted, and reacted to dif- ferently by individuals. Decision making is composed of a series of sub-decisions as people evaluate the threat, the risk to them- selves, and what they can do about it, adding complexity to the social process of evacuation decision making (Perry and Lindell 1991).

One approach to understanding variations in evacuation is to understand how individuals hear, understand, believe, personal- ize, confirm, and respond to warning (Mileti et al. 1975; Mileti and O’Brien 1992; Mileti and Beck 1975; Nigg 1993; Mileti and Sorenson 1990). Individuals are also influenced by their environ- ment, and social and psychological attributes (Mileti and O’Brien 1992). In a review of 20 years of hazard warning systems, Soren- son (2000) summarizes the types of social and psychological at- tributes that influence the warning process, although he notes that only a few of these factors can be manipulated to make the warn- ing process more effective (Sorenson and Vogt-Sorenson 2006). Response, using a “warning approach,” then, is seen as a function of perceptions that a warning is valid and the risk it conveys is real. Warning response is believed to vary with warning source, warning content, number of warnings, and of course, warning belief itself (Mileti et al. 1975) and is influenced by both sender and receiver characteristics (Sorenson and Vogt-Sorenson 2006). If individuals do not believe warnings are valid or the risk real, then the likelihood of response is decreased.

Clearly the process of warning has changed over the last de- cade, however, the question still remains of how warning infor- mation is used in an ever-evolving context where information is disseminated on various levels by a multitude of sources. Recent research evaluating evacuation responses for Hurricanes Fran and Bertha found that household evacuation decisions are being influ- enced more by media and other household characteristics than by actual warning (Dow and Cutter 1998), thus emphasizing the need to reexamine warning response models. Even before the Dow and Cutter (1998) finding, more complex models of evacu- ation compliance developed, with risk perception as the central focus and more consistent indicator of evacuation behavior (Baker 1991; Perry 1994; Dow and Cutter 1998).

Although the general trend has been for “modern-day” hurri- canes striking the United States to cause significantly more dam- age than deaths, our inability to understand the evacuation phenomena will leave thousands on highways trying to flee and tens of thousands in their homes in low-lying locations vulnerable to storm surge—as happened with Hurricane Katrina in New Or- leans. While no completely accurate estimates exist, some argue that approximately 100,000 people remained in New Orleans as the city flooded due to the levee breech (Brinkley 2006). With

over 1,300 casualties in Louisiana alone (Knabb et al. 2005), understanding how to better motivate evacuation is clearly criti- cal. In addition to the warning process, another key component to evacuation decision making focuses on how individuals perceive risk.

Risk Perception

As illustrated, warning plays a significant role in our understand- ing of evacuation decisions, yet, focusing on warning alone offers an incomplete picture of the complicated process that results in evacuation. Risk perception is also a vital piece of the puzzle. However, risk perception is complicated by the uncertainty within the situation itself (such as determining the probability of impact). The underlying issue is that evacuation decision making, and per- haps hazard decisions in general, are complex processes that are difficult to categorize. Understanding the decision-making pro- cess is complicated by the fact that the situation itself is often uncertain when making hazard decisions in general, and always uncertain when making hurricane decisions in particular. As a result, decision makers have a difficult time grasping and under- standing not only the probability of the event, but also the range of options available to them (Slovic et al. 1974; Burton et al.1978). Although emergency managers and others assume that people will act rationally—hear a warning, realize the danger conveyed in that warning, and leave when told to do so (because the cost of staying outweighs the benefit)—more often than not, many of those at greatest risk choose not to take protective mea- sures each time a warning is given.

Risk perception is one of the key factors in understanding the evacuation decision-making process. Knowledge about hazards alone is not enough to motivate action. Instead, information must be translated into a concrete conception of pending danger. Al- though risk can be seen as a technical notion calculated based on the probability of events and the magnitude of specific conse- quences (Kasperson et al. 1988), others define it based on its social meaning, characterized by worry, dread, angst, concern, or anxiety (Rogers 1982; Jaeger et al. 2001) whereas other research- ers view it as more of a social concept that takes into account context and culture in the interpretation of what is dangerous (Turner 1979; Tierney 1994;White 1994).

In his 1987 article, “Perception of risk,” Paul Slovic (1987,

1. 280) asserts that “whereas technologically sophisticated analysts employ risk assessment to evaluate hazards, the majority of citizens rely on intuitive risk judgments, typically called ‘risk perceptions.’ ” He argues that difficulties in understanding proba- bilistic processes, biased media coverage, misleading personal experiences, and the anxieties generated by life’s gambles cause uncertainty to be denied, risks to be misjudged, and judgments of fact to be held with unwarranted confidence. We may therefore conclude that models of individualized risk perception must in- clude social dimensions based on the decision-maker’s frames of reference.

Individuals process information through their own social lenses constructed by their particular cultural context, and as a result, different people may well interpret the same information and messages differently. Turner (1979) asserts that individuals often have problems assessing their own risk, and these risk perception problems derive from difficulties in obtaining and processing information. Understanding how people obtain and process information, then, is a vital part of understanding hurri-

70 / NATURAL HAZARDS REVIEW © ASCE / AUGUST 2007

cane warning response. We explore this in greater depth in the following.

Gilbert White (1994) presents a model of perception that takes into account social characteristics that influence how people de- termine their risk, in an effort to understand the interrelationships that result in risk perceptions. This perception is presented as a product of distinct dimensions including social system, decision maker, and environment, which each have their own unique com- ponents, while at the same time interacting with each other.

White’s (1994) social system dimension includes socioeco- nomic characteristics such as race, income, and age which occur in specific social contexts that affect the ultimate outcome, whether evacuation decisions or risk decisions. Similar context variables, included in this research as individual level indicators, are independent of the hazard event and represent the social land- scape of the decision maker, which ultimately influence decision outcomes. The decision-maker dimension focuses on the ability of decision makers to process and understand information. This notion is more psychological in nature and attempts to measure the cognitive ability of decision makers—do decision makers have the psychological ability to process the information being presented to them. And finally, the environmental dimension is characterized by knowledge of the magnitude, frequency, dura- tion, and location of the hazard, which then affects the decision maker. Risk perception, White argues, occurs at the nexus of these dimensions, and is a function of environmental factors, technological factors about the hazard, economic effect, and so- cial linkages affected by the social location of an individual (White 1994; Kasperson et al. 1988).

Although research has examined many of the factors that in- fluence risk perception, there is much that we do not know. In fact, Tierney (1994) notes that sociologists increasingly argue that existing research on risk can be criticized less for what it has found than for what it has failed to examine. She is most critical of the notion that there is some objective and knowable calcula- tion that individuals can assess as it relates to risk. She refutes the idea that individuals know their level of objective risk. In other words, even when presented with information such as elevation of home or location near the coast, individuals may still interpret that information through their social lenses, and as a result, their risk determination is not objective. From this stems a major criti- cism of an approach that focuses solely on risk communication theory.

Research both in and out of the hurricane context adds to our overall understanding of how individuals personalize risk. Hurri- canes present near term risks which are often difficult to perceive compared with risks of distant events (Mileti and Fitzpatrick 1993). However, although this may be true in general terms, Mileti and Fitzpatrick (1993) found that individuals were able to personalize risk of an earthquake prediction through a process that included seeking information from many sources and talking about the risk with others. As people learned about the earthquake prediction through educational material, they sought additional information that allowed them to create a “personal definition” of the risk they faced (Mileti and Fitzpatrick 1993, p. 86). Through the personalization process individuals transform abstract notions of risk into concrete personalized assessments of risk for them- selves, their families, and their households (Mileti and Sorenson 1987).

Personalization also relates to an individual’s ability to accu- rately identify their risk area (Zhang et al. 2004). In other words, simply receiving risk information is not sufficient to mediating response, as it requires the processing of information. This is

critical in today’s world as individuals have the ability to seek and interpret additional information on their own. For example, a variety of websites not only report official National Hurricane Center warning information, but also the details of the models the forecasters use to determine their official predicted path. As has been argued, then, individuals use this information to determine their personalized risk without relying on official warning sources to interpret the information.

Dash (2002) attempted to capture this notion of personalized risk by creating an index, utilizing the following questions: (1) “As it approached, how dangerous did Hurricane Georges seem to you then, in terms of death or serious injury?” and (2) “How concerned were you about damage or destruction to your home when Georges approached?” The scale was positively related to increased likelihood of evacuation. Respondents who thought the hurricane was dangerous were also asked why they thought so. Results of this follow-up question were coded for storm direction, storm strength, and damage already done by Hurricane Georges in the Caribbean. Modeled by themselves, all three predicted higher evacuation rates, although only the last appeared to be a unique predictor. The results indicate that respondents processed the in- formation they were receiving about the storm to contextualize the situation for themselves.

Tierney (1994) supports this notion of personalization of risk by focusing on risk as a social construction. “A social construc- tivist approach does not claim that there is no objective basis for believing that certain risks exist,” (Tierney 1994, p. 6) but does not focus solely on these objective notions. Instead Tierney (1994) proposes that “the basic¼task is to explain how social agents create and use boundaries to demarcate that which is dan- gerous” (Clarke and Short 1993, p. 379). Social and cultural factors, therefore, cannot be ignored when analyzing and under- standing risk. Information is processed within social contexts that influence how individuals assess the level of danger. As “poten- tial” threats become realized threats, and as abstract, vague ideas of potential damage become real, levels of danger may increase. At this juncture, decision making often becomes more compli- cated as decisions are in part influenced by risk perception, which itself is influenced by official messages characterizing the threat as real.

Gender is also an important factor in the personalization of risk, as Bateman and Edwards (2002) indicate. In a study of households making evacuation decisions for Hurricane Bonnie, they found that the situation of women with regard to both objec- tive (living in a mobile home, on a barrier island, close to water, or in a storm surge zone) and subjective perceptions of risk (per- ceived risk of flooding and from wind) was typically different from that of men. Their analysis parses these objective measures of risk and perceptions of it by gender, and shows that, in general, women have greater objective risk (living in riskier housing) and more realistic perceptions of it (Bateman and Edwards 2002, pp. 107, 116). Gender is often not measured by evacuation studies that focus on the household as the decision-making entity (unit of analysis), but where it is measured, female respondents are usu- ally more likely to report having evacuated (Morrow and Gladwin 2005).

Other important work found that risk perception is more im- portant than negative previous hurricane evacuation experiences. Although popular belief would seem to indicate that bad evacua- tion experiences result in a lower likelihood of evacuation for future storms, Dow and Cutter (1998) investigated the effects of the “Crying Wolf” syndrome (an evacuation order for a storm that misses). Their work examined evacuation behavior for two 1996

NATURAL HAZARDS REVIEW © ASCE / AUGUST 2007 / 71

storms in South Carolina, Hurricane Bertha (which had minimal effects), and 2 months later, Hurricane Fran. Dow and Cutter (1998, p. 245) found that “despite the difference in the storms, there was a considerable degree of consistency in individuals’ decision to evacuate.” They report that 39% of respondents evacuated for both storms, and another 37% remained in their homes for both storms. Only 3% evacuated for Hurricane Bertha but not for the subsequent Hurricane Fran. It is notable that, at least in overall evacuation percentages, the “cry wolf” phenom- enon did not occur for the Katrina evacuation in Louisiana. Evacuation rates were as good as or better than those for Ivan the year before, even though Ivan went elsewhere and during the evacuation many thousands of evacuees were stranded in traffic jams (Morrow and Gladwin 2005).

In later work, Dow and Cutter (2000, p. 144) found that a new “hurricane-savvy” population no longer needs to wait for govern- ment warning and advisories to make hurricane decisions. They argue that as Hurricane Floyd neared the coast, residents used factual and personal information such as storm intensity, landfall, past experience, and their own hurricane knowledge to make in- dependent assessments of how to respond to the storm. Included in this process was also an assessment of probable traffic delays. Dow and Cutter (2002) argue that such delays are becoming more important in the evacuation decision-making process as individu- als use various sources of information to assess the risk of leav- ing, as well as the risk of staying.

Dash and Morrow (2000) also investigated the effect of evacu- ation return delays on future evacuation plans. They hypothesized that those who evacuated and experienced lengthy delays at road- blocks would be less likely to evacuate for the next storm threat. The results, however, suggested that those who experienced the delays were less likely to be adversely affected than those who knew of the delays only through media reports. Similar to the conclusions drawn in the work by Dow and Cutter (1998), Dash and Morrow conclude that risk perception has greater saliency in the decision-making process than prior adverse experiences like traffic delays.

The goal, then, for evacuation researchers is to develop models that include key components of decision-making processes in- cluding risk perception. More important, however, is that risk becomes a product of not only the technical, but also the social. Renn et al. (1992, p. 139) suggest that a “novel and integrative framework is necessary to analyze the social experience of risk and to study the dynamic processing of risks by the various par- ticipants in a pluralistic society.” To do this, they argue that risk needs to be approached as both a social and technical concept through a “social amplification of risk” approach that tends to see risks from a broad perspective. Even though risk perception is a major factor in evacuation-related decisions, it is not the only influence; likewise, focusing only on individual-level variables fails to capture complexities of the decision-making process. Ap- proaching evacuation as a process and not as an outcome is key to understanding why some evacuate and some do not, and more important, to determining what can be done to motivate more compliance.

Risk perception, then, is a critical component in understanding how individuals decide to evacuate or to stay. Whether for those in an official evacuation zone with the expectation of leaving as a storm approaches or for “shadow evacuators” who perceive per- sonal danger despite not being in an evacuation zone, understand- ing how people decide that an event poses risk to themselves and family is critical to modeling evacuation behavior, and may shed some light on how to rework or retool messages that take some of

these components into consideration. It is important to understand how people transition from hearing evacuation orders (warnings) to deciding to evacuate—a process which combines what is known about warning compliance and risk perception.

Evacuation

As we see time and time again, for as much research as has been conducted on the issue of evacuation, our understanding of evacu- ation is extremely limited. Those expected to evacuate often do not, and those who should not evacuate (at least in the estimation of emergency managers) often do. The following focuses on the body of literature that specifically emphasized understanding evacuation compliance with a discussion of the small, but grow- ing literature focused on the evacuation decision-making process itself.

Research on evacuation has focused on the characteristics of those who evacuate and those who do not (Baker 1979; Cross 1979; Baker 1991;Fischer et al. 1995;Dow and Cutter 1998; Drabek 1999), or on difficulties associated with evacuation (Baker 1980; Mileti and Sorenson 1987). Other research, such as Perry and colleagues (1981), Gladwin and Peacock (1997), and Whitehead et al. (2000) attempted to model evacuation compli- ance, while Lindell et al. (2005) focused on household decision making as it relates to hurricane evacuation. It appears that over time, more complex models of evacuation compliance have de- veloped, often, but not always, with risk perception as a central focus and with more reliable indicators of evacuation behavior (Baker 1991; Perry 1994; Dow and Cutter 1998; Whitehead et al. 2000).

Historically, factors such as age of the decision maker (Mileti et al. 1975; Gruntfest et al. 1978; Perry 1979), presence of chil- dren or elderly in the household (Carter et al. 1983; Gladwin and Peacock 1997), gender (Bolin et al. 1996; Fothergill 1996; Bate-man and Edwards 2002), disability (Van Willigen et al. 2002), race and ethnicity (Drabek and Boggs 1968; Perry et al. 1982; Perry and Mushkatel 1986), and income (Schaffer and Cook 1972; Sorensen et al. 1987; Bolin 1986) have all been shown to influence evacuation outcomes. Depending on the situation, these factors can either motivate or constrain evacuation. For example, the presence of children in the household might influence parents to protect them from danger, yet the lack of resources to evacuate may hinder the ability to take protective measures. Decision mak- ers are influenced by these factors while determining risk, e.g., hearing notices of potential hazards and determining the safety of their structures, when interpreting warning messages, and finally when making the actual decision to take action. Both previous experience (Hutton 1976; Baker 1979; Perry and Greene 1982; Sorensen et al. 1987) and geographic location, such as proximity to highways and exit routes (Simpson and Riel 1981), play a similar role in understanding evacuation decision making. While these findings are important in our overall knowledge of the evacuation problem, other models focus less on these character- istics and more on the processes that lead to evacuation.

Perry et al. (1981, p. 1), for example, have created progres- sively more complex models of evacuation compliance by con- centrating on the “development of forecasting techniques and warning systems that generate a protective response” by those threatened. Rejecting rational choice and using emergent norm theory which focuses on the “development of situational norms and expectations that arise as a function of some crisis or change in the social or physical environment that renders traditional

72 / NATURAL HAZARDS REVIEW © ASCE / AUGUST 2007

norms inappropriate” (Perry et al. 1981, p. 27), four initial vari- ables emerged. These were: threat seen as real; level of perceived risk; possession of adaptive plan; and family context (family members all accounted for). In their initial model, they found a direct relationship between perceived threat, personal risk, and presence of an adaptive plan for evacuation, with perceived per- sonal risk the strongest predictor. This work played a key role in bringing social context to the forefront in evacuation modeling. Whereas these variables for the most part remain key factors in their more detailed models, in later work other factors also emerged as important. These variables include the warning mes- sage itself, previous experience, and respondents’ social networks. Although these models help address some key issues, they fail to include other socioeconomic factors such as the inability to garner resources that can seriously constrain decision making.

Gladwin and Peacock (1997), on the other hand, focus on contextual indicators in their work on evacuation compliance dur- ing Hurricane Andrew (1992, Florida), such as the key role of social factors on a household’s ability to respond to an evacuation order. In their work, they recognize the complex nature of evacu- ation and the factors that may influence compliance, and note that to respond, households must receive accurate information, pro- cess the information, determine their danger, make decisions, and gather the appropriate resources (Gladwin and Peacock 1997,

1. 54). At each one of these stages (which at times occur concur- rently), household decision makers must not only manage the facts of the situation (as they know them), but more important, do so with social constraints and influence. While households in a community, in this case South Florida, may seem to be working similarly toward the same goal, it is rare to find “cohesive com- pliance” (Gladwin and Peacock 1997, p. 54). They argue it is important to understand that except under extreme circumstances, households cannot be compelled to evacuate or to remain where they are, much less to prepare themselves for the threat. Even under extraordinary conditions many households have to be indi- vidually located and assisted or forced to comply. Segments of a population may fail to receive, ignore, or discount official re- quests and orders. Still others may not have the resources or wherewithal to comply.

Using survey data collected after Hurricane Andrew, Gladwin and Peacock (1997) found that three types of variables stand out as unique and significant predictors of evacuation: (1) being in an evacuation zone; (2) having demographic factors associated with small households and the presence of either elders (negative ef- fect) or children (positive effect); and (3) living in a single-family dwelling (negative effect). Their research found that having chil- dren in the household and being located in an evacuation zone closely rival each other in their relative importance as predictors. Yet, significantly missing from their models are specific measure- ments assessing perception of risk.

Although similarities exist with respect to the factors previ- ously found to be important for both those in and out of evacua- tion zones, a number of differences also emerge. As mentioned previously, household size, the presence of elders or children, and residence in a single-family dwelling remain strong predictors, regardless of a household’s location. For households within evacuation zones, however, number of years in South Florida and income become significant factors as well. In particular, the longer a household has resided in South Florida, the lower its odds of evacuating. Upper income households were much more likely to evacuate. It is also interesting to note that Black house- holds (of which the majority are African American) that reside in evacuation zones were less likely to evacuate, with their odds

being reduced by almost two-thirds compared to Anglos. These findings clearly suggest that very different factors influenced the evacuation decisions made by South Floridians living in different locations.

In addition, the work by Gladwin and Peacock (1997) is one of the pieces to explicitly deal with shadow evacuation (those who evacuated even though they might not have been required to). Assessing risk response becomes even more complex when the issue of shadow evacuators is acknowledged. Shadow evacuation is a serious concern because, as Hurricane Floyd in North Florida and Georgia, and Rita in Texas demonstrated, mass shadow evacuation can overwhelm evacuation highway capacity. Shadow evacuation is not well studied because most behavioral research has focused on people in storm surge and other risk zones who should evacuate, rather than people living in safer locations. More recent work on evacuation attempts to focus on the decision- making factors that influence the types of action individuals and households take when storms approach. One such approach mod- eled both real and hypothetical evacuation from Hurricane Bonnie and a similar scenario-based storm in coastal North Carolina (Whitehead et al. 2000). Whitehead et al. (2000) take into account objective and perceived risk in the development of their model, and focus on social, economic, and risk variables, and how these variables correlate with evacuation decision making. Whereas risk is central to their developed model, they broaden it to include economic and social variables such as income, gender, race, and the presence of at least one pet in the household.

Their results focused on both actual evacuation in Hurricane Bonnie and hypothetical evacuation for future storms. The find- ings indicate that respondents’ likelihood of evacuating for future storms depends on the intensity of the storm scenario. In other words, those who were given more intense storms in their hypo- thetical scenarios were more likely to indicate that they would evacuate than those who were given less severe storms. In addi- tion, households that received voluntary or mandatory evacuation orders were more likely to evacuate than those who did not re- ceive any order. Fear of flooding was also more likely to result in greater hypothetical evacuation compliance than fear of wind (Whitehead et al. 2000).

Dash (2002) extended this approach by assessing subjective risk measures on the Hurricane Georges evacuation in Monroe County (Florida Keys) and Miami-Dade County. Dash found: (1) if the respondents felt that they should do what was best for them even if authorities said otherwise, evacuation likelihood de- creased; (2) if the respondents knew that an evacuation order had been given, evacuation likelihood increased; (3) if the household had an evacuation plan before Georges, evacuation likelihood in- creased; (4) if the household had evacuated for Hurricane An- drew, evacuation likelihood increased; and (5) for larger family sizes, evacuation likelihood decreased and for households with young children, evacuation likelihood increased. These findings are significant for researchers attempting to improve predictive models. Questions, however, remained as to how these character- istics affect decision making. For example, do decision makers with large families think directly about these issues when decid- ing whether to evacuate or not, or are they simply a lens through which information is filtered?

In an attempt to focus more on these decision-making factors, Lindell et al. (2005) investigated evacuation decisions related to 2002’s Hurricane Lili in Louisiana. Their comprehensive study focused on both decision-making factors and evacuation time is- sues. Of particular interest was their hypothesis that “risk area residents rely on some information sources more than others”

NATURAL HAZARDS REVIEW © ASCE / AUGUST 2007 / 73

(Lindell et al. 2005, p. 172) and in a particular order, with local news media being used for information first and the internet being used last. In addition, they hypothesized that the most important information is based on environmental cues and the least impor- tant on evacuation impediments. To assess these hypotheses and others, they asked a series of questions on information sources and decision-making factors as related to environmental evacua- tion cues (e.g., proximity to coast and storm conditions), social cues (e.g., seeing peers evacuating, hearing announcements of watches and warnings, and hearing official evacuation recommen- dations), personal experience (e.g., previous experience with hurricanes and unnecessary evacuations), and evacuation impedi- ments (e.g., protecting home from looters, evacuation expenses, and being struck in traffic at landfall). Their findings, although supportive, for the most part, of the hypothesized relationships, found that whereas residents at risk primarily use news media for information, evacuation was more highly correlated with respon- dents receiving information from peers and local authorities. Similarly, whereas respondents indicated that environmental cues were most important in their decision making, social cues were actually more highly correlated to evacuation. More research needs to focus on these factors, paying particular attention to what types of information are consciously considered in the evacuation decision-making process.

An example of research more directly incorporating kinds of information considered in evacuation decision making is the eth- nographic decision model approach employed by Gladwin and her colleagues (Gladwin et al. 2001). In this approach, a general decision model—to evacuate or not evacuate in the face of a hurricane threat—is inductively derived from specific individual reports about recent evacuation decisions. Residents who had been in South Florida during both Hurricanes Andrew and Erin in 1992 and 1995 were interviewed. From the personal reports of 60 respondents, key individual decision models were constructed. Respondents were asked to mentally put themselves back into the situation in which they found themselves as the hurricane ap- proached, establishing a general overall sequence of events. With the assistance of the interviewer, respondents were then reguided through the various decision points that emerged from their sto- ries. Many of these decision points seemed to be automatic in that respondents could not remember making a decision (cf. Tversky 1972). When questioned, however, individuals were able to recall elements of the decision process, such as, “In those days when I heard the authorities put out an evacuation order, I just assumed I would leave. I didn’t question it. Now I might do it differently.” Based on these accounts as well as answers to the contrasting question, “Why did you decide to evacuate with Hurricane An- drew but not with Hurricane Erin?,” flow charts were constructed to model the decision of each household and then combined into a joint flow chart that modeled the decisions of all households. This method can be utilized for decision processes where induc- tive study reveals that most households face the same top-level information (e.g., information they get from television and other media sources about hurricane risk and evacuation orders) and constraints. The model was then adapted to a format suitable for a telephone survey and tested with a random sample of 954 South Florida residents.

What we can conclude from this brief overview of evacuation research is that when considering how to best warn individuals and households in today’s environment focus should continue to be placed on understanding how individuals and their households arrive at a decision to evacuate or not. Assessing how information is used and interpreted including warning messages, evacuation

suggestions, and storm knowledge should be emphasized; how- ever, this should be done with the realization that decision makers also consider a variety of different information such as feelings of safety of one’s home, family size, and previous experience. Inte- grating major decision-making factors such as individual-level indicators, event-oriented variables, and risk perception allows us to better conceptualize the process by which people come to re- alize their risk and take protective measures.

Future research to understand evacuation must move beyond understanding the characteristics of those who evacuate and those who do not, toward an understanding of what factors people con- sider as they make their decisions and how important those fac- tors are in the process. Today, many people hear that a storm is approaching the United States or the Caribbean, and on their own, seek information from weather web sites and determine their own risk (see Dow and Cutter 2000). We know little about how this new information is used. Do individuals have the capacity to un- derstand and process the information they use to make their evacuation decisions, and how does this information affect not only the decision to evacuate, but probably more important, their risk perception such as how safe their home is?

Where Do We Go from Here?

Much progress has been made, with important contributions from a range of behavioral research disciplines, in understanding evacuation behavior. For example, economists and psychologists coming from a rational-choice perspective have found systematic deviations from optimal behavior (Kunreuther and Linnerooth1984; Slovic et al. 1984; and others cited previously). Psycholo- gists and others using a social cognition framework have shown the importance of self-efficacy, particularly that which can be derived from a better understanding of forecast information (Burnett et al. 1997; Dow and Cutter 1998; Benight et al. 2004). Still other researchers study how information processing and veri- fication constraints affect the timing of evacuation decisions Lin- dell and Perry 1987; Sorenson 1991; Lindell and Perry 2004). Sociologists and anthropologists using multiple regression and ethnographic decision modeling have been more inductive and eclectic, opting to model any relevant variables that will make evacuation rates more predictable. We believe that all these ap- proaches are valuable and have contributed to a better understand- ing of the evacuation process. This leads us to the question: What should come next in evacuation research?

We believe the following three objectives to be important. First, more accurate and geographically focused prediction of evacuation rates is needed. This would give emergency managers a better idea of where evacuation orders would be followed and where they, along with forecasters and the media, should focus their efforts to improve communication of forecast information and the risks people face if they do not evacuate. Second, better prediction of evacuation rates would enable better estimation of potential hurricane consequences that depend on evacuation rates, including clearance times, shelter usage, and potential casualty rates. And third, more research should focus on understanding shadow or spontaneous evacuators. Understanding what factors those not at greatest objective risk consider when deciding to leave should help better target educational programs.

We should look much more closely at the content and flow of information from forecasters to decision makers (both officials who make evacuation calls and people who are supposed to evacuate when ordered). Whereas much is known about warning

74 / NATURAL HAZARDS REVIEW © ASCE / AUGUST 2007

and its process, new research contextualized for the information age may shed light on how information is utilized by decision makers. This is important as it assists forecasters and others down the information line in better shaping and communicating forecast messages. A good example is the current discussion of the “cone of probability” versus the forecast track. To decision makers, forecast information feeds into subjective indicators of risk and the process of assessing one’s risk. For example, although we can argue that those in mobile homes are more vulnerable, some still believe that mobile homes are safe. We must first understand how people come to their risk conclusions, including how they inter- pret warning messages and assess the safety of their homes, before we can effectively suggest changes to the messages themselves.

Finally, we must better incorporate time into evacuation mod- eling. In urban areas people often decide to evacuate too late, becoming caught in traffic jams that potentially expose them to greater risk than if they had stayed home. Or they may hear about the traffic and not evacuate when they should. With better tempo- ral modeling we can more easily predict what will get people to evacuate in a timely manner. Evacuation behavior models can also be used as inputs to traffic-clearing models if they incorpo- rate time into their predictions.

To move more rapidly toward meeting these objectives, what new research emphases might be added to the types of research already being conducted? One important task is to look more closely at what people are thinking and doing when they hear and act on forecast warning messages. How do people use warning information? How much do they simply take and accept what they hear, how much do they gather information and then inter- pret it for themselves, and how do they interpret the information? These questions are cognitive in nature; we therefore believe that more cognitive-based research should be incorporated to better understand evacuation thinking and decision making. People in- terpret warning messages in terms of their beliefs and knowledge. These are usually remembered as scenarios or stories (Schank and Abelson 1977), often taking the form of causal relations—beliefs about likely consequences of events that get modified through experience (see Cameron 2003 for information on health risk per- ceptions). Such “stories that teach an important lesson” may embed decision criteria and constraints that are hard to measure with purely quantitative surveys.

The traditional linear flow of hurricane warnings from fore- casters to the public has evolved “into a nonlinear system as citizens, enterprises, and local governments [come] to rely in- creasingly on sources outside institutional channels, especially broadcast media, the Internet, and peer-to-peer communications”(Gladwin et al., 2005). In this situation many players contribute knowledge and expertise to the evacuation decision process, in what has come to be called “socially distributed cognition.” Most of the research we discuss in this paper recognizes the social nature of evacuation decision making, but there has been very little work done on the socially distributed cognition involved. This latter research area examines situations where considerable risk is involved if decisions are not made correctly (Hutchins 1995). In such circumstances, correct information transmission, relying on commonly understood measuring scales, is critical. Relying on correct information may be difficult, however, in situ- ations such as during a hurricane warning where forecasters, emergency managers, and the public have very different levels of expertise and understanding of scales and diagrams representing hurricane risk (“track,” “probability cone,” etc.). More research of this type could improve prediction of evacuation rates. It is also

likely to provide a better rationale for selecting variables in build- ing models, leading to better evacuation rate predictions.

This research would also furnish more detailed mechanisms for operating temporal decision models at the household level and at the large-scale information flow level (Lindell and Prater 2002). It could also enable new types of evacuation decision simulations to be constructed in the same manner as expert sys- tem programs (Giarratano and Riley 1994) and model evacuation decisions over time as new information and constraints get ap- plied to evacuation decision makers.

Evacuation research is hampered by a weakness often encoun- tered in social science research: the problem of recall. After a hurricane hits or misses and time passes, people can experience difficulty remembering exactly what happened hour by hour and how their understanding of the situation changes throughout the decision process. Studies necessarily have been conducted after the fact and, unfortunately, some people rationalize that they made the best decision, altering their memory in the process. To address this issue, more systematic preevent research focused on evacuation needs to be conducted in areas prone to hurricanes. Whereas there is no guarantee that events will take place, under- standing the process that takes place as a storm approaches should significantly improve our understanding of evacuation and the effectiveness of warnings. We do not know enough about the relationship between evacuation intentions and actions. Quick re- sponse preevent research is important and researchers must be ready at a moment’s notice to get into the field collecting data on decisions during an event as well as right afterward. Investment in systematic pre- and postevent data collection in hurricane prone areas will begin to fill the gaps in our understanding. This could be accomplished by conducting a series of pilot studies through- out the East and Gulf coasts. Findings from these types of re- search are likely to assist in effectively developing better ways to motivate evacuation compliance as well as shape and enhance the development of educational programs and media guide lines.

References

Baker, E. J. (1979). “Predicting response to hurricane warnings: A re- analysis of data from four studies.” Mass Emerg., 4(1), 9–24.

Baker, E. J., ed. (1980). “Hurricanes and coastal storms.” College Rep.

No., 33, Florida Sea Grant, Gainesville, Fla.

Baker, E. J. (1991). “Evacuation behavior in hurricanes.” Int. J. MassEmerg. Disasters, 9(2), 287–310.

Bateman, J. M., and Edwards, B. (2002). “Gender and evacuation: Acloser look at why women are more likely to evacuate for hurricanes.” Nat. Hazards Rev., 3(3), 107–117.

Benight, C., Gruntfest, E., and Sparks, K. (2004). “Colorado wildfires 2002.” Quick Response Research Rep. No. 167, Natural Hazards Cen- ter, Univ. of Colorado, Boulder, Colo.

Bolin, R. (1986). “The 1986 California floods.” Quick Response ResearchRep. No. 02, Univ. of Colorado, Institute for Behavioral Research,Boulder, Colo.

Bolin, R., Jackson, M., and Crist, A. (1996). “Gender inequality, vulner- ability, and disasters: Theoretical and empirical considerations.” The gendered terrain of disasters, E. Enarson and B. H. Morrow, eds., Praeger, Westport, Conn.

Brinkley, D. (2006). The great deluge: Hurricane Katrina, New Orleans,

and the Mississippi Gulf Coast, William Morrow, New York.

Burnett, K. et al. (1997). “Measurement of perceived disruption during rebuilding following Hurricane Andrew.” J. Trauma Stress, 10(4), 673–681.

Burton, I., Kates, R., and White, G. F. (1978). The environment as hazard,Guildford, New York.

NATURAL HAZARDS REVIEW © ASCE / AUGUST 2007 / 75

Cameron, L. D. (2003). “Anxiety, cognition, and responses to health threats.” The self-regulation of health and illness behaviour, L. D. Cameron and H. Leventhal, eds., Routledge, New York.

Carter, M. T., Kendall, S., and Clark, J. P. (1983). “Household response to

warnings.” Int. J. Mass Emerg. Disasters, 1(1), 95–104.

Clarke, L., and Short, J. (1993). “Social organization and risk: Somecurrent controversies.” Annu. Rev. Sociol., 19, 375–399.

Cross, J. (1979). “The association between previous residence and hurri- cane hazard perception and adjustments.” Paper presented at the 75th Annual Meeting of the Association of American Geographers, Philadelphia.

Dash, N. (2002). “Decision-making under extreme uncertainty: Rethink-

ing hazard related perceptions and action.” Ph.D. dissertation, FloridaInternational Univ., North Miami, Fla.

Dash, N., and Morrow, B. H. (2000). “Return delays and evacuationorder compliance: The case of Hurricane Georges and the FloridaKeys.” Environ. Haz., 2(3), 119–128.

Dow, K., and Cutter, S. L. (1998). “Crying wolf: Repeat responses tohurricane evacuation orders.” Coastal Manage., 26(4), 237–252.

Dow, K., and Cutter, S. L. (2000). “Public orders and personal opinions:Household strategies for hurricane risk assessment.” Environ. Haz.,2(4), 143–155.

Dow, K., and Cutter, S. L. (2002). “Emerging hurricane evacuation is-sues: Hurricane Floyd and South Carolina.” Nat. Hazards Rev., 3(1),12–18.

Drabek, T. E. (1986). Human systems responses to disaster: An inventoryof sociological findings, Springer, New York.

Drabek, T. E. (1999). “Disaster-induced employee evacuation.” Program

on Environment and Behavior, Monograph No. 60, Institute of Behav- ioral Science, Univ. of Colorado, Boulder, Colo.

Drabek, T. E., and Boggs, K. (1968). “Families in disaster: Reactions and relatives.” J. Marriage Fam., 30(3), 443–451.

Fischer, H. W., Stine, G. F., Stoker, B. L., Trowbridge, M. L., and Drain,

36. M. (1995). “Evacuation behavior: Why do some evacuate, while others do not? A case study of the Ephrata, Pennsylvania (USA) evacuation.” Disaster Preven. Manage., 4(4), 30–36.

Fothergill, A. (1996). “Gender, risk and disaster.” Int. J. Mass Emerg.

Disasters, 14(1), 33–56.

Giarratano, J., and Riley, G. (1994). Expert systems: Principles and pro-gramming, 2nd Ed., PWS, Boston.

Gladwin, C., Gladwin, H., and Peacock, W. G. (2001). “Modeling hurri-

cane evacuation decisions with ethnographic methods.” Int. J. MassEmerg. Disasters, 19(2), 117–143.

Gladwin, H., Lazo, J., Morrow, B. H., Peacock, W. G., and Willoughby,H. E. (2005). “Social science research needs for the hurricane forecastand warning system.” Hurricane Forecast and Warning System SocialScience Research Agenda, (http://www.sip.ucar.edu/hurricane.jsp).

Gladwin, H., and Peacock, W. G. (1997). “Warning and evacuation: A

night for hard houses.” Hurricane Andrew: Gender, ethnicity and thesociology of disasters, B. H. Morrow and H. Gladwin, eds., Rout-ledge, New York, 52–74.

Gruntfest, E., Downing, T., and White, G. F. (1978). “Big Thompson Flood.” Working Paper No. 32, Institute of Behavioral Science, Univ. of Colorado, Boulder, Colo.

Hutchins, E. (1995). Cognition in the wild, MIT Press, Cambridge, Mass.

Hutton, J. (1976). “The differential distribution of death in disaster: A test of theoretical propositions.” Mass Emerg., 1(4), 261–266.

Jaeger, C. C., Renn, O., Rosa, E. A., and Webler, T. (2001). Risk, uncer- tainty, and rational action, Earthscan, London.

Kasperson, R. E., Ortwin, P., Slovic, P., Brown, H. S., Emel, J., Goble, R.,

Kasperson, J. X., and Ratick, S. (1988). “The social amplification ofrisk: A conceptual framework.” Risk Anal., 8(2), 177–204.

Knabb, R. D., Rhome, J. R., and Brown, D. P. (2005). “Tropical cyclonereport: Hurricane Katrina, 23–30 August, 2005.” Rep. of the NationalOceanic and Atmospheric Administration National Hurricane Center,(http://www.nhc.noaa.gov/pdf/TCR-AL122005_Katrina.pdf).

Kunreuther, H., and Linnerooth, J. (1984). “Low probability accidents.”

Risk Anal., 4(2), 143–152.

Lindell, M. K., Lu, J. C., and Prater, C. S. (2005). “Household decision- making and evacuation in response to Hurricane Lili.” Nat. Hazards Rev., 6(4), 171–179.

Lindell, M. K., and Perry, R. W. (1987). “Warning mechanisms in emer-gency response systems.” Int. J. Mass Emerg. Disasters, 5(2), 137–153.

Lindell, M. K., and Perry, R. W. (2004). Communicating environmentalrisk in multiethnic communities, Sage, Thousand Oaks, Calif.

Lindell, M. K., and Prater, C. S. (2002). “Risk area residents’ perceptions

and adoption of seismic hazard adjustments.” J. Appl. Soc. Psychol.,32(11), 2377–2392.

Mileti, D., and Beck, E. M. (1975). “Communication in crisis: Explaining evacuation symbolically.” Communication Res., 2(1), 24–49.

Mileti, D., and Fitzpatrick, C. (1993). The great earthquake experiment: Risk communication and public action, Westview, Boulder, Colo.

Mileti, D., and O’Brien, P. (1992). “Warning during disaster: Normalizing

communicated risk.” Soc. Probl., 39(1), 40–55.

Mileti, D. S., Drabek, T. E., and Haas, E. (1975). “Human systems inextreme environments: A sociological perspective.” Program on Tech-nology, Environment and Man, Monograph No. 21, Institute of Be-havioral Science, Univ. of Colorado, Boulder, Colo.

Mileti, D. S., and Sorensen, J. H. (1987). “Hazards and precautionary

behavior.” Taking care: Understanding and encouraging self- protective behavior, D. Weinstein, ed., Cambridge University Press, Cambridge, U.K., 189–207.

Mileti, D. S., and Sorensen, J. H. (1990). “Communication of emergencypublic warnings: A social science perspective and the state-of-the-artassessment.” ORNL-6609, Dept. of Energy, Oak Ridge NationalLaboratory, Oak Ridge, Tenn.

Morrow, B. H., and Gladwin, H. (2005). “Hurricane Ivan behavioral analysis, 2004 hurricane assessments.” U.S. Army Corps of Engineers, September 2005, (http://chps.sam.usace.army.mil/USHESdata/ Assessments/2004Storms/2004_hurricane_season_page.htm).

Nigg, J. M. (1993). “Risk communication and warning systems.” Proc.,

Int. Conf. on Natural Risk and Civil Protection, Commission of Eu- ropean Communities, 209–236.

Perry, R. W. (1979). “Evacuation decision making in natural disasters.”

Mass Emerg., 4(1), 25–38.

Perry, R. W. (1994). “A model of evacuation compliance behavior.” Di- sasters, collective behavior, and social organization, R. R. Dynes and

1. J. Tierney, eds., University of Delaware Press, Newark, Del.

Perry, R. W., and Greene, M. R. (1982). “The role of ethnicity in theemergency decision-making process.” Sociol. Inq., 52(4), 306–334.

Perry, R. W., and Lindell, M. K. (1991). “The effects of ethnicity onevacuation decision-making.” Int. J. Mass Emerg. Disasters, 9(1),47–68.

Perry, R. W., Lindell, M. K., and Greene, M. R. (1981). Evacuationplanning in emergency management, Lexington Press, Lexington,Mass.

Perry, R. W., Lindell, M. K., and Greene, M. (1982). “Threat perceptionand public response to volcano hazards.” J. Soc. Psychol., 116(2),199–204.

Perry, R. W., and Mushkatel, A. H. (1986). Minority citizens in disasters, University of Georgia Press, Athens, Ga.

Renn, O., Burns, W. J., Kasperson, J. X., Kaspserson, R. E., and Slovic,

160. (1992). “The social amplification of risk: Theoretical foundations and empirical applications.” J. Soc. issues, 48(4), 137–160.

Rogers, G. O. (1982). “Social status and perceived risk: Some social processes and risk perception.” Paper prepared for the 10th World Congress of Sociology, Mexico City.

Schaffer, R., and Cook, E. (1972). Human response to Hurricane Celia,

Environmental Quality Program, Texas A&M Univ., College Station, Tex.

Schank, R. C., and Abelson, R. P. (1977). Scripts, plans, goals, andunderstanding: An inquiry into human knowledge structures,Lawrence Erlbaum, Mahwah, N.J.

76 / NATURAL HAZARDS REVIEW © ASCE / AUGUST 2007

Simpson, R. H., and Riehl, H. (1981). The hurricane and its impact, LSUPress, Baton Rouge, La.

Slovic, P. (1987). “Perception of risk.” Science, 236(17), 280–285. Slovic, P., Fischhoff, B., and Lichtenstein, S. (1984). “Behavioral deci-

sion theory perspectives on risk and safety.” Acta Psychol., 56(1–3),193–203.

Slovic, P., Kunreuther, H., and White, G. F. (1974). “Decision processes, rationality, and adjustment to natural hazards.” Natural hazards: Local, national and global, Oxford University Press, New York, 187– 204.

Sorensen, J. H. (1991). “When shall we leave: Factors affecting the tim-

ing of evacuation departures.” Int. J. Mass Emerg. Disasters, 9(2),153–164.

Sorensen, J. H. (2000). “Hazard warning systems: Review of 20 years ofprogress.” Nat. Hazards Rev., 1(2), 119–125.

Sorensen, J. H., Vogt, B. M., and Mileti, D. S. (1987). “Evacuation: Anassessment of Planning and Research.” ORNL-6376, Oak Ridge Na-tional Laboratory, Oak Ridge, Tenn.

Sorensen, J. H., and Vogt-Sorenson, B. (2006). “Community processes;Warning and evacuation.” Handbook of disaster research, H. Rod-riguez, E. L. Quarantelli, and R. R. Dynes, eds., Springer, New York,183–199.

Tierney, K. J. (1994). “Sociology’s unique contributions to the study of

risk.” Paper presented at the 13th World Congress of Sociology,Bielefeld, Germany.

Tierney, K. J., Lindell, M. K., and Perry, R. W. (2001). Facing the un-

expected: Disaster preparedness and response in the United States,Joseph Henry Press, Washington, D.C.

Turner, B. A. (1979). “The social aetiology of disasters.” Disasters, 3(1),51–59.

Tversky, A. (1972). “Elimination by aspects: A theory of choice.” Psy-chol. Rev., 79(4), 281–299.

Van Willigen, M., Edwards, T., Edwards, B., and Hessee, S. (2002). “Riding out the storm: Experiences of the physically disabled duringHurricanes Bonnie, Dennis, and Floyd.” Nat. Hazards Rev., 3(3),98–106.

White, G. (1994). “Paths to risk analysis.” Environmental risks and haz- ards, S. Cutter, ed., Prentice-Hall, Upper Saddle River, N.J., 69–74.

Whitehead, J. C., Edwards, B., Van Willigen, M., Maiolo, J. R., Wilson,

K., and Smith, K. T. (2000). “Heading for higher ground: Factors affecting real and hypothetical hurricane evacuation behavior.” Envi- ron. Haz., 2(4), 133–142.

Zhang, Y., Prater, C. S., and Lindell, M. K. (2004). “Risk area accuracyand evacuation from hurricane Bret.” Nat. Hazards Rev., 5(3),115–120.

NATURAL HAZARDS REVIEW © ASCE / AUGUST 2007 / 77

All in-text references underlined in blue are linked to publications on ResearchGate, letting you access and read them immediately.

doi:10.1111/disa.12026

The rise of politics and the decline of vulnerability as criteria in disaster decisions of the United States, 1953–2009

1. Steven Daniels, PhD Department of Public Policy and Administration, California State University, Bakersfield, United States

This paper examines the shift from vulnerability to political responsiveness in presidential and gubernatorial disaster decisions in the United States from 1953–2009 (President Dwight D. Eisenhower to President Barack Obama) using annual request, declaration, and approval data from multiple sources. It makes three key conclusions: first, the 1988 Stafford Act expanded federal coverage to all categories of disasters, added a significant range of individual types of assistance, and provided extensive funding for recovery planning. Second, the election effects on disaster decisions increased over time whereas the impact of social and economic vulnerability (measured by scope of disaster) declined. Third, the changes affected governors more than presi- dents, and the choices of governors drove those of presidents. The analysis underscores the increasingly political nature of the disaster decision-making process, as well as the difficulty in emphasising mitigation and preparedness as intensively as response and recovery. Proactive inter- vention yields fewer political rewards than responsiveness.

Keywords: disaster management, economic vulnerability, gubernatorial decision- making, political responsiveness, presidential decision-making, social vulnerability

Introduction

Earlier research on presidential decision-making (Daniels and Clark-Daniels, 2002; Daniels, 2009a, 2009b) concluded that presidents and governors face the competing goals of political responsiveness and comprehensive vulnerability management when seeking and granting federal disaster assistance to states and counties. It found that Presidents Gerald Ford (1974–77) and Jimmy Carter (1977–81) employed both cri- teria in executive disaster decisions, but that political responsiveness was already a critical factor in the 1970s. Many commentators have noted the mounting influence of political choices on escalating disaster costs in the 1990s and beyond (U.S. General Accounting Office, 1995, 1998; Sylves, 1996; Platt, 1999; Downton and Pielke, Jr.,

2001; Garrett and Sobel, 2003; Reeves, 2011; Sylves and Búzás, 2007).

This paper examines the lessening impact of social and economic vulnerability (as measured by scope of disaster) and the increasing impact of political responsiveness in presidential disaster decisions for the period from 1953–2009 (President Dwight D. Eisenhower to President Barack Obama). Scope of disaster in this context represents the size of the disaster (intensity) as well as economic (damages and government payouts) and social (deaths and injuries) effects. Whereas earlier analyses traced the

Disasters, 2013, 37(4): 669−694. © 2013 The Author(s). Disasters © Overseas Development Institute, 2013

Published by John Wiley & Sons Ltd, 9600 Garsington Road, Oxford, OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA

consequences of the competing influences of vulnerability, disaster scope, and respon- siveness for individual political choices, the current assessment focuses on the cumu- lative impact of those influences on the annual number of requests and declarations and the percentage of requests granted. One reason for the shift is the absence of Federal Emergency Management Agency (FEMA)1 information on presidential deci- sions to decline requests since 2006.2

The presidential disaster declaration process

An understanding of the decision model requires at least some knowledge of the process that presidents use to declare disasters. The decision-making process to deter- mine which political jurisdictions receive federal disaster assistance has evolved gradu- ally since the passage of the Federal Disaster Act (1950). Under the original legislation, states could only request assistance to repair public facilities (called, confusingly, public assistance) damaged by natural disasters. The federal government did not assist individuals until the Disaster Relief Act of 1974 created the Individual and Family Grant Program under FEMA’s Federal Disaster Assistance Program. The federal government responded only to natural disasters and occasional chemical and com- mercial disasters until the passage of the Robert T. Stafford Disaster Relief and Emergency Assistance Act (Stafford Act) in 1988, allowing it to assist in disasters of any cause.

Despite the gradual expansion of the scope of federal disaster assistance, the formal process has remained consistent (Settle, 1990; U.S. General Accounting Office, 1995; FEMA, 2008). When an incident or disaster occurs, state and local governments evaluate the damage and determine the capability of the local jurisdiction and the state to respond. If the state decides that the disaster exceeds the capability of the local and state governments to respond adequately, the governor contacts the regional office of the federal disaster agency (FEMA since 1979). The regional office then conducts a preliminary damage appraisal itself. In many cases, the local, state, and federal agencies assess the damage jointly. The governor formally requests assistance from the president, certifying that the disaster is beyond the capability of the state to respond and that the state has provided all required assistance and followed all necessary procedures. The regional office summarises the information collected during the preliminary assessment and forwards the material to headquarters. Based on the summary material from the regional office, the director of the federal disaster agency recommends an action to the President of the United States. The director of the federal disaster agency forwards a disaster packet combining the governor’s request, relevant political information about congressional and gubernatorial representation, and the agency’s recommendation. Depending on the administration, White House staff (usually domestic policy advisers and the White House Counsel’s office) may provide additional recommendations. The final decision to grant or deny aid rests with the president. Grants of assistance go out under the president’s signature. The FEMA administrator forwards denials. The governor may appeal any denial. The

two critical decisions that are at the core of the process are: (i) the governor’s decision to request assistance from the president; and (ii) the president’s decision to grant or decline the request.

In practice, for the vast majority of disaster requests (well in excess of 90 per cent), the actual process follows the formal process closely. Even in these cases, however, local, state, and federal agencies may conduct damage assessments separately or jointly. Catastrophic disasters frequently short-circuit the process, prompting declarations before all evidence is gathered. Disasters with predictable lead times (typically hurri- canes) may trigger declarations before the disaster actually occurs. Other disasters may unfold more slowly, prompting the gradual addition of more counties to the declara- tion and the provision of assistance first to public organisations, and later to individuals.

Vulnerability, responsiveness, and executive disaster decisions

The straightforward nature of the disaster declaration process disguises considerable complexity in the choices confronting individual governors and presidents contend- ing with unique disaster situations. Throughout the period of formal federal disaster assistance (1950–2012), governors and presidents faced a critical dilemma: disaster decisions face the inherent conflict between the evolving standard of comprehensive vulnerability management used by emergency management professionals and the pressures for political responsiveness exerted on elected officials. Emergency manage- ment professionals have embraced a comprehensive or integrated emergency man- agement model concentrating on mitigation, preparedness, response, and recovery (Mileti, 1999; Daniels and Clark-Daniels, 2002; McEntire et al., 2002; Schneider, 2002; Trim, 2004; McEntire, 2005). Under the model, the ultimate objective of dis- aster management is the development of ‘holistic and integrated activities directed toward the reduction of emergencies and disasters by diminishing risk and suscepti- bility and building of resistance and resilience’ (McEntire et al., 2002, p. 273; see also Mileti, 1999). This concept suggests that elected and institutional decision-makers should focus disaster decisions on the reduction of physical, social, or organisational liabilities (risk and susceptibility) and on increasing physical, social, or organisational capabilities to respond to disaster (resistance and resilience). ‘Vulnerability reduction requires long-term changes in regulatory policy aimed at reducing economic devel- opment in areas especially susceptible to catastrophes’ (Daniels and Clark-Daniels, 2002, p. 227). In addition, it requires rewarding jurisdictions for improving the resist- ance of physical infrastructure and the resilience of vulnerable human populations to potential catastrophes. In short, disaster policy should be proactive, anticipating and mitigating potential disasters.

By contrast, disasters have a number of political characteristics that make them

ideal arenas for political intervention and selective application of emergency manage- ment principles. First, disasters are focusing events. They combine objective criteria

(scope, severity, and visibility), extensive media coverage, symbolic characteristics (rarity, unpredictability, and victim vulnerability), and significant negative exter- nalities (Birkland, 1997, pp. 21–46, 2006, pp. 1–30; Kingdon, 2002; Schneider, 1995, pp. 14–18). This combination of impact, media attractiveness, symbolism, and sig- nificant spill-over effects makes catastrophic disasters prime opportunities for the exercise of gubernatorial and presidential leadership. Indeed, the public increasingly expects both the state and the federal government to act.

Second, governors, state legislatures, presidents, and congresses regard disaster policy as distributive (Lowi, 1964). Governors and presidents can treat disaster relief as a preferment to dispense for political motives that may bear little direct linkage to risk or susceptibility. Several authors have linked re-election to the granting of disaster recommendations (Downton and Pielke, Jr., 2001; Daniels and Clark-Daniels, 2002; Garrett and Sobel, 2003; Daniels, 2009a, 2009b; Reeves, 2011).

Furthermore, Members of Congress can see disaster declarations as ‘pork’, as ben- efits for their districts. Every disaster declaration packet forwarded to the president includes letters of support from every affected Member of Congress. While many of these expressions of support are routine, their absence can trigger the denial of an application for assistance (Eizenstat and Daft, 1980). In short, disaster policy frequently is reactive, responding to a disaster and providing leadership.

The political attractiveness of reactive spending versus proactive mitigation sug- gests that both the president and Congress derive significant political benefits from the ability to bestow preferments and lobby disaster declaration decisions directly. The effects of reaction are immediate and politically attractive; the effects of mitiga- tion are long term and heavily discounted. One important outcome has been the gradual nationalisation of the emergency management process (National Academy of Public Administration, 1993; Donahue and Joyce, 2001). This nationalisation has changed the decision dynamics for governors and state legislatures as well. State decision-makers routinely anticipate that the governor will ask the federal govern- ment for assistance with any disaster that has even a moderate impact on local and state resources. Many commentators have expressed concern about the ‘moral hazard’ generated by the current system of emergency management: the tendency of the system to reward jurisdictions for ignoring physical risk and social or organisational susceptibility to disaster (Platt, 1999, pp. xv–xviii, 37–41; O’Toole, 2005).

Previous research on disaster decisions

This process appears to have accelerated since the passage of the Stafford Act in 1988. Garrett and Sobel (1993) and Reeves (2011) have argued that the Act changed the dynamics of the disaster declaration process, making it even more political. Both sets of authors suggest that, since 1988, presidents have used disaster declarations to reward differentially states with close partisan balances and many Electoral College votes. Although no authors have discussed the governor’s role in this process, the consequences of disaster decisions in the presidential election process are not lost on state governors and may well influence the requests that they make.

arrGett and Sobel ( 1993) and Reeves (2011) based most of their conclusions on an analysis of declarations by the administrations of Presidents George H.W. Bush (1989–93) and Bill Clinton (1993–2001). Sylves and Búzás (2007) extended the analy- sis to the period from 1953–2003, and focused on the governor’s request as the unit of measurement. They found that both primary type of disaster (scope and vulnera- bility) and political factors (responsiveness) influenced the president’s decision to grant a governor’s request for a major disaster. Disasters involving hurricanes and typhoons, a Democratic president, a Republican governor, and an election year all increased the probability of a declaration. By contrast, human-caused disasters, a Republican president, a Democratic governor, and a non-election year all decreased the prob- ability of a declaration. Overall, Sylves and Búzás (2007) concluded that presidents granted disaster requests based on need more than on politics. Nevertheless, politics played a significant role.

Daniels (2009a, 2009b) extended the analysis to 2005 and included all gubernato- rial requests (including emergency requests). The research concluded that the passage of the Stafford Act boosted the propensity of subsequent presidents to grant indi- vidual disaster requests (a 14 percentage point rise) despite no significant amplification in the overall scope of disasters after 1989. Economic and social vulnerability influ- enced disaster choices indirectly by affecting the overall scope of the disaster. Both governors and presidents were sensitive to upcoming presidential elections, request- ing more assistance and granting it more often. Indeed, the probability of granting a major disaster request rose steadily in line with proximity to the election. Finally, even at the individual level, the president’s decision whether to grant a major disas- ter request is contingent on the governor’s decision to ask. An increase in requests generally produced an increase in major disaster declarations because the approval rates remained constant. Any hesitation in asking led to hesitation in authorising the request.

Multiple modes of vulnerability in disaster decisions Despite the importance of the dilemma confronting governors and presidents who must request assistance and pass judgement on it, it is important not to overstate the

importance of these decisions in the general scheme of executive decision-making.

For most governors and presidents, disaster requests represent largely second-tier, routine decisions. The pressure to establish a routine for disaster decisions is enormous. The pressure arises because of the large volume of requests: 2,642 major disaster and emergency requests between 1953 and 2005. The acceleration of requests (one every

18.1 days during the 1950s versus one every 4.8 days during the 2000s) has also gener- ated pressure to routinise the process. Finally, the fact that 77 per cent of requests are associated with windstorms and flood-related events makes the development of standard operating procedures attractive to both governors and presidents. Hence, administra- tions have developed the process of gubernatorial request and presidential considera- tion outlined above. During any week, therefore, disaster requests represent only one of literally hundreds of such decisions that presidents are required to make.

eveNrtheless, the status of disasters as focusing events and the increasing expec- tation that the state and the federal government will respond to local requests for assistance means that apparently routine executive choices can rapidly become cata- strophic failures of decision-making. Both President George H.W. Bush and President George W. Bush (2001–09) discovered to their detriment the political importance of maintaining an emergency management agency committed to effective response and recovery from catastrophic disasters (National Academy of Public Administration, 1993; Select Bipartisan Committee to Investigate the Preparation for and Response to Hurricane Katrina, 2006; Townsend, 2006). President Clinton learned this lesson well, appointing a FEMA director ( James Lee Witt) with extensive emergency man- agement experience and a close personal relationship with the president, and accord- ing the FEMA director Cabinet status. This status allowed President Clinton to respond more quickly when confronted with catastrophic disasters (Daniels and Clark-Daniels, 2001).3 However, it also boosted the likelihood that the president would grant disas- ter assistance under circumstances that might well be within the capabilities of the state or the local jurisdiction (Sylves, 2005; Sylves and Búzás, 2007). Both President Clinton and President George W. Bush set a then-record for disaster declarations

• during the election years of 1996 and President Obama (2009–present) broke that record in 2010 with 81 declarations (FEMA, 2011) and again in 2011 with 98 declarations. The dynamics outlined above created a set of incentives for guber- natorial and presidential decisions on disaster requests that greatly increased the number of requests submitted by the governor and granted by the president. The incentives also boosted the percentage of requests granted.

A vulnerability and responsiveness model of annual disaster decisions

The full presidential decision model of Daniels (2009a, 2009b) contained 17 inde- pendent variables, including: FEMA region; measures of social, demographic, and economic vulnerability; scope and type of disaster; measures of political responsive- ness; and decision cues by both the governor and the president. At key stages of the decision process, the explained variance ranged from 20–40 per cent. Many of these variables identify factors unique to a particular type of disaster, geographical or political jurisdiction, or decision cue (such as FEMA region, state-level demographic and economic variables, the party and ideology of the governor and congressional del- egation, the type of disaster, or the gubernatorial and presidential response time). Significant sources of variation in these variables disappear when disaster requests are aggregated from individual cases to annual totals. This aggregation suggests that the number of explanatory factors relevant for assessing annual disaster data is relatively small. The period over which the federal government has granted disaster assistance (57 years or data points through 2009) also supports the argument for a limited number of variables.

Key decisions in the disaster process

For each year, three outcomes arise from gubernatorial and presidential decisions:

• the number of requests that governors make; (ii) the number of declarations that presidents grant; and (iii) the percentage of requests that presidents The first outcome reflects the key choices confronting governors during a disaster event. The second reflects the president’s response to the governor’s choices and the available information gathered by the state and by FEMA. The third represents the president’s overall assessment of the merits of the governor’s requests.

External factors in the decision process

The time frame of federal disaster assistance reflects two distinct periods: prior to and after the passage of the Stafford Act. The initial period, from 1953–88, mirrors the gradual nationalisation of the system described earlier. The expansion from public assistance (repair and replacement of public facilities) to individual assistance under the Disaster Relief Acts of 1970 and 1974 could have produced significant variations but it did not appear to change the basic decision calculus of the request and declara- tion process (Daniels, 2009a, 2009b). Instead, the main break point during this period seems to be the widespread disasters of 1972 (Hurricane—later Tropical Storm— Agnes), 1973 (major flooding of the lower Mississippi River), and 1974 (Midwest Super Tornado outbreak). The geographical scope of these disasters generated unu- sually high numbers of requests across many states during these three years, produc- ing a spike in both requests and declarations between 1972 and 1974. In addition, 1977 saw a large number of requests for snow and ice emergencies. The 1972–74 disasters split the first period of disaster assistance into two sub-periods: 1953–71 and 1972–88.

The second period covers the passage of the Stafford Act of 1988, which expanded FEMA’s authority to all classes of disasters. Although the actual number of human- caused disasters requested and granted since 1989 has been very small, the Act appears to have changed the dynamics of federal disaster decisions. Several disaster manage- ment researchers have suggested that the Stafford Act altered the decision incen- tives of the intergovernmental emergency management system (Garrett and Sobel, 2003; Reeves, 2011). One might anticipate, then, that the results of those changes may affect the decision calculus of both governors and presidents. In particular, this author expected the following patterns of decision-making:

• Hypothesis 1: the passage of the Stafford Act increased requests, declarations, and approval rates for

Decision-making trends

Within each period, governors and presidents became aware, or FEMA and its pred- ecessors made them aware, of the pattern of past disaster declarations. When they made their requests, governors recognised the pattern of requests that presidents had

granted over the previous year or two. Presidents also were sensitive to the pattern of requests made by governors. In short, the process provided incentives for governors to adjust their request pattern and for presidents to modify their approval or denial pattern to reflect not only disaster type and scope but also the political actions taken by previous governors and presidents. Consequently, annual requests, declarations, and percentage approvals might well demonstrate trends over the three periods of federal disaster relief discussed above. Previous research (Daniels, 2009a, 2009b ) sug- gests the following hypotheses:

• Hypothesis 2a: the annual number of requests increased in the periods 1953–71

and 1989–2009, but decreased from 1972–88 (reflecting the peak during 1972–74).

• Hypothesis 2b: the annual number of declarations increased in the periods 1953– 71 and 1989–2009, but decreased from 1972–88.
• Hypothesis 2c: the approval rate remained stable from 1953–88 but increased after the passage of the Stafford

Scope of disaster

The comprehensive vulnerability management model suggests that presidents should respond to disasters by making choices that diminish ‘risk and susceptibility’ and build ‘resistance and resilience’ (McEntire et al., 2002, p. 273; see also Mileti, 1999). Ideally, the decrease in vulnerability and the increase in capability should drive the design of the mitigation, preparedness, response, and recovery process. In practice, the pressure for political responsiveness shifts the balance towards response and recovery and away from mitigation and preparedness (National Academy of Public Administration, 1993; Dymon and Platt, 1999; Donahue and Joyce, 2001; Downton and Pielke, Jr., 2001; McEntire et al., 2002; Garrett and Sobel, 2003; Sylves and Búzás, 2007; Daniels 2009a, 2009b; Reeves, 2011).

The president will receive more credit and attract less blame for helping vulner- able constituents than for preventing constituents from building or living where they please. Therefore, the best that supporters of comprehensive vulnerability manage- ment can hope for is that the president will react to the scope of the disaster (that is, the intensity, the number of deaths and injuries, the estimated economic damage, and the potential government bill) rather than the relative political power of the stake- holders. These expectations suggest the following hypotheses:

• Hypothesis 3a: increases in the overall magnitude of disasters in a year and in the average scope of disasters investigated by the governors will augment the number of requests by
• Hypothesis 3b: increases in the overall magnitude of disasters in a year and in the average scope of disasters investigated and requested by the governors will augment the number of declarations by
• Hypothesis 3c: increases in the overall magnitude of disasters in a year and in the average scope of disasters investigated and requested by the governors will augment the presidential approval

Political responsiveness

Although the effects of congressional and presidential ideology and governor’s party affiliation cancel out when the analysis focuses on annual figures, two political fac- tors remain relevant in predicting requests, declarations, and approval rates: (i) the president’s political party; and (ii) the occurrence of a request during a presidential election year. Sylves and Búzás (2007) concluded that Democratic presidents had a greater probability of granting a declaration between 1953 and 2003. Several authors (Garrett and Sobel, 2003; Sylves and Búzás, 2007; Daniels, 2009a, 2009b; Reeves, 2011) have demonstrated that the number of declarations rises during presidential election years. No research exists on the impact of presidential party and election year on governor’s requests, although the presence of a Democratic president has probably increased both gubernatorial requests and presidential approval rates because of the greater odds of a declaration under a Democratic president. This study anticipates that these phenomena will affect annual disaster decisions as well as individual-level decisions. This suggests the following hypotheses:

• Hypothesis 4a: the presence of a Democratic president increases the number of major disaster requests by
• Hypothesis 4b: the presence of a Democratic president increases the number of major disaster
• Hypothesis 4c: the presence of a Democratic president increases the approval rate of major disaster
• Hypothesis 5a: the number of major disaster requests increases during presiden- tial election
• Hypothesis 5b: the number of major disaster declarations increases during presi- dential election
• Hypothesis 5c: the approval rate of major disaster requests increases during pres- idential election

The effect of governor’s choices on presidential decisions

Finally, since the disaster decision process is a two-stage procedure requiring judge- ments by the state, territorial governor, or District of Columbia Mayor and the President of the United States, the governor’s initial request is likely to affect the president’s subsequent declarations. For annual totals, two competing models suggest themselves. The first assumes that the president reacts in the short term to guberna- torial requests. In this model the governor’s requests drive the president’s declarations. The more requests state governors make, the more declarations presidents approve. In this scenario, the president’s approval rate does not change. That is, as the number of requests mounts, the president responds by increasing the number of declarations while keeping approval rates constant. This implies the following hypotheses:

• Hypothesis 6a: the number of requests by the governor strongly influences the number of declarations by the

• Hypothesis 6b: the number of requests by the governor does not affect the per- centage of declarations that the president

The second model assumes that the president takes a longer-term view of the decision process (more proactive). The number of requests made each year by the governor exerts some pressure on the president to approve more disasters; however, presidents temper larger numbers of requests by granting a smaller percentage of them. This model assumes that the president attempts to mitigate the impact of larger numbers of requests by raising the stringency of approval standards. This model implies the following hypotheses:

• Alternative hypothesis 6a: the number of requests by the governor influences moderately the number of declarations by the
• Alternative hypothesis 6b: Increases in the number of requests from the gov- ernor decrease the approval rate by

Analysis

Data set

The data set for this analysis was composed of 57 cases, one for each year between 1953 and 2009. The dependent variables were the annual number of major requests made, the annual number of major disaster declarations granted, and the percent- age of major disasters approved by Presidents Dwight Eisenhower, John Kennedy, Lyndon Johnson, Richard Nixon, Gerald Ford, Jimmy Carter, Ronald Reagan, George H. Bush, Bill Clinton, George W. Bush, and Barack Obama between 1 January 1953 and 31 December 2009. The independent variables included a categorical vari- able for the three disaster periods (for declarations and requests), a dummy variable for pre- and post-Stafford Act (for approval rates), number of years, logged total US annual natural disaster damage, the average scope of a disaster for the governor’s dis- aster requests, the party of the president (Republican=+1; Democratic=-1), the pres- ence of a presidential election year (1=election year; 0=non-election year), and the annual number of major requests made (only for declarations and percentage approvals).

Data were gathered on independent and dependent variables for this period from multiple sources:

• FEMA annual major disaster declaration totals (FEMA, 2010).
• DARIS 2 runs on 17 July 1999 and 27 December 1999 (for requests), from FEMA and Freedom of Information Act requests in 2004 and 2006 (FEMA, 1999a, 1999b).
• Total annual natural disaster damage from EM-DAT (CRED, 2009).
• Information on the scope of a disaster (disaster intensity, damage estimates for indi- vidual disasters, number of deaths and injuries, and FEMA and predecessor agency payout amounts) from multiple sources:

• alifCornia disaster information from the California Governor’s Office of Emer- gency Services (2007);
  • Ohio disaster information from the Ohio Emergency Management Agency (2008);
  • Spatial Hazard Events and Losses Database for the United States (SHELDUS 0) (Hazards and Vulnerability Research Institute, University of South Carolina, 2012);
  • Hurricane data from the National Hurricane Center (2007; see also Pielke and Landsea, 1998; Blake et al., 2005);
  • Earthquake and volcano data from the S. Geological Survey (National Earth- quake Information Center, U.S. Geological Survey, 2006; Volcano Hazards Program, U.S. Geological Survey, 2007);
  • Climate and flood-related data from the ‘Storm Events’ database of the National Climatic Data Center (2007b);
  • Climate and flood-related data from monthly and annual issues of Storm Data

(National Climatic Data Center, 1953–2005);

• Disaster data gathered from the White House central files, the White House Office of Records Management, disaster files at the Lyndon Johnson Presidential Library, Nixon Materials Project, Gerald Ford Presidential Library, and Jimmy Carter Presidential Library, and central files at the Ronald Reagan Presidential Library and George Bush Presidential Library;
• Palmer Drought Severity Indices from Climate Data Online at the National Climatic Data Center (2007a);
• Flood data from the S. Geological Survey’s Publication Warehouse (Wells, 1959a, 1959b, 1962; Hendrix, 1964a, 1964b, 1964c, 1964d; Rostvedt, 1965a, 1965b, 1968, 1970, 1971, 1972a, 1972b; Rostvedt et al., 1968, 1970; Reid et al., 1975; Perry, Aldridge, and Ross, 2001); and
• FEMA and predecessor agency payout information and declared county infor- mation from the All about Presidential Disaster Declarations website of the Public Entity Risk Institute (2008).

Included are all major disaster requests for all political jurisdictions, including the 50 states, the District of Columbia, all territories, and all affiliated commonwealths and republics in the South Pacific and the Caribbean. Excluded are requests for emer- gency declarations, which have somewhat different decision dynamics, and fire sup- pression requests, which use different resources.

Some total damage estimates were missing from EM-DAT (CRED, 2009), for 1959, 1963, and 1988. In addition, because FEMA did not report information on disaster requests after 2005, the data set also was missing information on the number of requests and the scope of disaster requests from 2006–09. Of the 570 data points in the dataset (10 variables, 57 cases), less than two per cent of the information was missing, although four of the missing points include the missing disaster requests from 2006–09 noted above. Multiple imputation analysis with five analysis data sets served as replacements for the missing values. According to IBM SPSS Statistics (2012, p. 1):

heTpurpose of multiple imputation is to generate possible values for missing values, thus creating several ‘complete’ sets of data. Analytic procedures that work with mul- tiple imputation datasets produce output for each ‘complete’ dataset, plus pooled output that estimates what the results would have been if the original dataset had no missing values. These pooled results are generally more accurate than those provided by single imputation methods.

Measurement

The measurement of most variables was straightforward. One category of variables required special analysis: scope of disaster, for which two indicators were employed. The first was a measure of the overall annual scope of disaster. A gradual increase in the number, scope, and intensity of disasters of all types over a period of years would very likely increase the number of disaster requests by governors, the number of disasters declared by presidents, and the percentage of disasters approved by presidents. To assess this possibility, annual natural disaster damage for the US was gathered from EM-DAT, the international disaster database (CRED, 2009). The US dollar figures were adjusted for inflation (2006 dollars to make the data consistent with the other scope of disaster variable) and then the data was logged (to base 10) to account for extreme data values.

The second measure was the average scope of each disaster request by state gov- ernors. Just as the overall scope of disasters might boost requests, declarations, and approvals, increases in the typical scope of each disaster confronting a governor might well have a similar effect. A measure of the average annual scope of disaster was derived using information on the individual scope of disaster from FEMA’s Disaster Declarations database, 1953–2005 (Daniels, 2009c). Information was aggregated annu- ally for five indicators of scope of disaster. The first was intensity of the disaster on a five-point scale, using various disaster scales (the Fujita Tornado (0–5), Saffir-Simpson Hurricane (1–5), Palmer Drought Severity (0–5+), Northeast Snow Intensity (1–5), the earthquake moment magnitude scale (1–4.9=1, 5–5.9=2, 6–6.9=3, 7–7.9=4, 8+=5), and Volcanic Explosivity (0–5+) indexes) to categorise each disaster. If a flood affected more than one state, the number of states affected was assigned; otherwise, one (1) was assigned. Three (major) was assigned to the Watts riots (1965), the Rodney King riots (1992), and the first bombing of the World Trade Center in New York (1993), and five (catastrophic) to the 11 September 2001 attacks in New York and Virginia; otherwise, one (1) was assigned to all human-caused disasters. The remaining indi- cators were damage, deaths, injuries, and disaster agency payouts all logged to base 10 to reduce skewed distributions. A single factor was generated using principal compo- nents analysis with mean substitution and the variable was recoded to run from 0 to 100 per cent.

Statistical model

Preliminary analyses were performed on trends in requests, declarations, turn downs, and percentage approvals. The impact of scope of disaster on requests, declarations, and turn downs over time was examined. Finally, three analyses of covariance were

run for requests, declarations, and percentage approvals by year and separate regres- sions were conducted for each disaster period for each dependent variable. Table 1 reports the summary data; Table 2 reports the top 25 disasters by scope; and Tables 3–5 report the analyses of covariance for the entire period and broken down by dis- aster period.

Results

Trends in requests, declarations, and turn downs

Figure 1 shows the trends in major disaster requests, declarations, and turn downs from

1953–2009. It contains break points for 1972 and for the passage of the Stafford Act.

Several things are im-

Figure 1. Trends in annual federal disaster requests, declarations, and turn downs, 1953–2009*

Note: * Request and turn down figures for 2006–09 imputed.

Source: author.

Figure 2. Percentage approvals of major disaster requests, 1953–2009

Source: author.

mediately apparent. First, there were relatively low levels of requests, declara- tions, and turn downs in the period from 1953–71. In only one year during this period (1969) did the number of requests exceed 40; declarations never ex- ceeded 30. In the period from 1972–88, there were much higher initial num- bers of requests and decla- rations in 1972, 1973, and 1974, and there was much more volatility in all three variables, though the trend during the entire period was for requests, declara- tions, and turn downs to decline. The last year of the Reagan Administra- tion (1988) witnessed the lowest number of requests and declarations since the Eisenhower Administra- tion. After the passage of the Stafford Act in 1988, the number of requests and declarations increased sig- nificantly. By contrast, the

Table 1. Mean annual requests, declarations, and approvals for the three major disaster periods between 1953 and 2009

Disaster policy periods	Mean number of disaster requests per year	Mean number of disaster declarations per year	Major disasters as a percentage of requests
1953–71	25	17	66.8%
1972–88	48	30	62.7%
1989–2009	63	50	78.7%
1953–2009 overall mean	46	33	70.0%

absolute number of turn downs gradually declined. Patterns in the post-Stafford Act period suggest a substantial change in presidential decision-making.

These variations also manifest themselves in the trends in percentage approvals of major disaster requests (see Figure 2).

The period from 1953–88 saw a gradual decline in approval rates; however, this was a consequence of the high approval rates of the Eisenhower Administration from 1953–56. Ignoring the first term of the Eisenhower Administration, there is little in the way of an overall trend in approval rates. The most significant characteristic of annual approvals during this period was their extreme volatility. Approval rates varied from a low of 35 per cent to a high of 90 per cent. Following the passage of the Stafford Act, though, annual approval rates varied more narrowly, from 68 to 90 per cent. Variation during a presidential election year after the Stafford Act (1992, 1996, 2000, 2004, and 2008) was even narrower, at only 81–90 per cent. The most notable observation from the post-Stafford Act period is the gradual increase in approval rates and the decline in variation of approval rates in the period.

Table 1 summarises the changes over the period of federal disaster relief.

Despite the clear downward trends from 1972–88, the average number of requests and declarations increased from the first to the second period by 92 and 76 per cent, respectively. Approval rates declined slightly (67 to 63 per cent), but not significantly. The post-Stafford years witnessed a smaller percentage rise in both, but the number of declarations (67 per cent) increased more than the number of requests (31 per cent). From the early period to the Stafford Act era, average annual requests and average annual declarations increased by 152 per cent and 194 per cent, respectively. Approval rates rose by 14 percentage points between the pre- and post-Stafford eras.

The declining influence of scope of disaster on decision-making

State disaster requests with the greatest overall scope were identified using the FEMA Disaster Declarations database, 1953–2005 (Daniels, 2009c). These are presented in Table 2.

Twenty-two of the top 25 disaster requests in the US between 1953 and 2005 related to natural disasters. Hurricanes accounted for nine of the state requests, tornadoes for eight, earthquakes for four, and floods for one. The three human-caused disasters were

Table 2. The top 25 disasters by overall scope

Disaster	State	Incident date	Scope of disaster*	Disaster intensity	Number of deaths	Number of injuries	Total damage (uSD 2006)	FEMA payout (uSD 2006)
New York: terrorist attack	New York	11 September 2001	100.0%	Catastrophic (5)	2,751	6,000	92,951,090,743	9,919,056,786
Louisiana: Hurricane Katrina	Louisiana	29 August 2005	94.8%	Major (3)	1,577	7,508	53,681,848,632	19,102,000,000
Northridge earthquake	California	17 January 1994	81.8%	Major (3)	51	9,000	60,689,655,172	9,560,306,451
Hurricane Camille	Mississippi	17 August 1969	80.9%	Catastrophic (5)	133	4,733	2,772,760,870	411,370,725
Rapid City flood of 1972	South Dakota	9 June 1972	78.9%	Catastrophic (5)	237	2,932	797,758,454	111,907,010
Alaskan earthquake	Alaska	28 March 1964	78.6%	Catastrophic (5)	115	2,058	3,981,045,752	364,081,652
Mississippi: Hurricane Katrina	Mississippi	29August 2005	77.0%	Major (3)	238	104	25,883,945,169	9,210,471,197
Virginia: terrorist attack	Virginia	11 September 2001	76.7%	Catastrophic (5)	184	4,099	1,131,221,719	29,923,415
Loma Prieta earthquake	California	18 October 1989	76.2%	Major (3)	63	3,757	9,630,818,620	1,411,125,838
San Fernando earthquake	California	9 February 1971	73.4%	Major (3)	65	2,000	2,580,808,081	928,808,570
Hurricane Audrey	Louisiana	18 April 1957	73.1%	Devastating (4)	195	2,979	539,568,345	23,289,410
Hurricane Audrey	Texas	18 April 1957	72.3%	Devastating (4)	195	3,118	539,568,345	13,796,306
Florida: Tropical Storm Bonnie and Hurricane Charley	Florida	11 August 2004	71.2%	Devastating (4)	16	804	14,319,148,936	2,013,648,985
Hurricane Andrew	Florida	24 August 1992	71.1%	Catastrophic (5)	15	194	37,911,301,860	2,399,041,210
Mississippi Delta tornadoes of 1971	Mississippi	4 February 1971	71.1%	Catastrophic (5)	117	1,475	167,212,121	21,242,105
Hurricane Camille aftermath	Virginia	19 August 1969	71.0%	Major (3)	113	4,021	788,043,478	46,686,642
Super tornado outbreak of 1974	Kentucky	29 March 1974	71.0%	Catastrophic (5)	72	1,306	295,915,966	51,682,736
Super tornado outbreak of 1974	Ohio	1 April 1974	70.7%	Catastrophic (5)	42	1,396	1,077,974,790	75,501,736
Palm Sunday tornadoes of 1965	Indiana	11 April 1965	70.6%	Devastating (4)	137	1,796	4,987,660,256	9,167,611
Red River Valley tornado outbreak of 1979	Texas	10 April 1979	70.2%	Devastating (4)	54	1,812	795,337,143	100,429,097
Super tornado outbreak of 1974	Alabama	1 April 1974	70.1%	Catastrophic (5)	78	950	581,315,126	30,110,612
Hurricane Betsy	Louisiana	8 September 1965	70.1%	Major (3)	76	662	8,140,127,389	250,402,108
Super tornado outbreak of 1974	Indiana	1 April 1974	70.0%	Catastrophic (5)	47	832	4,015,126,050	46,787,089
Rodney King riots	California	29 April 1992	70.0%	Major (3)	53	2,383	1,154,401,154	209,655,429
Oklahoma tornado outbreak of 1999	Oklahoma	3 May 1999	69.8%	Catastrophic (5)	40	678	1,342,240,291	113,894,253

Note: * Scope of disaster is a single-factor scale of disaster intensity, logged number of deaths, logged number of injuries, logged damage, and logged FEMA payouts scaled to 100 per cent.

Figure 3. The diminishing impact of scope of disaster on decision-making

Source: author.

the Rodney King riots in California and the terror- ist attacks in New York and Virginia.

Despite the devastating nature of these events, the scope of the typical disas- ter request did not corre- spond with the intensity. For the period 1953–2005, the average scope of the typical disaster request was 33 per cent, meaning a dis- aster intensity of 1.7, dam- ages of USD 12 million, FEMA payouts of USD 6 million, one death, and

four injuries. This suggests that governors ask for federal assistance for disasters that, in most cases, are well within the capacity of state and local governments. These low numbers would not be so important if a clear difference in scope existed between declarations and turn downs; however, the proportions were 37 and 28 per cent, respectively—remember that 70 per cent of requests were declarations.

Although these distinctions are small, they might not matter if presidents were becoming more discriminating and proactive over time. Has the difference in scope between declarations and turn downs increased over the period of federal disaster assistance? The test of this hypothesis appears in Figure 3.

Contrary to expectations, the scope of disaster for declarations has decreased whereas the scope of disaster for turn downs has increased. In short, scope of disaster has become less important to both governors and presidents.

Explaining annual requests, declarations, and approval rates

Annual requests by governors

If the influence of scope of disaster has declined for both governors and presidents, what has happened with regard to the impact of political decisions? Using the model of decision-making outlined in the previous section, the study assessed the influence of the passage of the Stafford Act, decision-making trends, scope of disaster (across all disasters and individually), and political responsiveness (presidential party and elec- tion year) on the annual number of disaster requests made by state governors from 1953–2009. The results appear in Table 3, broken down by disaster era.

The overall equation (including the three disaster eras and interactions for all inde- pendent variables across eras) explained 87 per cent of the variance in the annual number of gubernatorial requests. For each of the three periods, the explained vari- ance is 54, 78, and 59 per cent, respectively.

Table 3. The effects of scope of disaster and political responsiveness on annual disaster requests, 1953–2009a

rviables	1953–71b		1972–88c		1989–2009d
Constant	-25.226		33.647		41.845
Trends
Number of years	0.725	***	-2.762	*****	1.548	*****
Economic vulnerability
Logged total US annual natural disaster damagese	4.783	***	5.587		-3.546
Average scope of annual major disaster requestsf	-0.040		1.156		-0.500
Political responsiveness
Party of the president (Republican=+1; Democratic=-1)	-0.559		-1.440		-0.872
Presidential election year	-0.320		-4.652		11.266	*
Period R-squared	0.542	**	0.775	****	0.587	**
Total R-squaredg	0.868	*****

Notes:

a Scope of disaster and disaster requests contained missing data for 2006–09. The coefficients represent pooled estimates from a five-analysis multiple imputation model.

b The period prior to the expansion of federal assistance to individuals.

c The first three years of the second period had unusually high declaration totals because of multiple- state disasters (Hurricane Agnes in 1972, Mississippi floods in 1973, and tornadoes in 1974).

d The Robert T. Stafford Disaster Relief and Emergency Assistance Act passed on 23 November 1988, expanding the rules for disaster assistance.

e Total annual natural disaster damages in the US reported to EM-DAT (logged).

f Factor scale combining five-category intensity and total logged damages, FEMA payouts, deaths, and injuries, rescaled to a 100 per cent scale.

g R-squared for full univariate analysis of covariance model allowing period interactions for all covariates.

* p<0.10; **p<0.05; ***p<0.01; ****p<0.005; *****p<0.001.

hreTe significant decision-making trends appear in the data:

• between 1953 and 1971, for each additional year, the number of requests increased by nearly three-quarters of one
• between 1972 and 1988, though, each additional year produced a decline of almost three requests per
• finally, the advent of the Stafford Act coincided with a rise of 5 requests per year by governors.

For the scope of disaster variables, a 10-fold variation in annual disaster damage between 1953 and 1971 generated nearly five additional requests per year. The effect jumped to nearly six additional requests per year between 1972 and 1988, although not significantly. However, the impact of scope of disaster largely disappeared and was

even reversed to some degree after 1988. The political variables proved irrelevant until after the passage of Stafford Act, when a presidential election year produced an aver- age of more than 11 additional disaster requests by governors.

Annual declarations by presidents

Once the governor has made his or her request, the president must then decide whether to grant a major disaster or to turn it down in some form. In addition to the original set of variables, the number of requests made by governors clearly will have a bearing on the president’s choices. The explanatory equations for annual declara- tions between 1953 and 2009 by disaster era are shown in Table 4.

Table 4. The effect of scope of disaster and political responsiveness on the annual number of disaster declarationsa

vriables	1953–71b		1972–88c		1989–2009d
Constant	-22.778		-26.302		-17.701
Trends
Number of years	-0.246		-0.316		0.033
Economic vulnerability
Logged total US annual natural disaster damagese	0.696		1.627		-0.057
Average scope of annual major disaster requestsf	0.505	**	0.611		0.362
Political responsiveness
Party of the president (Republican=+1; Democratic=-1)	-0.365		3.563	*	0.614
Presidential election year	3.829	*	2.611		4.692	**
Governor’s decisions
Governor’s requests for assistance	0.697	*****	0.533	**	0.835	*****
Period R-squared	0.770	****	0.751	**	0.965	*****
Total R-squaredg	0.957	*****

Notes:

a Scope of disaster and disaster requests contained missing data for 2006–09. The coefficients represent pooled estimates from a five-analysis multiple imputation model.

b The period prior to the expansion of federal assistance to individuals.

c The first three years of the second period had unusually high declaration totals because of multiple- state disasters (Hurricane Agnes in 1972, Mississippi floods in 1973, and tornadoes in 1974).

d The Robert T. Stafford Disaster Relief and Emergency Assistance Act passed on 23 November 1988, expanding the rules for disaster assistance.

e Total annual natural disaster damages in the US reported to EM-DAT (logged).

f Factor scale combining five-category intensity and total logged damages, FEMA payouts, deaths, and injuries, rescaled to a 100 per cent scale.

g R-squared for full univariate analysis of covariance model allowing period interactions for all covariates.

* p<0.10; **p<0.05; ***p<0.01; ****p<0.005; *****p<0.001.

heTfull analysis of covariance model explained 96 per cent of the variance in annual declarations. The explained variance for each of the periods was 77 per cent (1953–71), 75 per cent (1972–88), and 97 per cent (1989–2009). The yearly trends largely disappeared from the presidential declaration equation because presidents demonstrated no trends that were independent of the patterns in the governor’s original requests. As was the case with the governors’ requests, the effects of scope appeared to manifest themselves principally between 1953 and 1988: a two percent- age point increase in the scope of the disaster generated one additional declaration per year.

Whereas significantly higher request rates for governors occurred only during election years after 1989, the effects for presidents were spread across the entire period (although not significantly between 1972 and 1988). Elections led to four addi- tional declarations per election year between 1953 and 1971, three additional declara- tions per election year between 1972 and 1988, and five additional declarations per election year after 1989. In the latter period, this was in addition to the 11 additional gubernatorial requests. Given the 83.5 per cent approval rate of governors’ requests during this period, the total effect was more than 14 additional declarations. During the period 1972–88, Republican presidents granted roughly four more disaster requests per year. This was largely due to President Carter’s discontent with the disaster man- agement process (Eizenstat and Daft, 1980).

Finally, governors’ requests had a profound effect on presidential decisions. Between 1953 and 1971, each additional request had a 70 per cent probability of producing a declaration (all other things being equal). Between 1972 and 1988, the probability was 53 per cent. After 1989, it was 84 per cent.

Percentage approval by presidents

The model explained percentage approvals less well than annual requests and decla- rations. The explained variance for the full analysis of covariance model was 46 per cent, but it was only 32 per cent for the period from 1953–88 and 50 per cent after 1989. The approval rate declined by approximately one-half of a percentage point per year between 1953 and 1988, although this reflects the high approval rates between 1953 and 1956. The approval rate increased by one-eighth of a percentage point per year after 1989, but it was not significant. Between 1953 and 1988, a one percentage point increase in scope of disaster produced a 2.5 percentage point rise in approval rates. Consequently, moving from 40 per cent scope to 50 per cent scope would boost the approval rate by 25 per cent. This effect declined by about 75 per cent after 1989

and hence no longer had a significant impact.

Republicans were slightly more likely to approve disaster requests but not signifi- cantly so. Presidential elections were worth about nine percentage points throughout the entire period. Notably, the number of governor requests had no significant effect on approval rates during any period, suggesting that the approval rate remained con- stant regardless of the number of requests.

Table 5. The effect of scope of disaster and political responsiveness on disaster approval ratesa

vriables	1953–88b		1989–2009d
Constant	-33.534		54.669
Trends
Number of years	-0.468	*	0.126
Economic vulnerability
Logged total US annual natural disaster damagesd	-1.809		-0.879
Average scope of annual major disaster requestse	2.534	****	0.616
Political responsiveness
Party of the president (Republican=+1; Democratic=-1)	2.736		1.360
Presidential election year	10.857	*	8.093	**
Governor’s decisions
Governor’s requests for assistance	0.035		0.050
Period R-squared	0.318	*	0.496	*
Total R-squaredf	0.464	*****

Notes:

a Scope of disaster and disaster requests contained missing data for 2006–09. The coefficients represent pooled estimates from a five-analysis multiple imputation model.

b The period prior to the passage of the Stafford Act.

c The Robert T. Stafford Disaster Relief and Emergency Assistance Act passed on 23 November 1988, expanding the rules for disaster assistance.

d Total annual natural disaster damages in the US reported to EM-DAT (logged).

e Factor scale combining five-category intensity and total logged damages, FEMA payouts, deaths, and injuries, rescaled to a 100 per cent scale.

g R-squared for full univariate analysis of covariance model allowing period interactions for all covariates.

* p<0.10; **p<0.05; ***p<0.01; ****p<0.005; *****p<0.001.

Implications

The results highlighted several key issues concerning analysis of gubernatorial and presidential disaster decisions.

The Stafford Act

Nearly all of the evidence suggests that the Stafford Act changed profoundly the dynamics of both gubernatorial and presidential disaster decisions. The Stafford Act expanded federal coverage to all categories of disasters, added a significant range of individual types of assistance, and provided extensive funding for recovery planning. It appeared to alter the dynamics of the disaster decision-making process. Requests, declarations, and approval rates all increased. There was a dramatic decline in varia- tion of approval rates across years.

Economics and politics

The impact of election-year politics on disaster decisions increased over time. The effect of economic factors, especially vulnerability (as measured by scope of disaster), declined. Following the passage of the Stafford Act, governors made 11 additional requests per year during presidential election years. Presidents averaged between three and five additional declarations per election year throughout the period.

The effect of scope on requests vanished after 1989. In fact, it reversed, although the negative effect was not significant. Higher scope disasters had a lower probability of approval. The effect of scope on declarations fell by approximately 35 per cent; the effect of scope on approvals decreased by more than 75 per cent. In short, gover- nors and presidents exchanged politics for vulnerability as the dominant decision- making criterion.

Governors and presidents

The changes affected governors more than they affected presidents. First, at the col- lective level at least, governors shifted nearly entirely from scope of disaster as the basis for making a request to election politics as the foundation of choice. Governors also apparently developed sensitivity to the preferences of presidents, as suggested by the significant trends during all three disaster periods. Presidents, meanwhile, main- tained a somewhat more balanced set of criteria; nevertheless, scope of disaster became of less influence as election-year politics maintained its sway.

Second, governors’ collective choices had a significant, and perhaps controlling, effect on presidential choices. Presidential approval rates did rise significantly after the passsage of the Stafford Act; however, within each period, approval rates remained relatively constant, meaning that increases in requests by governors would drive increases in declarations by presidents. At the aggregate level, presidential response was almost exclusively reactive. The number of requests did not prompt presidents to alter their decision-making calculus.

Conclusion

This analysis covers all disaster requests, declarations, and approvals from 1953–2009. It highlights the increasingly expansive and political nature of the disaster decision- making process, as well as the difficulty that the emergency management system faces in emphasising mitigation and preparedness as intensively as response and recovery. The pressures that have driven the expansion of the federal role in disaster relief have made the adoption of a proactive focus a difficult long-term goal.

The most important finding of this project may well be the unpredictability of presidential disaster decisions. The deliberate vagueness of the various pieces of author- ising legislation and the enormous discretion that this legislation grants to the president reduce the likelihood of standard criteria for determining grants of federal assistance. This discretion means that presidents may award disaster assistance for whatever reasons

they deem to be demographically, economically, politically, or socially relevant. Even with regard to collective approval rates, 60 per cent of the variance remains unex- plained. The result appears to have been a gradual federalisation of disaster assistance as presidents find more reasons to grant more disasters. The probability of granting a disaster has risen gradually since the passage of the Stafford Act. Prior to 1988, approval rates for major disaster requests (as opposed to emergencies) averaged about 64 per cent. Since 1988, approval rates have averaged 79 per cent.4 This has occurred despite the fact that little has changed in relation to scope of disaster. Such a system prob- ably makes political jurisdictions more—rather than less—vulnerable because the decisions reward risky behaviour (Platt, 1999).

Correspondence

1. Steven Daniels, PhD, Department of Public Policy and Administration, California State University, 9001 Stockdale Highway, BDC 20, Bakersfield, CA 93311-1022, United States. Telephone: +1 661 654 2318; e-mail: rdaniels@csub.edu

Endnotes

¹ Throughout this paper ‘FEMA’ and ‘Federal Emergency Management Agency’ refer not only to the current agency but also to any of its predecessor organisations: Office of Defense Mobilization, Office of Defense and Civilian Mobilization, Office of Civil and Defense Mobilization, Office of Emergency Planning, Office of Emergency Preparedness, and Federal Disaster Assistance Admin- istration (National Archives, 2011).

² Personal communication with Kirby R. Rowe, FEMA–FOIA (Freedom of Information Act) Officer,

14 January 2010.

³ Staff members of the Clinton White House reported that James Lee Witt had more direct contact with the president, through Secretary of the Cabinet Thurgood Marshall, Jr., than any other member of the Cabinet. Interview with Kris M. Balderston, Special Assistant to the President for Cabinet Affairs, Deputy Assistant to the President, and Deputy Secretary to the Cabinet (1995–2001), Old Executive Office Building, Washington, DC, 17 July 1999.

⁴ The percentages reported earlier included both emergency and major disaster requests.

References

Birkland, T. (1997) After Disaster: Agenda Setting, Public Policy, and Focusing Events. Georgetown Uni- versity Press, Washington, DC.

Birkland, T. (2006) Lessons of Disaster: Policy Change after Catastrophic Events. Georgetown University Press, Washington, DC.

Blake, E.S., J.D. Jarrell, E.N. Rappaport, and C.W. Landsea (2005) ‘The deadliest, costliest, and most intense United States tropical cyclones from 1851 to 2004’. http://www.aoml.noaa.gov/hrd/ Landsea/deadly/index.html (accessed on 2 August 2013).

California Governor’s Office of Emergency Services (2007) Origins and Development: A Chronology,

1917–1999. California Governor’s Office of Emergency Services, Mather, CA.

CRED (Centre for Research on the Epidemiology of Disasters) (2009) ‘EM-DAT: The International Disaster Database’. CRED, Brussels. http://www.emdat.be (accessed on 2 March 2010).

Daniels, R.S. (2009a) ‘Vulnerability and responsiveness in U.S. disaster policy 1953–2005’. March. Annual Meeting Paper. Western Political Science Association, Vancouver, BC.

Daniels, R.S. (2009b) ‘Vulnerability and responsiveness in U.S. disaster policy, The impact of the Stafford Act’. April. Annual Meeting Paper. Midwest Political Science Association, Chicago, IL. Daniels, R.S. (2009c) ‘FEMA Disaster Declaration Database 1953–2005: damage estimates and disas- ter declaration decisions for Presidents Eisenhower through George W. Bush’. Machine readable

data file. R.S. Daniels (producer and distributor). California State University, Bakersfield, CA. Daniels, R.S. and C.L. Clark-Daniels (2001) ‘Transforming the Federal Emergency Management

Agency: the renewal and revitalization of FEMA’. In M.A. Abramson and P.R. Lawrence (eds.)

Transforming Organizations. Rowman and Littlefield, Lanham, MD. pp. 57–89.

Daniels, R.S. and C.L. Clark-Daniels (2002) ‘Vulnerability reduction and political responsiveness: explaining executive decisions in U.S. disaster policy during the Ford and Carter administrations’. International Journal of Mass Emergencies and Disasters. 20(2). pp. 225–253.

Donahue, A.K. and P.G. Joyce (2001) ‘A framework for analyzing emergency management with an application to federal budgeting’. Public Administration Review. 61(6). pp. 728–740.

Downton, M.W. and R.A. Pielke, Jr. (2001) ‘Discretion without accountability: politics, flood damage, and climate’. Natural Hazards Review. 2(4). pp. 157–166.

Dymon, U.J. and R.H. Platt (1999) ‘U.S. federal disaster declarations: a geographical analysis’. In

R.H. Platt (ed.) Disasters and Democracy: The Politics of Extreme Events. Island Press, Washington, DC. pp. 47–67.

Eizenstat, S. and L. Daft (1980) Recommended Turndown for Major Disaster Assistance – Montana. Memo- randum for the President, June 13, and Attachments. Carter Presidential Records, White House Central Files. Subject Files: Disasters, Box DI-11. Folder DI2/ST 26 1/20/77-1/20/81 Executive. Jimmy Carter Presidential Library, Atlanta, GA.

FEMA (Federal Emergency Management Agency) (1999a) ‘Declarations and turndowns (detail)’. DARIS 1.2 run, 7 July. FEMA, Washington, DC.

FEMA (1999b) ‘Federal Emergency Management Agency: designated areas’. DARIS 1.2 run,

27 December. FEMA, Washington, DC.

FEMA (2008) ‘The disaster process and disaster aid programs’. 24 March. http://www.fema.gov/ hazard/dproc.shtm (accessed on 24 March 2008).

FEMA (2010) ‘Annual major disaster declarations totals’. 8 March. http://www.fema.gov/news/ disaster_totals_annual.fema (accessed on 8 March 2010).

FEMA (2011) ‘2010 federal disaster declarations’. 20 April. http://www.fema.gov/news/disasters.fema?

year=2010 (accessed on 8 May 2011).

Garrett, T.A. and R.S. Sobel (2003) ‘The political economy of FEMA disaster payments’. Economic Inquiry. 41(3). pp. 496–509.

Hazards and Vulnerability Research Institute, University of South Carolina (2012) ‘Spatial Hazard Events and Losses Database for the United States (SHELDUS)’. http://webra.cas.sc.edu/hvri/ products/sheldus.aspx (accessed on 2 August 2013).

Hendrix, E.L. (1964a) Summary of Floods in the United States during 1956. Water Supply Papers: 1530. United States Government Printing Office, Washington, DC. http://pubs.usgs.gov/wsp/1530/ report.pdf (accessed on 15 March 2006).

Hendrix, E.L. (1964b) Summary of Floods in the United States during 1957. Water Supply Papers: 1652-C. United States Geological Survey, Washington, DC. http://pubs.usgs.gov/wsp/1652c/report.pdf (accessed on 15 March 2006).

Hendrix, E.L. (1964c) Summary of Floods in the United States during 1958. Water Supply Papers: 1660-B. United States Geological Survey, Washington, DC. http://pubs.usgs.gov/wsp/1660b/report.pdf (accessed on 15 March 2006).

Hendrix, E.L. (1964d) Summary of Floods in the United States during 1959. Water Supply Papers: 1750-B. United States Geological Survey, Washington, DC. http://pubs.usgs.gov/wsp/1750b/report.pdf (accessed on 15 March 2006).

IBM SPSS Statistics (2012) ‘Help – IBM SPSS Statistics: case studies: missing values option: multiple imputation’. IBM SPSS Statistics, Version 21. Release 21.0.0.0. IBM Corporation, Armonk, NY. Kingdon, J.W. (2002) Agendas, Alternatives, and Public Policies. Second edition. Longman Classics,

New York, NY.

Lowi, T. (1964) ‘American business, public policy, case studies, and political theory’. World Politics.

16(4). pp. 677–715.

McEntire, D.A. (2005) ‘Why vulnerability matters: exploring the merit of an inclusive disaster reduc- tion concept’. Disaster Prevention and Management. 14(2). pp. 206–222.

McEntire, D.A., C. Fuller, C.W. Johnson, and R. Weber (2002) ‘A comparison of disaster paradigms: the search for a holistic policy guide’. Public Administration Review. 62(3). pp. 267–281.

Mileti, D. (1999) Disaster by Design: A Reassessment of Natural Hazards in the United States. Joseph Henry Press, Washington, DC.

National Academy of Public Administration (1993) Coping with Catastrophe: Building an Emergency Management System to Meet People’s Needs in Natural and Manmade Disasters. National Academy of Public Administration, Washington, DC.

National Archives (2011) ‘Guide to federal records: records of the Federal Emergency Management Agency [FEMA]’. Record Group 311, 1955–89. http://www.archives.gov/research/guide-fed- records/groups/311.html (accessed on 7 December 2011).

National Climatic Data Center (1953–2005) Storm Data. http://www.ncdc.noaa.gov/oa/climate/ sd/#ABOUT (accessed on 2 August 2013).

National Climatic Data Center (2007a) ‘NNDC Climate Data Online’ http://lwf.ncdc.noaa.gov/ oa/climate/onlineprod/drought/xmgr.html (accessed on 8 August 2007).

National Climatic Data Center (2007b) ‘Storm Events’. http://www.ncdc.noaa.gov/stormevents/ (accessed on 2 August 2013).

National Earthquake Information Center, U.S. Geological Survey (2006) ‘National Earthquake Information Center – NEIC’. http://earthquake.usgs.gov/regional/neic/ (accessed on 12 April 2006).

National Hurricane Center (2007) ‘National data archive’. http://www.nhc.noaa.gov/pastall.shtml (accessed on 8 August 2007).

O’Toole, J. (2005) ‘U.S. help for disaster victims goes from nothing to billions’. 2 October. Pittsburgh Post-Gazette. http://www.post-gazette.com/stories/news/us/us-help-for-disaster-victims-goes- from-nothing-to-billions-603084/ (accessed on 7 August 2013).

Ohio Emergency Management Agency (2008) State of Ohio Disaster History (1964–2007) (Presidential Disaster Declarations). http://ema.ohio.gov/Documents/OhioMitigationPlan/Appendix_A_State_ of_Ohio_Disaster_History_Chart.pdf (accessed on 7 August 2013).

Perry, C.A., B.N. Aldridge, and H.C. Ross (2001) Summary of Significant Floods in the United States, Puerto Rico, and the Virgin Islands 1970 through 1989. Water Supply Papers: 2502. United States Govern- ment Printing Office, Washington, DC. http://pubs.usgs.gov/wsp/2502/report.pdf (accessed on 10 April 2006).

Pielke, R.A., Jr. and C.W. Landsea (1998) ‘Normalized hurricane damages in the United States:

1925–1995’. Weather and Forecasting. 13. pp. 621–631.

Platt, R.H. (1999) DisastersandDemocracy:The Politics of Extreme Natural Events. Island Press, Washington, DC. Public Entity Risk Institute (2008) ‘All about presidential disaster declarations’. http://www.

peripresdecusa.org (accessed on 21 March 2008).

Reeves, A. (2011) ‘Political disaster: unilateral powers, electoral incentives, and presidential disaster declarations’. Journal of Politics. 73. pp. 1154–1171.

Reid, J.K. et al. (1975) Summary of Floods in the United States during 1969. Water Supply Papers: 2030. United States Government Printing Office, Washington, DC. http://pubs.usgs.gov/wsp/2030/ report.pdf (accessed on 10 April 2006).

Rostvedt, J.O. (1965a) Summary of Floods in the United States during 1960. Water Supply Papers: 1790-B. United States Government Printing Office, Washington, DC. http://pubs.usgs.gov/wsp/1790b/ report.pdf (accessed on 15 March 2006).

Rostvedt, J.O. (1965b) Summary of Floods in the United States during 1961. Water Supply Papers: 1810. United States Geological Survey, Washington, DC. http://pubs.usgs.gov/wsp/1810/report.pdf (accessed on 15 March 2006).

Rostvedt, J.O. (1968) Summary of Floods in the United States during 1962. Water Supply Papers: 1820. United States Geological Survey, Washington, DC. http://pubs.usgs.gov/wsp/1820/report.pdf (accessed on 15 March 2006).

Rostvedt, J.O. (1970) Summary of Floods in the United States during 1964. Water Supply Papers: 1840-C. United States Government Printing Office, Washington, DC. http://pubs.usgs.gov/wsp/1840c/ report.pdf (accessed on 15 March 2006).

Rostvedt, J.O. (1971) Summary of Floods in the United States during 1966. Water Supply Papers: 1870-D. United States Geological Survey, Washington, DC. http://pubs.usgs.gov/wsp/1870d/report.pdf (accessed on 15 March 2006).

Rostvedt, J.O. (1972a) Summary of Floods in the United States during 1967. Water Supply Papers: 1880-C. United States Geological Survey, Washington, DC. http://pubs.usgs.gov/wsp/1880c/report.pdf/ (accessed on 15 March 2006).

Rostvedt, J.O. (1972b) Summary of Floods in the United States during 1968. Water Supply Papers: 1970-B. United States Government Printing Office, Washington, DC. http://pubs.usgs.gov/wsp/1970b/ report.pdf (accessed on 15 March 2006).

Rostvedt, J.O. et al. (1968) Summary of Floods in the United States during 1963. Water Supply Papers: 1830-B. United States Geological Survey, Washington, DC. http://pubs.usgs.gov/wsp/1830b/ report.pdf (accessed on 15 March 2006).

Rostvedt, J.O. et al. (1970) Summary of Floods in the United States during 1965. Water Supply Papers: 1850-E. United States Geological Survey, Washington, DC. http://pubs.usgs.gov/wsp/1850e/ report.pdf (accessed on 15 March 2006).

Schneider, R.O. (2002) ‘Hazard mitigation and sustainable community development’. Disaster Pre- vention and Management. 11(2). pp. 141–147.

Schneider, S.K. (1995) FlirtingwithDisaster:PublicManagementinCrisisSituations. M.E. Sharpe, Armonk, NY. Select Bipartisan Committee to Investigate the Preparation for and Response to Hurricane Katrina (2006) A Failure of Initiative: The Final Report of the Select Bipartisan Committee to Investigate the Preparation for and Response to Hurricane Katrina. United States Government Printing Office, Washington, DC. Settle, A.K. (1990) ‘Disaster assistance: securing presidential declarations’. In R.T. Sylves and W.L. Waugh, Jr. (eds.) Cities and Disasters: North American Studies in Emergency Management. Chapter 2.

Charles C. Thomas, Springfield, IL. pp. 33–56.

Sylves, R.T. (1996) The Politics and Administration of Presidential Disaster Declarations. Quick Response Report 86. Natural Hazards Research and Applications Information Center, Institute of Behavioral Science, University of Colorado, Boulder, CO.

Sylves, R.T. (2005) Presidents, Disasters and Policy. Public Entity Risk Institute, Fairfax, VA.

Sylves, R.T. and Z.I. Búzás (2007) ‘Presidential disaster declaration decisions 1953–2003: what influ- ences odds of approval?’. State and Local Government Review. 39(1). pp. 3–15.

Townsend, F.F. (2006) The Federal Response to Hurricane Katrina: Lessons Learned. February. http:// library.stmarytx.edu/acadlib/edocs/katrinawh.pdf (accessed on 25 March 2010).

Trim, P.R.J. (2004) ‘An integrative approach to disaster management and planning’. Disaster Prevention and Management. 13(3). pp. 218–225.

U.S. General Accounting Office (1995) Disaster Assistance: Information on Declarations for Urban andRural Areas. GAO/RCED-95-242. United States Government Printing Office, Washington, DC.

U.S. General Accounting Office (1998) Disaster Assistance: Information on Federal Costs and Approaches toReducingThem. GAO/T-RCED-98-139. United States Government Printing Office, Washington, DC.

Volcano Hazards Program, U.S. Geological Survey (2007) ‘Welcome to the Volcano Hazards Program (VHP) of the U.S. Geological Survey’. http://volcanoes.usgs.gov (accessed on 8 August 2007). Wells, J.V.B. (1959a) Summary of Floods in the United States during 1953. Water Supply Papers: 1320-E.

United States Government Printing Office, Washington, DC. http://pubs.usgs.gov/wsp/1320e/ report.pdf (accessed on 15 March 2006).

Wells, J.V.B. (1959b) Summary of Floods in the United States during 1954. Water Supply Papers: 1370-C. United States Government Printing Office, Washington, DC. http://pubs.usgs.gov/wsp/1370c/ report.pdf (accessed on 15 March 2006).

Wells, J.V.B. (1962) Summary of Floods in the United States during 1955. Water Supply Papers: 1455-B. United States Geological Survey, Washington, DC. http://pubs.usgs.gov/wsp/1455b/report.pdf (accessed on 15 March 2006).

doi:10.1111/disa.12059

Decision-making and evacuation planning for flood risk management in the Netherlands

Bas Kolen PhD, MsC Senior Consultant Disaster and Flood Risk Management, HKV Consultants, The Netherlands, and Ira Helsloot PhD Professor, Radboud University, The Netherlands

A traditional view of decision-making for evacuation planning is that, given an uncertain threat, there is a deterministic way of defining the best decision. In other words, there is a linear relation between threat, decision, and execution consequences. Alternatives and the impact of uncer- tainties are not taken into account. This study considers the ‘top strategic decision-making’ for mass evacuation owing to flooding in the Netherlands. It reveals that the top strategic decision- making process itself is probabilistic because of the decision-makers involved and their crisis managers (as advisers). The paper concludes that deterministic planning is not sufficient, and it recommends probabilistic planning that considers uncertainties in the decision-making process itself as well as other uncertainties, such as forecasts, citizens responses, and the capacity of infrastructure. This results in less optimistic, but more realistic, strategies and a need to pay atten- tion to alternative strategies.

Keywords: decision-making, evacuation, flood risk, the Netherlands

Introduction

Evacuation is a potential measure to reduce loss of life in a time of disaster or threat of disaster. People, animals, and goods that can be moved might be saved, but it can be costly in terms of time, money, and credibility (Bourque et al., 2006). Evacuation is a potential measure to address the risk of flooding. When a delay occurs, not every- one can reach the desired destination in time (Urbina and Wolshon, 2003; Barendregt et al., 2005; Jonkman, 2007; Kolen and Helsloot, 2012). The response to Hurricane Katrina in New Orleans, Louisiana, United States, in 2005 demonstrated that people and some movable items of property might be saved, but goods will be affected by flooding, and economic processes will come to a halt (Vrijling, 2009). The cost of an evacuation in the case of a hurricane in the US can exceed USD one million per mile of the coast because of commerce, productivity, and direct losses (Wolshon et al., 2005). With regard to credibility, this involves addressing concerns about the qual- ity and sources of information, the discrepancy between timely warnings and later but more accurate warnings (Dow and Cutter, 2000), and the impact of false alarms (Gruntfest and Carsell, 2000).

Floods often are described as the most deadly of all natural disasters (Alexander,

1993). The mortality rate for different kinds of flooding has been shown to be related

Disasters, 2014, 38(3): 610−635. © 2014 The Author(s). Disasters © Overseas Development Institute, 2014

Published by John Wiley & Sons Ltd, 9600 Garsington Road, Oxford, OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA

to the time available for the implementation of measures ( Jonkman and Kelman, 2005). Furthermore, its relationship with lead time is highlighted in research conducted after other kinds of disasters, such as earthquakes (see, for example, Alexander, 2004). This paper focuses on large-scale flooding caused by extreme water levels on rivers (rain, snow) and storm surges in tidal areas. The expected lead team for these events is relatively long compared to flash floods. The research shows that, owing to uncertainties in forecasting, different geotechnical mechanisms of failure of levees, and different hydraulic conditions, multiple events can happen. For example, extreme events as worst credible events (ten Brinke et al., 2010) can occur as can smaller events when the hydraulic load is much less. Furthermore, the lead time can vary between days and unforeseen events (Barendregt et al., 2005; Jonkman, 2007). By taking these uncertainties into account (as a part of reality), a raft of flood scenar- ios for emergency management can be defined that represents all possible situations

(Kolen et al., 2011).

Human interventions can reduce the consequences of a flood. When people leave the potentially exposed area or move to relatively safe places, such as shelters on high ground, they are less vulnerable to a flood. In addition, traffic management and mass communication campaigns can be implemented in the case of a threat to maximise the possibility of mass evacuation. These human interventions require decisions by citizens and authorities—all of which are based primarily on information about a possible threat, the expected consequences of the threat, and the anticipated outcomes of emergency measures. Authorities can influence the ‘physical environment’ that creates boundary conditions for a later response and increases the effectiveness of emer- gency measures. This paper uses the term ‘transition phase’ to characterise this period. The central issue for authorities (as well as for the general public) is how and when to respond to large-scale flooding. A previous evacuation with or without a flood, a missed call (Gruntfest and Carsell, 2000; Grothmann and Reusswig, 2006), or other circumstances in society (as depicted in the survey presented in this paper) might influ- ence future credibility and response and hence loss of life. However, other evidence suggests that this relation is less important (Baker, 1991; Lindell, Lu, and Prater, 2005). According to Kunreuther et al. (2002), this is because people fail to learn from the past. For low frequency events, as in the Netherlands, one can question whether learning is relevant for citizens because the length of the return period for such events might be far greater than a lifetime. For authorities, though, learning is relevant and explicit attention is needed as citizens in a democratic society expect public leaders to introduce responsible measures based on available information and knowledge in a crisis (Boin et al., 2005). Research shows that exercises and training to stimu- late the correct response have to be based on plausible scenarios (Alexander, 2000).

These scenarios and possible responses have to be recognised and accepted within the realm of planning.

The decision-making process for mass evacuation is influenced by great uncertain- ties and consequences. Research following Hurricane Katrina, for instance, assessed whether or not the earlier involvement of national organisations would have lessened

the impacts (Parker et al., 2009). More insight is needed, therefore, into the effective- ness of different strategies for mass evacuation and the mechanism governing how (and when) to decide on a specific strategy based on the probability of flooding, available time, the characteristics of different areas, and the effect of uncertainties (Kolen and Helsloot, 2012).

Objective and overview

Almost all of the literature (see below) focuses on the relationship between available time and required time for evacuation. This relation is used to define the moment at which to call for an evacuation; planning documents that support decision-making are based mainly on deterministic assumptions about a linear relation between threat/ decision/execution, which is applied to all possible events. Almost never addressed is the role of the decision-making process itself in strategy choice and the moment of decision-making and the impact on evacuation planning. Owing to the effect of ambiguity and current risk perception, a deterministic approach to decision-making processes might not result in minimisation of loss of life or social consequences.

This paper focuses on the impact of ambiguity and uncertainties in decision- making on mass evacuation strategies in the event of flood risk. This is called ‘top strategic decision-making’ and it centres on preventive or vertical evacuation and the moment at which to initiate such a strategy. The literature review below shows that evacuation planning for flooding concentrates on preventive evacuation because this is the best-case strategy. In a best-case strategy optimistic assumptions are used to define the threat and all operational measures. There is no attention to other evac- uation strategies even though they might be more effective in reducing potential loss of life. A better understanding of the role of the decision-maker and the crisis manager in the decision-making process in relation to evacuation strategy choices can result in more realistic evacuation planning and improve the effectiveness of evacu- ation. Consequently, this study concentrated on the:

• concept of top strategic decision-making and the transition phase;
• primary information for top strategic decision-making for evacuation;
• key factors in the decision-making process that affect strategy choices; and
• willingness to act according to the decision

The results are discussed within the context of evacuation planning.

The challenges for the Dutch decision-maker

The people of the Netherlands live in a delta that is largely below sea level. Historically, the country has concentrated principally on flood prevention, resulting in a flood defence system with the highest safety standards in the world. This paper examines flooding in the Netherlands caused by extreme discharges from rivers and/or storm

surges. Forecasting models and early warnings are used to alert crisis organisations and citizens to the need to implement protective measures. However, the time avail- able for evacuation based on forecast and failure mechanisms and the time needed for decision-making are uncertain (Barendregt et al., 2005). Flooding, as a natural hazard, can happen under many different scenarios (ten Brinke et al., 2010; Kolen et al., 2011).

Preparing for flood disasters on a national scale means preparing for extreme— but very unlikely—events (ten Brinke et al., 2010) that involve multiple decision- makers. Critical (forecasted) water levels in the Netherlands that could initiate a decision-making process are foreseen to occur less than once in a lifetime (Ministry of Transport, Public Works, and Water Management, 2008). Hence, the Dutch lack frequent experience of these events as well as other comparable threats with a short lead time in which to initiate a mass evacuation. Most relevant is the evacuation of some 250,000 persons in 1995 owing to the level of river waters. The main reason to select preventive evacuation in 1995 was the message from the water boards that they could not guarantee any longer the strength of the dikes (van Duin et al., 1995). In the days and hours before this announcement, there was a sense of urgency among the authorities and members of the general public because of the rising water levels in the rivers. Subsequently, discussions took place about the need for mandatory evacu- ation (Meurs, 1996).

An evaluation of the water safety policy in 2004 showed that the Netherlands is not prepared for extreme flooding (RIVM, 2004; see also ten Brinke, Bannink, and Ligtvoet, 2008a). As a result, the Government of the Netherlands (Ministry of the Interior and Kingdom Relations and Ministry of Transport, Public Works, and Water Management, 2005, 2006) sought to address the need for improved preparation. National (BZK, 2007; Kolen et al., 2007; Kolen, Vermeulen, and van Bokkum, 2008; LOCC, 2008; Ministry of Transport, Public Works, and Water Management, 2008; Wegh, 2008) and regional (Brabant, Veiligheidsregio Midden- en West- Brabant, 2008; Hulpverleningsregio Haaglanden, 2008; Veiligheidsregio Zeeland, 2008; TMO, 2009b) authorities prepared emergency plans for flood prevention and large-scale preventive evacuation. Drafts and first-generation plans were tested in a nationwide exercise entitled ‘Waterproef’ between 3 and 7 November 2008 (TMO, 2009b). The Government of the Netherlands (Ministry of the Interior and Kingdom Relations and Ministry of Transport, Public Works, and Water Management, 2009) concluded that, owing to this planning, research, and testing, the country was better prepared for flooding, although improvements still could be made, such as in evac- uation planning.

A complete preventive evacuation of large coastal areas generally is not possible. Road capacity is not appropriate for a preventive evacuation within a realistic early warning time frame. The provinces of North and South Holland, the most valuable part of the Netherlands in terms of economic processes, need the most time for preventive evacuation. Other provinces, such as Zeeland, require less time because they are less populated. In most circumstances, however, they still need more than

one day. Other evacuation strategies are highlighted in planning documents as pos- sible alternatives strategy, but they are not (yet) taken into account. A preventive evacuation owing to river flooding in polder areas takes less time than one for coastal flooding in highly populated areas, such as the province of South Holland, and fore- casts of water levels are more uncertain in the case of a storm surge than in relation to high discharges on rivers. In general, these areas can evacuate in time, although exceptions can manifest themselves because of unexpected events.

Earlier research by the authors (Kolen and Helsloot, 2012) underlined the need to consider three alternative strategies:

• Preventive evacuation: the organisation and movement of people from a poten- tially exposed area to a safe location outside of this area before the start of a
• Vertical evacuation: the organisation and movement of people inside the area under threat to shelters or safe havens before the start of a
• Shelter in place (or hide): the organisation and movement of people to the upper levels of residential buildings at the location before the start of a

In addition to preventive evacuation, safe havens, shelters, vertical evacuation, and support to augment the self-reliance of citizens should be taken into account in order to reduce loss of life and the impacts of evacuation (Haynes et al., 2009b; Kolen and Helsloot, 2012; Kolen et al., 2013). Decision-makers, though, have to make choices about evacuation strategy. These are only relevant when they can be implemented before people in the threatened area initiate an evacuation on their own. Research shows that an increase in the time available for evacuation is more critical for its effectiveness than further improvements in how organisations connect their plan- ning, how people behave during an evacuation, or how infrastructure is used during an evacuation (Kolen et al., 2013).

The challenge facing decision-makers is how to deal with the positive (reduction in loss of life in a flood) and negative (economic and social disruption) consequences of evacuation, which are related to the uncertainty of a flood and the time available. The size of a flood (for example, 20 per cent of the Netherlands might be flooded in one event), the magnitude of the evacuation, and the possible autonomous response of the public increase complexity.

To gain more insight into the decision-making process and the role of decision- makers and crisis managers during an evacuation, this paper evaluates the ‘Waterproef’ large-scale exercise of 2008. In addition, a survey of all 431 Dutch mayors and 95 international crisis managers who advise decision-makers was conducted between January and March 2010.

The study focused on the situation in the Netherlands for two key reasons:

• complete preventive evacuation is not always possible so there is a need to consider different evacuation strategies; and
• the magnitude of the threat in the Netherlands means that a large number of local and national decision-makers and crisis managers are

Dutch mayors of municipalities are responsible for emergency planning in their community and safety region. When their communities are under threat they have a role to play in the evacuation decision-making process, or they can be confronted by the consequences of evacuation in their community or safety region or the deci- sions of the national authorities.

The 95 crisis managers who participated in the survey all have a role in crisis centres and inform and advise decision-makers. The survey of crisis managers took place during the visiting programme of the international exercise ‘EU Floodex’ on 23–24 September 2009. All respondents were given some background information on mass evacuation in the Netherlands and a description of the different types of evacuation described in this paper. Furthermore, it was assumed that all respond- ents were aware of the possibility of flood risk (because of their involvement in the exercise and their profession).

The research endeavour aimed to gain more insight into the impact of decision- makers and crisis managers. Although one can question whether decision-makers in the Netherlands will respond to a crisis in the same way as they answered the ques- tions in the survey, the results are important for emergency planning since they clarify the way in which they think.

Experience and knowledge of decision-making for flood-related mass evacuation in the Netherlands is scarce. Flooding is a (very) low-frequency event in the coun- try, and the public perception is limited (Terpstra, 2009). However, authorities have paid increasing attention to the consequences of flooding since Hurricane Katrina in the US in August 2005 (ten Brinke, Bannink, and Ligtvoet, 2008a).

Future preparation for mass evacuation (such as emergency planning, exercises, and research) could challenge the results of the survey. Moreover, according to the literature, a lack of knowledge of heuristics and biases in dealing with uncertainties in these situations can influence the decision-making processes negatively (Tversky and Kahneman, 1974). Nevertheless, the survey serves to reveal the current percep- tions of decision-makers and crisis managers of mass evacuation in the Netherlands. For emergency planning, this range of perceptions is particularly important.

Literature review of different mass evacuation strategies Much attention has been paid to preventive evacuation in the event of a possible flood, yet almost none has been devoted to other strategies. In New Orleans, for

instance, ‘a complete evacuation of the city has been the cornerstone of hurricane preparedness planning for the region’ (Wolshon, 2006, p. 28). During Hurricane Katrina it was clear that not all citizens could, or wanted to, leave the area in time. Therefore ‘shelters of last resort’, such as the Ernest N. Morial Convention Center and the Mercedes-Benz Superdome, were opened (CNN, 2005). Post Katrina, New Orleans Mayor Ray Nagin declared that shelters of last resort would not be used again in the future (CNN, 2006). None were opened during Hurricane Gustav of August–September 2008. Those who did not leave the area in time arranged their own shelter, such as building on high areas (in the French Quarter).

Wolshon (2006, p. 28) describes hurricane-related evacuation in the US as an ini- tiative to ‘move people away from danger’, but he notes that it might not be possible to evacuate everyone preventively. Hence, there is an implicit need for other strategies. The need for strategies other than preventive evacuation has been addressed in contemporary literature. Following an analysis of 50-year flash-flooding in Australia, Haynes et al. (2009a, p. 9) concluded that: ‘in cases where evacuation may lead to increased exposure to danger and a suitable refuge exists for suitable occupants, shel- tering in place may be a better option’. In addition, they pointed out that: ‘At the moment, the literature cannot unequivocally support one option over another, in part due to the fact that because evacuation is such a well-established emergency management strategy, literature about policy alternatives is relatively thin on the ground. What the literature does show is that neither strategy is without risk and more research is needed to guide decision-making by emergency managers. In the end, emergency managers and the people directly at risk need to be able to assess the

relative risks of alternative strategies’.

Emergency planners in the Netherlands concentrate only on preventive evacu- ation (as illustrated by the Waterproef exercise) (TMO, 2009a). Jonkman (2007) mentions the possibility of vertical evacuation, but he focuses only on preventive evacuation. What is more, emergency planning for coastal areas of the provinces of North and South Holland assumes that coordination could solve the problems associ- ated with the limited capacity of the infrastructure and restricted lead time. The Coordination plan dikering area 14 states that: ‘as long as there is no national operation evacuation plan a large scale preventive mass evacuation seems not possible in case of coastal flooding’ (South-Holland, Province, et al., 2010, p. 24). Earlier research, however, reveals that, even in a perfect situation, a complete preventive evacuation is not possible (Barendregt et al., 2005; BZK, 2008b; Maaskant et al., 2009).

The success of an evacuation strategy depends on the relation between the time available and the time required to execute the strategy and emergency measures. Time available depends on a combination of availability of forecasts and their use by decision-makers and experts (van Zuilekom, van Maarseveen, and van der Doef, 2005; Jonkman, 2007; Kolen and Helsloot, 2012). Meanwhile, analyses show that the time required for an evacuation can be reduced by introducing several extra emergency measures, such as mass communication and traffic management. The time needed to evacuate the ‘Islands of Zealand and South Holland’ in the event of a possible storm surge, for example, can be decreased from a worst-case scenario of approximately 55 hours to some 25 hours. Although a complete preventive evacu- ation for the entire Dutch coast still remains impossible, national traffic management can lower the time needed to evacuate 50 per cent of the population by almost a full day (from 44 to 27 hours). A theoretical mathematical solution that optimises the use of roads and utilises the behaviour of people as a variable predicts that the time required can be reduced by up to 48 hours (Kolen and Helsloot, 2012). Furthermore, the contra-flow system of New Orleans shows that it is possible to decrease the time required for evacuation (DHS, 2006; Wolshon, 2006),

Other strategies, such as vertical evacuation, might result in less loss of life because shelters are relatively safe areas. By comparing lead and required time and deter- mining loss of life in a preventive and vertical evacuation, it has been shown that a vertical evacuation in the coastal areas of the provinces of North and South Holland will result in less loss of life than a preventive evacuation, except in the very opti- mistic (best) case of an exceptional (very uncertain) lead time and a perfect logistical operation (Kolen et al., 2013). For other less-populated areas, whether a preventive or a vertical evacuation results in minimum of loss of life depends on the time available and the implemented measures (Kolen and Helsloot, 2012).

Case study: ‘Waterproef’

Scope of the exercise

The Ministry of the Interior and Kingdom Relations, the Ministry of Transport, Public Works, and Water Management, and Taskforce Management Flooding organised ‘Waterproef’ between 3 and 7 November 2008. It was the first national exercise on flooding and mass evacuation and was held after a two-year programme of improve- ments for flood preparedness (TMO, 2009b). This study assesses the part of the exer- cise related to decision-making on evacuation during coastal flooding.1

Several national (crisis centres and Rijkswaterstaat) and regional (safety regions and water boards) organisations took part in the exercise. A public panel also partici- pated, providing feedback on the communications of the authorities and the media (de Jong and Helsloot, 2010). Waterproef focused on the choice of decision-making strategy for evacuation four days before an expected dike failure and possible flood. A scenario was developed that described a possible storm surge that could cause large- scale flooding along the coast of the Netherlands—a best-case scenario was used to reduce complexity and to present preventive evacuation of coastal areas as a serious option. The development of the threat and of all decisions between the first warning (eight days before the possible flood) and the day of the exercise was described in a start document based on existing emergency planning (Kolen, Vermeulen, and van Bokkum, 2008). Decisions concerning evacuation were foreseen four days before the expected moment of failure of levees. During the exercise regional decision- makers called for an evacuation, but this was not supported at the national level.

Even though an exercise is a constructed situation and is not developed for scientific research, and it is so large that it cannot be controlled (Helsloot, Scholtes, and Warners, 2010), lessons can be learned that are applicable to top strategic decision-making for mass evacuation by taking its circumstances into account. An evaluation of the exercise offers a unique (because of the presence of observers) view of national crisis management.

Lessons learned during preparations for Waterproef

Waterproef illustrates the difficulties that decision-makers face in dealing with uncer- tainties and in employing an integrated approach. Three alternatives for mass evac- uation presented to the national decision-makers sparked debate among them. This

resulted in a decision to implement a totally new strategy (not prepared in advance)— involving the evacuation of non-self-reliant individuals and the families of first responders; others had to wait—that delayed the preventive evacuation by at least a day. One can also question whether the decision made was realistic or even counterproductive. At the same time, regional crisis centres advocated complete preventive evacua- tion because they were not aware of the other options. In addition, the decision- makers of one region decided to call for an evacuation on their own (based on their own risk perceptions and responsibilities), even though they were aware of the national

decision-making process.

The combined impact of all decisions on an evacuation (in terms of a reduction in loss of life or risk) has not been defined in planning and was not addressed dur- ing the exercise. Consequently, they were not taken into account in the study. The decision-making process was dominated by perceptions and expectations regarding the effectiveness of an evacuation and the cooperation of others.

Waterproef also showed that the personal opinions of crisis managers vary strongly and are influenced by available information over time. Analysis of the development of top strategic decision-making, as recorded in the starting document, which was the input for the exercise, reveals substantial differences in when and how to inform others. Although this starting document (BZK and VenW, 2008) was based on avail- able emergency planning, several crisis managers still questioned the point at which to inform decision-makers and when to call for certain emergency measures. This produced several extreme opinions, such as ‘directly after the detection of a possible storm surge’ up to the ‘moment to call for a mass evacuation’. Arguments presented included ‘decision-makers are too busy and not willing to spend time’, ‘it is not serious enough’, ‘media pressure will force them to meet directly’, and ‘because of the lack of resources and the possible consequences’. The matter has a serious political dimension because the most logical move, ‘to wait and see’, can become a dramatic decision, as less time is available for evacuation, resulting in greater loss of life.

Top strategic decision-making: key factors and parameters

The concept and the transition phase

Top strategic decision-making for mass evacuation deals with (i) when to initiate an evacuation and (ii) the type of evacuation (preventive, vertical evacuation, use of shelters, and the creation of optimal (or better) circumstances for evacuation). The top strategic decision-maker is at the apex of the decision-making tree—the Ministerial Policy Team vis-à-vis flooding in the Netherlands. Top strategic decision-makers will be confronted automatically with many choices, great uncertainties, and myriad consequences in all circumstances. Uncertainties occur, for example, in predicting flooding (size and probability of occurrence) (ten Brinke et al., 2010; Kolen et al., 2011), the effectiveness of emergency measures (Kolen and Helsloot, 2012), and the responses of other stakeholders (local authorities, first responders, and citizens).

When forecasts become clearer and uncertainties decline (see Figure 1), people and decision-makers start to act. The autonomous response of citizens can lead to overload or the inefficient use of road capacity and available equipment and can place limitations on authorities in implementing further mitigating measures. Several models describe the possible responses of citizens to a natural hazard based on the interaction of environmental, individual, and social processes (Lindell and Perry, 1992; Sorensen, 2000; Grothmann and Reusswig, 2006; Kolen et al., 2013). It is known that not all people act directly after receiving a flood warning and that it takes time before people start to evacuate (Lindell et al., 2002). Since floods do not respect administrative boundaries multiple decision-makers are involved. The autono- mous response of these decision-makers can result in counterproductive measures as well as the less optimal use of available resources and infrastructure.

Emergency measures have to be implemented before the combined consequences of the autonomous responses of others (citizens, organisations) create boundary con- ditions for evacuation. The impact of these ‘top strategic decisions’ depends on the possibility of establishing circumstances that facilitate a future response by citizens and several stakeholders. Rasmussen, Brehmer, and Leplat (1991) describe this pro- cess as reflective decision-making: the decision has to be made in relation to the decisions of others. These top strategic decisions have to be made based on informa- tion on forecasts and scenarios for evacuation and before people start to act. These decisions involve a transition from normal life to a mass evacuation mode. This period is called the transition phase (see Figure 1). Future decisions are made during an operation in the context of the evacuation mode.

During the top strategic decision-making phase, one can already speak of a ‘crisis’. This is defined as the moment when policymakers experience ‘a serious threat to the basic structures or the fundamental values and norms of a system, which under time pressure and highly uncertain circumstance necessitates making vital decisions’ (Rosenthal, Charles, and Hart, 1989, p. 10). Preparations for a future mass evacuation have the objective of transporting as many people, animals, and movable goods to the

Figure 1. Conceptual illustration of the transition phase for top strategic decision-making

Source: authors.

safest possible place before the event (in this case, a flood). In addition to flooding, which is considered to be a national crisis in the Netherlands (Helsloot and Scholtens, 2007), and hurricanes in the US (Cole, 2008; Parker et al., 2009), mass evacuation itself also has to be seen as a crisis.

To reduce the consequences of a possible flood, decision-makers can opt to initiate another crisis: the evacuation itself. An example is a ‘shadow evacuation’ (a non- authorised evacuation), as seen in the US during a chlorine spill in Graniteville, South Carolina (Mitchell, Cutter, and Edmonds, 2007), and during Hurricane Rita in Houston, Texas (DHS, 2006), in 2005.

The transition phase reveals the role of top strategic decision-making for mass evacuation: to create the conditions needed for the taking of other decisions and for the implementation of emergency measures in the near future by authorities, emer- gency services, and citizens. The following are examples of top strategic decisions with regard to communication policy and operational measures:

• Communicate with the public about the risk, the impacts, possible emergency measures, and
• Take policy decisions to influence other authorities. Warn the relevant national and regional authorities (if not warned already). Define the go/no-go decision and strategy for evacuation (preventive, vertical, shelter in place, or a combination). Inform other authorities about the risks and consequences (and timelines) of the threat, as well as about possible emergency measures, the impact of uncertainties, how to call for assistance, juridical arrangements, and international
• Implement operational emergency measures to adapt the environment, including implementing national traffic management, identifying the availability of routes, assigning regions that will offer public shelters, and prioritising the use of limited available (national)

For evacuation planning, authorities in surrounding areas have an important part to play in supporting evacuation operations, such as traffic management, providing shelter, and delivering equipment and services (Wolshon, 2006; Ministry of Transport, Public Works, and Water Management, 2008; Wegh, 2008). Emergency planning (BZK, 2008a; Ministry of Transport, Public Works, and Water Management, 2008, 2009), research post evacuation (Jonkman, 2007; Kolen and Helsloot, 2012), experience of Hurricane Katrina (Parker et al., 2009), and exercises (TMO, 2009a) in the Netherlands indicate that proactive and direct involvement at the national level is necessary to increase the effectiveness of emergency measures following a national disaster.

The mass evacuation of Rivierenland in 1995 and the response to Hurricane Katrina in 2005 underscore the importance of and the difficulties associated with top stra- tegic decision-making, including involving relevant partners in time and when to call for a preventive evacuation. During Katrina, some people did not want to evac- uate because they hoped or they assumed that, as in the past, the hurricane would not hit their area (Parker et al., 2009). Earlier involvement at the national level, such as by the Federal Emergency Management Agency or the Red Cross, might have reduced

some of the consequences (Parker et al., 2009). The top strategic decision to involve the national level early in creating better conditions for a response was a lesson learned in New Orleans, as well as in the Netherlands (van Duin et al., 1995). During Hurricane Gustav of 2008, the national level made a concerted effort to be on top of the situ- ation and to show its concern (Cole, 2008). Although this is not clear evidence of an increase in the effectiveness of top strategic decision-making, such action surely affects the perceptions of professionals and members of the general public. Another key factor that contributed to the Gustav response was recent experience of Katrina. When time is limited, other strategies for mass evacuation, such as vertical evacu- ation and implementation of national traffic management (if implemented in time), might be more attractive (van Noortwijk and Barendregt, 2004; Wolshon, 2006; Jonkman, 2007). Whether to implement them is down to the decision-makers involved.

Primary information for top strategic decision-making for evacuation

The literature shows that, when relevant stakeholders have contact with one another, an optimal decision-making process can be implemented in which the right people work on the right objective at the right moment with the right information (Aldunate, Pena-Mora, and Robinson, 2005). One can question whether or not information can or will ever be completely available. At the moment the information is analysed, new information is, by definition, available because of the ongoing nature of the disaster or threat. It is also impossible to know whether all relevant information is available when taking the number of stakeholders (such as citizens, crisis centres, and first responders) into account.

In a Western society, some tasks of government are spread across several national and regional (semi-) governmental organisations; others are privatised. In normal day-to-day life, these organisations implement their own emergency measures based on their policies. The theory of ‘Distributed Decision Making’—defined as the design and coordination of connected decisions (Schneeweiss, 2003)—describes the optimi- sation of multiple decisions in a situation involving multiple interests of organisations. The theory assumes that society is differentiated in such a way that a central body cannot control it via a hierarchical relation. The theory becomes more relevant when more stakeholders make decisions. Thus, decision-makers should take ‘other decisions’ into account so they do not frustrate other decisions.

In the case of the threat of flooding, time is available to share information and to discuss the decision-making process. Therefore, the theory of ‘Natural Decision Making’, which describes how people act in a disaster (Fjellman, 1976), applies only to top strategic decision-makers and their crisis managers. First responders are not confronted immediately with the need to act: during this phase, the flood has not occurred yet and evacuation has still to commence. Aside from top strategic decision- makers, no one else faces the consequences directly, so they are not required to make any immediate decisions. This might result in calls for further information gathering and in a delay in decision-making. Critical moments could pass, meaning that some emergency measures, such as a preventive evacuation, can no longer be introduced.

A major flood event, involving all stakeholders, as noted above, is a national crisis in the Netherlands (Helsloot and Scholtens, 2007; BZK, 2008c). In the best case, stakeholders in the dynamic organisation are completely aware of the available infor- mation and execute the (centrally) chosen strategy perfectly. However, decision- makers have to deal with imperfect and uncertain information and are confronted by the decisions of others. The key question is: what kind of information do they need during the transition phase? Another fundamental matter concerns the prioritising of sources of information (see Table 1).

The results show that the survey participants (mayors and crisis managers) all tend to prefer a risk-based approach, according most value to the probability, impact, and effectiveness of possible strategies. In addition, all placed great importance on expert advice. Public pressure and the economic and social consequences of decisions were considered to be less important. This means that decision-makers tend towards a rational approach and rely on experts for advice to support decision-making.

In a potential mass evacuation, however, one can conclude that the capacities of the emergency services are far outweighed by the population that needs to be served. Evacuation might reduce loss of life and the cost of lost movable goods, but it cannot decrease the expected damage to fixed goods, such as agricultural land and houses. Hence, authorities and emergency services have to prioritise and deal with limited resources.

Table 1. ‘Determine the importance of each item to the decision-making process about mass evacuation on a scale of 1 to 5 (1 = no importance, 2 = less important, 3 = important 4 = very important 5 = most important)’

Parameter	Decision-makers		Crisis managers
Response rate of decision-makers: 38% Response rate of crisis managers: 59%	Expected value	Standard deviation	Expected value	Standard deviation
Probability of flooding	4.2	0.8	3.9	0.8
Size of the threatened area	3.5	0.8	3.6	0.8
Time available until failure of defence system	4.0	0.7	4.1	0.7
Public pressure	2.9	0.6	2.9	0.7
Effectiveness of a strategy	3.8	0.8	3.8	0.8
Economic impact of an unnecessary evacuation	3.0	0.7	2.9	0.8
Social impact of an unnecessary evacuation	3.3	0.7	2.9	0.8
Accountability for decisions made	3.4	0.8	3.2	0.8
Required leadership	3.5	0.8	4.0	0.8
Expert advice	3.9	0.7	3.8	0.7

Hindsight permits an examination of the best decisions in a flood event or in a situ- ation when a flood did not occur. The aftermath also influences public opinion on the response of decision-makers. Thus the second question in the survey focused on the parameters expected to be most important following an evacuation when a flood did or did not occur (see Table 2).

The results show that a reduction in loss of life is seen as a more important param- eter than a decrease in damage. Evacuation happens more frequently than flooding in the Netherlands (HKV Consultants, 2010). In the case of an evacuation that is not followed by a flood event, more attention is paid to accountability of decision-makers and cooperation between authorities than when the flood does happen, although pre- vention of loss of life remains important (instead of prevention of damage).

Table 2. ‘What are the 3 factors that contribute most to whether an evacuation decision was “right” in (1) a situation after a flood and (2) after a false alarm?’

Parameter	Decision-makers				Crisis managers
Response rate of decision-makers: 23% Responserate of crisis managers: 48%	Contributionto top three in case of a flood	Contributionto top three in case of a false alarm (no flood)	Contributionto top one after a flood	Contributionto top one in case of a false alarm (no flood)	Contributionto top three in case of a flood	Contributionto top three in case of a false alarm (no flood)	Contributionto top one after a flood	Contributionto top one in case of a false alarm (no flood)
Prevention of casualties (loss of life)	97%	47%	92%	40%	87%	48%	87%	44%
Prevention of damage	48%	31%	0%	3%	46%	26%	0%	0%
Availability of public shelters and care	63%	22%	1%	0%	48%	17%	0%	3%
Cooperation between authorities and emergency response units	24%	40%	0%	14%	28%	35%	6%	6%
Support of self-reliance	17%	13%	3%	3%	15%	13%	0%	3%
Accountability of authorities	21%	61%	2%	25%	24%	48%	6%	9%
Public perception	12%	47%	1%	6%	22%	41%	0%	16%
Perception of media	5%	31%	0%	6%	13%	52%	0%	16%
Impact of consequences of flooding outside the flood zone	12%	6%	0%	1%	17%	15%	0%	0%

Key factors in the decision-making process that affect strategy choices

By definition, top strategic decision-making for mass evacuation in the case of a threat of flooding is a low-frequency event for citizens and decision-makers. Most (devel- oped) deltas in the world already have a combination of prevention and emergency management (based on available emergency management for other threats) that influ- ences flood risk. Hence, a basic level of protection is already available.

Decision-makers also have to decide which information to use, and they have to assign a value to information (Boin et al., 2005). Decision-makers (in multiple teams) and crisis managers can provide simultaneously multiple frames of reference about a certain phenomenon. This is called ‘ambiguity’. Some literature describes it as uncertainty (Dewulf et al., 2005, Brugnach et al., 2008), whereas other works state that ambiguity is not a part of uncertainty but is ‘removed on the level of words by linguistic conventions’ (Bedford and Cooke, 2001, p. 19). The risk of linguistic prob- lems increases when risk perception or awareness is limited. Given the continuing struggle to raise awareness of flood risk management among decision-makers (ten Brinke et al., 2008b), and the low perception of risk among the general public (Terpstra, 2009), ambiguity might affect evacuation-related decision-making. Above all, these decision-makers are trained daily in a normal situation in how to take decisions on their own, and they focus on measures that are known and that are common to them. These measures, though, might be less effective in reducing loss of life in a flood and a mass evacuation. Because of ambiguity, therefore, it cannot be guaranteed that all decision-makers will execute a strategy as foreseen. In addition, it cannot be guaranteed that all relevant stakeholders will cooperate with the decision- making process. As a result measures can be counterproductive.

Table3. ‘What is the impact of an external issue on the outcome of the decision-making process for mass evacuation?’

Parameter	Decision-makers				Crisis managers
Response rate of decision-makers: 30% Responserate of crisis managers: 59%	No effect	Delay in decision-making process	Speed up decision-making process	Change of strategy	No effect	Delay in decision-making process	Speed up decision-making process	Change of strategy
Large-scale flu	70%	7%	7%	16%	38%	21%	20%	21%
Pandemic flu	38%	13%	18%	32%	18%	14%	20%	48%
Animal diseases (such as foot-and-mouth disease)	41%	13%	11%	34%	30%	25%	25%	20%
Economic crisis	85%	7%	1%	7%	48%	34%	9%	9%
False alarm in previous year	49%	32%	3%	16%	27%	50%	13%	11%

Table 3 shows the impacts of external issues (such as events in the past or actual circumstances in society) on the chosen evacuation strategy of decision-makers. (Tables 4 and 5, moreover, highlight the influence of different perceptions of risk infor- mation by decision-makers and how they influence top strategic decision-making.) Table 3 clearly reveals that actual circumstances have a strong bearing on top strategic decision-making. Most of the circumstances presented in Table 3 cause a delay in decision-making, and so less time is available to execute an evacuation. These circum- stances also result in the consideration of alternative strategies. The actual circum- stances in a society cannot be gauged in advance in planning documents, meaning that decision-makers have to be able to take them into account in real time. Circumstances that affect human well-being and trust in the government (with regard to false alarms) seem to influence the decision-making process more than economic circumstances.

Table 4. ‘In a situation when the forecast models show the first indications of a possible flood 4 days in advance and the time required for a successful preventive evacuation is approximately one day: When (1 = Certainly, 2 = Probably, 3 = Probably not, 4 = Not at all) should you decide to (A) start to develop several alternatives for evacuation for later decision-making, (B) advise the public to evacuate and (C) call for a mandatory evacuation?’

Parameter	Start planning process				Advised evacuation				Mandatory evacuation
Response rate of decision-makers: 23% Responserate of crisis managers: 54%	Decision-makers		Crisis managers		Decision-makers		Crisis managers		Decision-makers		Crisis managers
	Expected value	Standard deviation	Expected value	Standard deviation	Expected value	Standard deviation	Expected value	Standard deviation	Expected value	Standard deviation	Expected value	Standard deviation
Directly after first signals from forecast models	1.9	0.8	1.9	0.9	3.1	0.9	3.1	0.9	3.6	0.6	3.3	0.9
Later, when experts address the threat as ‘serious’	1.7	1.0	1.7	0.8	2.4	0.9	2.1	0.7	2.8	1.0	2.6	0.8
Later, when public opinion addresses the threat as ‘serious’	2.5	1.1	2.2	1.0	2.7	0.9	2.4	0.7	3.1	0.7	2.9	0.9
Later, when the risk increases to a low probability (10%)	2.4	1.2	2.4	1.1	2.8	0.8	2.6	0.8	3.1	0.7	3.0	0.8
Later, when the risk increases to an average probability (10%–25%)	2.2	1.1	2.1	1.0	2.3	0.9	2.2	0.9	2.7	0.8	2.5	0.8
Later, when the risk increases to a large probability (25%–50%)	2.0	1.2	1.7	1.1	1.7	0.8	1.7	0.9	1.9	0.9	2.0	0.8
Later, when the flood is almost certain (>50%)	1.9	1.2	2.0	1.3	1.3	0.7	1.6	1.0	1.4	0.8	1.6	1.0

Key factors in the decision-making process that affect strategy

Table 4 contains the survey results for when decision-makers and their crisis manag- ers, given enough time, would start emergency planning for an evacuation and call for a mandatory or advised evacuation. An interesting outcome is that decision- makers tend to initiate emergency planning directly when information about a threat is available. This means that the warnings of experts should be presented to them and not be kept from them. Given the information, the decision-makers can decide to prepare (and how to prepare) for a possible nearby event (top strategic decision- making). Table 4 shows that decision-makers only call for an evacuation when the risk increases because of a rise in the probability of flooding. In addition, it reveals that planning, and therefore crisis management structures, are activated more quickly than evacuation decisions are made. In general, a mandatory evacuation requires a higher probability of flooding than an advised evacuation. ‘Probability’ influences the decision to call for an evacuation, yet it also depends strongly, or even more so, on

Table5. ‘What probability of flooding is necessary to be able to choose a certain type of evacuation in a situation 1.5 days before the possible flooding with the knowledge that 1 day is required at minimum?’

Parameter	No opinion	1 (Very low probability: 1–5%)	2 (Low probability: 5–10%)	3 (Average probability: 20–30%)	4 (High probability: 30–40%)	5 (Very High probability: 40–50%)	6 (Almost certain: > 50%)	Expected value	Standard deviation
	Response rate of decision-makers: 28%
A preventive evacuation, instead of a vertical evacuation or shelter in place?	1%	5%	8%	16%	19%	20%	31%	4.4	2.3
A vertical evacuation to a safe haven inside the threatened area, instead of a preventive evacuation or shelter in place?	2%	3%	11%	31%	20%	20%	13%	3.8	1.8
Shelter in place instead of other forms of evacuation	4%	37%	43%	7%	8%	0%	1%	1.9	0.9
	Response rate of crisis managers: 53%
A preventive evacuation, instead of a vertical evacuation or shelter in place?	11%	7%	5%	25%	13%	13%	27%	4.1	2.6
A vertical evacuation to a safe haven inside the threatened area, instead of a preventive evacuation or shelter in place?	9%	0%	5%	13%	23%	34%	16%	4.5	1.2
Shelter in place instead of other forms of evacuation	14%	9%	14%	18%	13%	16%	16%	3.7	2.7

‘expert advice’ (and less on public opinion). Consequently, it is recommended that experts, as well as crisis managers, be included in decision-making.

In relation to the transition phase, this gives the decision-maker the opportunity to influence the people who have to evacuate. This can occur in three key ways: he/she can demonstrate involvement by activating planning, by recommending an evacu- ation when the risk increases, and by calling for a mandatory evacuation.

An extra element of uncertainty in the preparation for flooding and mass evac- uation pertains to how decision-makers and crisis managers deal with risks and uncertainties. Table 5 presents the chosen strategy in relation to the probability of the occurrence of a disaster when just enough time is available for a preventive evacuation; other strategies (with a lower economic impact) could be considered as well, though. It illustrates clearly that the outcome and the speed of a decision-making process depends on the actual probability of flooding and the risk perception of those involved. Based on the same risk, decision-makers tend to choose a variety of strategies, such as pre- ventive evacuation, vertical evacuation, or shelter in place. When more stakeholders on different levels are involved (plus decision-makers and crisis managers), this auto- matically creates a climate for time-consuming discussions or delayed or contradictory decisions. Thus, different timelines for decision-making have to be taken into account, as well as different strategies because of the possible behaviour of the decision-makers and the consequences of decisions.

Expectations of decision-makers to act as decided

Decision-makers and crisis managers were asked about their willingness to cooper- ate as a citizen with the chosen strategy and how they expected their neighbours to behave. Table 6 shows that approximately 25 per cent of respondents would not react to the evacuation call of the government. Expectations concerning the behaviour of neighbours were more pessimistic. Crisis managers assumed that 48 per cent of the people would not pursue the expected strategy of the government; decision-makers were less pessimistic, expecting 64 per cent of them to respond. The literature also

Table 6. ‘How should you respond as a citizen, as a member of a family, to a call for evacuation by the authorities in a situation when the possibility to evacuate preventively exists but you are ordered to respond alternatively?’

Parameter	Decision-makers		Crisis managers
Response rate of decision-makers: 29% Response rate of crisis managers: 59%	Yes	No	Yes	No
Delay moment of departure for preventive evacuation in favour of other strategies	77%	23%	75%	25%
Shelter in place and prepare yourself in your own house so other, more threatened people can evacuate preventively	76%	24%	79%	21%
How would your neighbour respond? Do you expect him to make the same choices as you and your family?	64%	36%	48%	52%

demonstrates that not all people act as advised by the government (the non-compliance rate). During Hurricane Katrina, for instance, some 20 per cent of the people did not leave the New Orleans area (Wolshon, 2006). Furthermore, the evacuation of the Rivierenland area in 1995 (van Duin et al., 1995; Meurs, 1996) did not lead to the removal of all people. Significant non-compliance rates are also known to exist for hurricanes in the US (between 35 and 64 per cent) (Lindell et al., 2002). Planning has to take into account, therefore, that not all people will comply with the chosen strategy. Realistic planning considers possible scenarios for people who do not comply with an evacuation instruction.

Discussion: how to structure evacuation planning to cope with the possible outcome of decision-making

Time span of the transition phase

During the transition phase the authorities can consider adapting the infrastructure, reallocating means and rescue workers, and informing the public. The created cir- cumstance increase the later effectiveness of emergency measures. The time available for top strategic decision-making and the period of the transition phase cannot be defined in advance because such moments depend on the availability of forecasts and the speed of sense-making by decision-makers and the public and the capacity of the infrastructure. Hence, a delay in deciding to evacuate is also a decision with a poten- tial major impact: emergency measures taken at a later stage could be less effective, not effective, or even counterproductive. Using the same line of argument, Boin et al. (2005) state that a non-decision is equal to a decision. Given the lead time for flood- ing and the slow onset of the event, the alternative strategy of ‘delaying the decision’ should be made explicit to decision-makers, and the consequences should be reviewed. Owing to the different risk perceptions of decision-makers and crisis managers, the uncertainties, and the lack of time, the creation of better circumstances for decision- making is recommended. When the situation is considered directly as a national crisis and crisis management structures are activated to connect initiatives and to identify

realistic measures, the performance rate of an evacuation increases.

The involvement of decision-makers and crisis managers

Another factor that might influence top strategic decision-making is the busy agen- das of decision-makers and policymakers. This was highlighted before Waterproef in preparatory discussions about when to involve decision-makers. Following the detection and recognition of a low probability but high impact threat, experts have to be able to put the warning on the agenda to create the boundary conditions for the start of top strategic decision-making. Of course, this will clash directly with other issues; debates will arise about the need for the action. The uncertainty of the threat and low risk perception mean that there is a risk of delaying or ignoring the warning.

Accountability and the need for probabilistic preparation

Public leaders are expected to take care of citizens in liberal democracies (in contrast to non-democratic societies) (Boin et al., 2005). Although it is clear to all stake- holders that the capacities of the authorities are limited in the case of a mass event such as a flood (BZK, 2008b), the public expects to be warned and it expects emergency measures to be taken to reduce the possible impact. Emergency management, therefore, has to maximise the use of available means and infrastructure to prevent casualties, damage, and the capability to return to a normal situation (resilience). Emergency management has to be able to adapt the environment to an evacuation mode in time.

During the survey, decision-makers and crisis managers addressed the importance of the parameters of leadership and accountability in a false alarm and the impor- tance of risk reduction in a flood. Leadership and accountability are also key drivers (in addition to probabilities, costs, and benefits) in the sphere of preparation. For the Netherlands, as well as parts of Australia and the US, the focus is on preventive evacuation, based on the assumption of a certain window of opportunity. Emergency planning focuses on one deterministic strategy based on a pre-defined event: using a chosen flood event, and after identifying the moment when measures to evacuate (all preventive) have been taken, and by whom, a timeline can be defined to con- nect all decisions at the strategic, tactical, and operational level by all stakeholders (including citizen response). This is deterministic planning, often based on best- case scenarios, which have a lead time that is sufficient to implement all measures. Other realistic events, such as a shorter lead time (Barendregt et al., 2005), a larger threatened area (ten Brinke et al., 2010), and disruptions in the decision-making process, are almost not foreseen in emergency preparation, yet they have to be taken into account. Thus, planning does not cover all possible strategies to minimise loss of life and damage.

Public leaders are also responsible, therefore, for realistic planning. The possible consequences of decisions, as well as uncertainties, should be considered. A leader has to initiate probabilistic planning instead of deterministic planning when uncer- tainties and ambiguity can result in different choices during top strategic decision- making to minimise the risk of loss of life and damage.

Owing to a lack of experience of all possible situations, preparations for mass evacuation should be based on relevant international knowledge and data. This can be combined with expertise in local characteristics to avoid the production of ‘fan- tasy’ documents.

Just as training and education are required to develop a culture of emergency management (Alexander et al., 2009), it is also important for the culture that train- ing and education are realistic. When all decision-makers are trained and educated in preventative evacuation in best-case circumstances, other evacuation strategies, such as vertical evacuation, or even no evacuation, are unlikely because they are unknown to the decision-makers and their crisis managers. This is true even when these other strategies might result in less loss of life in a real-world situation.

The decision taken during Waterproef to delay the evacuation of self-reliant indi- viduals while evacuating non-self-reliant individuals and the families of first responders could be seen as unrealistic planning because the effect was not considered. After the evacuation of 1995, it was concluded that a time-phased evacuation strategy of a threatened area (first this group, then the next) does not work in practice because most people will act primarily based on their own assessment (van Duin et al., 1995). The decision taken during Waterproef, combined with the press conference, in reality might be the signal to start a spontaneous evacuation directly. The result could be a preventive evacuation but without the support of mitigating emergency measures. One can question, therefore, whether the level of preparedness of decision-makers and organisations increased because they were not confronted by the consequences of their decision during the exercise.

The role of decision-makers, crisis managers, and experts

The survey shows that the crisis manager has almost the same perception as the decision-maker. Both support the need for experts to explain the threat and possible responses and strongly depend on them for decision-making. Some differences can be identified, though: decision-makers tend to accord more value to risk-based data (probabilities, consequences in terms of casualties and damage) combined with leadership, whereas crisis managers also attach importance to the perceptions of the public and the media. This might introduce extra elements to the decision-making process, increasing the likelihood of establishing meaning and framing and reduc- ing the possibility of risk-based decision-making.

Many crisis teams from different authorities will be involved in a mass evacuation. It is recommended that the decision-making process be simplified to prevent loss of time owing to interference. Experts could advise decision-makers directly, whereas crisis managers could support this process and implement decisions. Their role is less to inform decision-makers and more to make things work.

Conclusions and recommendations

A traditional view of decision-making is that, given a probable but uncertain threat, there is a deterministic way of defining the best decision. This approach is not suf- ficient for mass evacuation due to potential flooding. Decision-making for mass evacuation cannot be seen as a deterministic process. Different timelines for decision- making have to be taken into account, as do different strategies. Owing to uncer- tainties, including the decision-making process itself, and ambiguity, a probabilistic approach (and the development of event trees) is required. This results in realistic planning. A probabilistic approach connects the measures of several stakeholders.

It is unrealistic to assume that all involved stakeholders (and local decision-makers in this network) will act in accordance with the strategy chosen by the top strategic decision-maker. In addition, the time needed for the decision-making process cannot

be determined in advance, and it depends on the involved decision-makers. That time is limited and it might exclude some evacuation strategies (as forms of preventive evacuation) or make them less effective, leading to an increase in loss of life or damage. When ambiguity is considered as a linguistic problem it can be prevented. However, because of low risk perception, the struggle to keep decision-makers involved and the low frequency of event ambiguity have to be taken into account in emergency planning. These variables could influence the effectiveness of strategies or result in

different strategies.

This does not mean that all deterministic planning documents are not realistic. Different strategies could be chosen based on the same information, depending on the decision-makers involved. In addition, local circumstances in a society also influ- ence the decision-making process. When uncertainties are considered, one can conclude that a deterministic approach covers all possible events. As shown in the survey, though, the impact of uncertainties, including the role of decision-makers, necessitates a probabilistic approach to evacuation planning when the objective is to minimise loss of life and damage.

The Waterproef case study demonstrates a lack of attention to the consequences of top strategic decisions and the impact of evacuation on effectiveness and relations with other stakeholders. One can question, therefore, whether such exercises increase or decrease preparedness when no feedback is forthcoming about the comparison between the chosen strategy and actual or planned consequences.

Top strategic decision-making focuses on creating optimal conditions for a response by all stakeholders to a future threat. These emergency measures have to be imple- mented in the ‘transition phase’, to create boundary conditions for future responses. To support top strategic decision-making, the situation has to be considered directly as a crisis, even in the case of a low probability threat. The top strategic decision- maker has to initiate crisis management mechanisms and involve other local decision- makers to enhance the possibility of connected planning. This is a top-down approach that differs from more classic ways of initiating crisis management mechanisms (using a bottom-up approach). When clear visual signals of nearby events are unavailable and public awareness is minimal, the decision has to be taken by decision-makers who are informed by experts. A delay in decision-making might be realistic, but it

can also result in a less effective response.

The study shows that the most important information variables for top strategic decision-making on evacuation are the probability of flooding, the consequences in terms of loss of life and possible economic damage, and leadership and accountability. Accountability with regard to taking care of citizens when things go wrong is also one of the requirements of realistic evacuation planning in a democratic country. Preparations should aim to maximise available means and infrastructure to reduce the risk of loss of life and damage.

This research illustrates that crisis managers have almost the same perceptions as decision-makers. It is recommended that the decision-making process for a mass evac- uation be simplified to prevent loss of time and to address properly risks, costs and

benefits, and uncertainties. It is suggested that experts be added directly to teams of decision-makers, whereas crisis managers should focus on supporting this process and implementing decisions.

Correspondence

Bas Kolen, HKV Consultants, PO Box 2120, 8203 AC Lelystad, The Netherlands. Telephone: +31 320 294231/+31 6 10962684; e-mail: Kolen@hkv.nl

Endnotes

¹ For more background information, see the official evaluation of the exercise (Cappelleveen and Ven, 2009; TMO, 2009c) and the response of the Cabinet on flood preparedness (Ministry of the Interior and Kingdom Relations, 2009).

References

Aldunate, R.G., F. Pena-Mora, and G.E. Robinson (2005) ‘Collaborative distributed decision making for large scale disaster relief operations: drawing analogies from robust natural systems’. Complexity. 11(2). pp. 28–28.

Alexander, D. (1993) Natural Disasters. UCL Press, London.

Alexander, D. (2000) ‘Scenario methodology for teaching principles of emergency management’.

Disaster Prevention and Management. 9(2). pp. 89–97.

Alexander, D. (2004) ‘Vulnerability to landslides’. In T. Glade, M.G. Anderson, and M.J. Crozier (eds.) Landslide Hazard and Risk. Wiley, Chichester. pp. 175–198.

Alexander, D., L. Bramati, and M. Simonetta (2009) ‘Emergency preparedness training and education in Lombardy region, Italy: survey of supply and demand’. Natural Hazards Review. 10(3). pp. 77–83. Baker, E.J. (1991) ‘Hurricane evacuation behavior’. International Journal of Mass Emergencies and Disasters.

9(2). pp. 287–310.

Barendregt, A., J.M. van Noortwijk, M. van der Doef, and S.R. Holterman (2005) Determining the time available for evacuation of a dike-ring area by expert judgement. ISSH – Stochastic Hydraulics 2005. 23 and 24 May. http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.65.6439&rep=

rep1&type=pdf (accessed on 18 March 2014).

Bedford, T. and R. Cooke (2001) ProbabilisticRiskAnalyses:FoundationsandMethods. Cambridge University Press, Cambridge.

Boin, A., P. ’t Hart, E. Stern, and B. Sundelius (2005) The Politics of Crisis Management: Public Leadership under Pressure. Cambridge University Press, Cambridge.

Bourque, L.B., J.M. Siegel, M. Kano, and M.M. Wood (2006) ‘Weathering the storm: the impact of hurricanes on physical and mental health’. The Annals of the American Academy of Political and Social Science. 604(1). pp. 121–159.

Brugnach, M., A. Dewulf, C. Pahl-Wostl, and T. Taillieu (2008) ‘Toward a relational concept of uncertainty: about knowing too little, knowing too differently, and accepting not to know’. Ecology and Society. 13(2). http://www.ecologyandsociety.org/vol13/iss2/art30/ (accessed on 18 March 2014). BZK (Ministry of the Interior and Kingdom Relations) (2007) National Crisis Plan: Extreme Water

Levels and Flooding (in Dutch). BZK, The Hague.

BZK (2008a) National Crisis Plan: Extreme Water Levels and Flooding. Communication Strategy (in Dutch). BZK, The Hague.

BZK (2008b) Kamerstukken II, 2007–2008, 30821 nr. 6. BZK, The Hague.

BZK (2008c) Nationaal Crisisplan Hoogwater en Overstromingen. BZK, The Hague.

BZK and VenW (Ministry of Transport, Public Works, and Water Management (2008) Exercise Waterproef 3 November: Coastal Flooding. Start Information. BZK, The Hague.

Cappelleveen, E. and J. Ven (2009) Evaluation Exercise Waterproef: Summary (in Dutch). Twynstra and Gudde, Amersfoort.

CNN (Cable News Network) (2005) ‘New Orleans braces for monster hurricane’. 29 August. http:// www.cnn.com/2005/WEATHER/08/28/hurricane.katrina/ (accessed on 18 March 2014).

CNN (2006) ‘New Orleans disaster plan: get out of town’. http://www.cnn.com/2006/US/05/02/ hurricane.plan/index.html (accessed on 18 May 2014).

Cole, J. (2008) Hurricane Gustav – Testing the Lessons Learned from Katrina. RUSI Analysis. 3 September. https://www.rusi.org/analysis/commentary/ref:C48BEB8D3402BE/ (accessed on 18 May 2014). Dewulf, A., M. Craps, R. Bouwen, T. Taillieu, and C. Pahl-Wostl (2005) ‘Integrated management of natural resources: dealing with ambiguous issues, multiple actors and diverging frames’. Water,

Science and Technology. 52(6). pp. 115–124.

DHS (Department of Homeland Security) (2006) Mass Evacuation: Developing a Contraflow Plan. DHS, Washington, DC

Dow, K. and C.L. Cutter (2000) ‘Public orders and personal opinions: household strategies for hurri- cane risk assessment’. Environmental Hazards. 2. pp. 143–155.

Fjellman, S.M. (2976) ‘Natural and unnatural decision-making: a critique of decision theory’. Ethos.

4(1). pp. 73–94.

Grothmann, T. and F. Reusswig (2006) ‘People at risk of flooding: why some residents take precau- tionary action while others do not’. Natural Hazards. 38(1–2). pp. 101–120.

Gruntfest, E.C. and K. Carsell (2000) The Warning Process: Toward an Understanding of False Alarms. University of Colorado at Colorado Springs, Colorado Springs, CO.

Hulpverleningsregio Haaglanden (2008) Coördinatieplan Overstromingen Haaglanden, Versie 1.3, 24 April

2008 (in Dutch). Hulpverleningsregio Haaglanden, The Hague.

Haynes, K. et al. (2009a) ‘Shelter-in-place versus evacuation in flash flood environments’. Paper presented at the ‘49th Annual floodplain management authorities conference and 6th biennial Victorian flood conference’, Albury, Australia, 17–20 February 2009. http://floods.org.au/wp- content/uploads/Session-7-paper-23-Coates.pdf (accessed on 18 March 2014).

Haynes, K. et al. (2009b) ‘Shelter-in-place versus evacuation in flash floods’. Environmental Hazards.

8(4). pp. 291–303.

Helsloot, I. and A.C.J. Scholtens (2007) Description of National Crisis Management (in Dutch). BZK, The Hague.

Helsloot, I., A.C.J. Scholtens, and E. Warners (2009) ‘Leerpunten van nationale crisisoefeningen’ (in Dutch). Recht, bestuur en organisatie van hulpdiensten. 6(4). pp. 154–165.

HKV Consultants (2010) Figures for Evacuation Cost–Benefit Analyses. Memorandum PR1919-10 (in Dutch). HKV Consultants, Lelystad.

Jong de, M. and I. Helsloot (2010) ‘The effects of information and evacuation plans on civilian response during the national Dutch flooding exercise “Waterproef”’. In S.P. Hoogendoorn et al. (eds.) First International Conference on Evacuation Modelling and Management. Elsevier, The Hague. pp. 153–162. Jonkman, S.N. (2007) Loss of Life Estimation in Flood Risk Assessment. Theory and Applications. PhD thesis.

Delft University of Technology, Delft.

Jonkman, S.N. and I. Kelman (2005) ‘An analysis of the causes and circumstances of flood disaster deaths’. Disasters. 29(1). pp. 75−97.

Kolen, B. and I. Helsloot (2012) ‘Time needed to evacuate the Netherlands in the event of large- scale flooding: strategies and consequences’. Disasters. 36(4). pp. 700–722.

Kolen, B., C. Vermeulen, and Y. van Bokkum (2008) Interdepartmental Operational Guideline Waterproef, Version 1.0. (in Dutch). Project 13. HKV lijn in water, Lelystad.

Kolen, B., J.K. Leenders, J. Van Der Schaaf, and D.M. Van Dijk (2007) Roles of National Governmental Organisations in Case of Mass Evacuation in the Netherlands. Guidelines for Evacuation and an Emergency Plan (Coordination and Decision-making) for the National Level (in Dutch). HKV, Oranjewoud Save, Lelystad. Kolen, B., M. Bos, J.M. Zoethout, S. Nieuwenhuis, and K. Van Ruiten (2011) ‘The use of different flood scenarios including worst cases for emergency planning and flood preparedness’. Paper presented at the Fifth International Conference on Flood Management (ICFM5), Tokyo, Japan,

27–29 September 2011.

Kolen, B., M. Kok, I. Helsloot, and B. Maaskant (2013) ‘EvacuAid: a probabilistic model to determine the expected loss of life for different mass evacuation strategies during flood threats’. Risk Analysis. 33(7). pp. 1312–1333.

Kunreuther, H. et al. (2002) ‘High stakes decision making: normative, descriptive and prescriptive considerations’. Marketing Letters. 13(3). pp. 259–268.

Lindell, M.K. and R. Perry (1992) Behavioral Foundations of Community Emergency Planning. Hemisphere, Washington, DC.

Lindell, M.K., C.S. Prater, R.W. Perry, and J.Y. Wu (2002) EMBLEM: An Empirically Based Large-scale Evacuation Time Estimate Model. Hazard Reduction and Recovery Center, Texas A&M University, College Station, TX.

Lindell, M.K., J. Lu, and C. Prater (2005) ‘Household decision making and evacuation in response to Hurricane Lili’. Natural Hazards Review. 6(4). pp. 171–179.

LOCC (National Operational Centre) (2008) National Operational Plan: Evacuation for Exercise Waterproef

(in Dutch). LOCC, Driebergen.

Maaskant, B., B. Kolen, R. Jongejan, S.N. Jonkman, and M. Kok (2009) Fractions of Preventive Evacuation in Case of Flooding in the Netherlands (in Dutch). HKV Consultants, Lelystad.

Meurs, R. van (1996) Extreme Water Levels. The Power of Farmer Organisations in River Delta (in Dutch). Scheffers, Herwijnen.

Ministry of the Interior and Kingdom Relations and Ministry of Transport, Public Works, and Water Management (2005) Strategies for Flood Disaster Control along the Rivers Rhine and Meuse (in Dutch). Ministry of the Interior and Kingdom Relations and Ministry of Transport, Public Works, and Water Management, The Hague.

Ministry of the Interior and Kingdom Relations and Ministry of Transport, Public Works, and Water Management (2006) Cabinet Standpoint Disaster Management Floods (in Dutch). Ministry of the Interior and Kingdom Relations and Ministry of Transport, Public Works, and Water Management, The Hague.

Ministry of the Interior and Kingdom Relations and Ministry of Transport, Public Works, and Water Management (2009) Reaction of the Dutch Cabinet on the Results of Taskforce Management Flooding (in Dutch). Ministry of the Interior and Kingdom Relations and Ministry of Transport, Public Works, and Water Management, The Hague.

Ministry of Transport, Public Works, and Water Management (2008) Emergency Plan: ‘Extreme Water Level and Storm Surge’. Guidelines for a National Approach (in Dutch). Ministry of Transport, Public Works, and Water Management, The Hague.

Ministry of Transport, Public Works, and Water Management (2009) National Situational Report. Flooding: Roles and Responsibilities of the National Commission of Flooding. HKV lijn in water, Lelystad. Mitchell, J.T., S.L. Cutter, and A.S. Edmonds (2007) ‘Improving shadow evacuation management: case study of the Graniteville, South Carolina chlorine spill’. Journal of Emergency Management. 5(1).

24. 28–24.

Parker, C., E. Stern, E. Paglia, and C. Brown (2009) ‘Preventable catastrophe? The Hurricane Katrina disaster revisited’. Journal of Contingencies and Crisis Management. 17(14). pp. 206–220.

Rasmussen, J., B. Brehmer, and J. Leplat (1991) Distributed Decision Making: Cognitive Models for Cooperative Work. John Wiley and Sons, Chichester.

RIVM (National Institute for Public Health and the Environment) (2004) Dutch Dikes and Risk Hikes. A Thematic Policy Evaluation of Risks of Flooding in the Netherlands. RIVM, Bilthoven.

Rosenthal, U., M.T. Charles, and P.T. Hart (1989) Coping with Crisis: The Management of Disasters, Riots and Terrorism. Charles C. Thomas, Springfield, IL.

Schneeweiss, C. (2003) ‘Distributed Decision Making – A unified approach’. European Journal of Operational Research. 150(2). pp. 237–252.

Sorensen, J.H. (2000) ‘Hazard warning systems: review of 20 years of progress’. Natural Hazards Review. 1(2). pp. 119–125.

South-Holland, Province et al. (2010) Coordination plan dikering area 14, February 2010 (in Dutch). ten Brinke, W.B.M., B.A. Bannink, and W. Ligtvoet (2008a) ‘The evaluation of flood risk policy in

the Netherlands’. Proceedings of the ICE – Water Management. 161(4). pp. 181–188.

ten Brinke, W.B.M., G.E.M. Saeijs, I. Helsloot, and J. van Alphen (2008b) ‘Safety chain approach in flood risk management’. Municipal Engineer.161(4). pp. 181–188.

ten Brinke, W.B.M., B. Kolen, A. Dollee, H. Van Waveren, and K. Wouters (2010) ‘Contingency planning for large-scale floods in the Netherlands’. Journal of Contingencies and Crisis Management. 18(1). pp. 55–69.

Terpstra, T. (2009) Flood Preparedness: Thoughts, Feelings and Intentions of the Dutch Public. University of Twente, Enschede.

TMO (Taskforce Management Flooding) (2009a) First Impression: Exercise Waterproef. TMO, The Hague. TMO (2009b) Final Report (in Dutch). TMO, The Hague.

TMO (2009c) First Impression Exercise Waterproef (Task Force Management Flooding). The Hague. Tversky, A. and D. Kahneman (1974) ‘Judgment under uncertainty: heuristics and biases’. Science.

185(4157). pp. 1124–1131.

Urbina, E. and B. Wolshon (2003) ‘National review of hurricane evacuation plans and policies: a comparison and contrast of state practices’. Transportation Research Part A: Policy and Practice. 37(1). pp. 257–275.

van Duin, M.J., M.J. Bezuyen, and U. Rosenthal (1995) Evacuation in Case of Extreme Water Levels: Self-reliance and Care of Authorities. COT, University of Leiden andErasmus University Rotterdam, Leiden.

van Noortwijk, J. and A. Barendregt (2004) Available and Necessary Time for Evacuation in Case of Possible Flooding (in Dutch). HKV lijn in water, Lelystad.

van Zuilekom, K.M., M.F.A.M. van Maarseveen, and M.R. van der Doef (2005) ‘A decision support system for preventive evacuation of people’. In P. Van Oosterom, S. Zlatanova, and E.M. Fendel (eds.) Geo-information for Disaster Management. Springer Berlin, Heidelberg. pp. 229–253.

Veiligheidsregio Midden- en West-Brabant (2008) Regionaal Basisplan Overstromingen, Samen voorbereiden op overstromingen, v

doi:10.1111/disa.12059

Decision-making and evacuation planning for flood risk management in the Netherlands

Bas Kolen PhD, MsC Senior Consultant Disaster and Flood Risk Management, HKV Consultants, The Netherlands, and Ira Helsloot PhD Professor, Radboud University, The Netherlands

A traditional view of decision-making for evacuation planning is that, given an uncertain threat, there is a deterministic way of defining the best decision. In other words, there is a linear relation between threat, decision, and execution consequences. Alternatives and the impact of uncer- tainties are not taken into account. This study considers the ‘top strategic decision-making’ for mass evacuation owing to flooding in the Netherlands. It reveals that the top strategic decision- making process itself is probabilistic because of the decision-makers involved and their crisis managers (as advisers). The paper concludes that deterministic planning is not sufficient, and it recommends probabilistic planning that considers uncertainties in the decision-making process itself as well as other uncertainties, such as forecasts, citizens responses, and the capacity of infrastructure. This results in less optimistic, but more realistic, strategies and a need to pay atten- tion to alternative strategies.

Keywords: decision-making, evacuation, flood risk, the Netherlands

Introduction

Evacuation is a potential measure to reduce loss of life in a time of disaster or threat of disaster. People, animals, and goods that can be moved might be saved, but it can be costly in terms of time, money, and credibility (Bourque et al., 2006). Evacuation is a potential measure to address the risk of flooding. When a delay occurs, not every- one can reach the desired destination in time (Urbina and Wolshon, 2003; Barendregt et al., 2005; Jonkman, 2007; Kolen and Helsloot, 2012). The response to Hurricane Katrina in New Orleans, Louisiana, United States, in 2005 demonstrated that people and some movable items of property might be saved, but goods will be affected by flooding, and economic processes will come to a halt (Vrijling, 2009). The cost of an evacuation in the case of a hurricane in the US can exceed USD one million per mile of the coast because of commerce, productivity, and direct losses (Wolshon et al., 2005). With regard to credibility, this involves addressing concerns about the qual- ity and sources of information, the discrepancy between timely warnings and later but more accurate warnings (Dow and Cutter, 2000), and the impact of false alarms (Gruntfest and Carsell, 2000).

Floods often are described as the most deadly of all natural disasters (Alexander,

1993). The mortality rate for different kinds of flooding has been shown to be related

Disasters, 2014, 38(3): 610−635. © 2014 The Author(s). Disasters © Overseas Development Institute, 2014

Published by John Wiley & Sons Ltd, 9600 Garsington Road, Oxford, OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA

to the time available for the implementation of measures ( Jonkman and Kelman, 2005). Furthermore, its relationship with lead time is highlighted in research conducted after other kinds of disasters, such as earthquakes (see, for example, Alexander, 2004). This paper focuses on large-scale flooding caused by extreme water levels on rivers (rain, snow) and storm surges in tidal areas. The expected lead team for these events is relatively long compared to flash floods. The research shows that, owing to uncertainties in forecasting, different geotechnical mechanisms of failure of levees, and different hydraulic conditions, multiple events can happen. For example, extreme events as worst credible events (ten Brinke et al., 2010) can occur as can smaller events when the hydraulic load is much less. Furthermore, the lead time can vary between days and unforeseen events (Barendregt et al., 2005; Jonkman, 2007). By taking these uncertainties into account (as a part of reality), a raft of flood scenar- ios for emergency management can be defined that represents all possible situations

(Kolen et al., 2011).

Human interventions can reduce the consequences of a flood. When people leave the potentially exposed area or move to relatively safe places, such as shelters on high ground, they are less vulnerable to a flood. In addition, traffic management and mass communication campaigns can be implemented in the case of a threat to maximise the possibility of mass evacuation. These human interventions require decisions by citizens and authorities—all of which are based primarily on information about a possible threat, the expected consequences of the threat, and the anticipated outcomes of emergency measures. Authorities can influence the ‘physical environment’ that creates boundary conditions for a later response and increases the effectiveness of emer- gency measures. This paper uses the term ‘transition phase’ to characterise this period. The central issue for authorities (as well as for the general public) is how and when to respond to large-scale flooding. A previous evacuation with or without a flood, a missed call (Gruntfest and Carsell, 2000; Grothmann and Reusswig, 2006), or other circumstances in society (as depicted in the survey presented in this paper) might influ- ence future credibility and response and hence loss of life. However, other evidence suggests that this relation is less important (Baker, 1991; Lindell, Lu, and Prater, 2005). According to Kunreuther et al. (2002), this is because people fail to learn from the past. For low frequency events, as in the Netherlands, one can question whether learning is relevant for citizens because the length of the return period for such events might be far greater than a lifetime. For authorities, though, learning is relevant and explicit attention is needed as citizens in a democratic society expect public leaders to introduce responsible measures based on available information and knowledge in a crisis (Boin et al., 2005). Research shows that exercises and training to stimu- late the correct response have to be based on plausible scenarios (Alexander, 2000).

These scenarios and possible responses have to be recognised and accepted within the realm of planning.

The decision-making process for mass evacuation is influenced by great uncertain- ties and consequences. Research following Hurricane Katrina, for instance, assessed whether or not the earlier involvement of national organisations would have lessened

the impacts (Parker et al., 2009). More insight is needed, therefore, into the effective- ness of different strategies for mass evacuation and the mechanism governing how (and when) to decide on a specific strategy based on the probability of flooding, available time, the characteristics of different areas, and the effect of uncertainties (Kolen and Helsloot, 2012).

Objective and overview

Almost all of the literature (see below) focuses on the relationship between available time and required time for evacuation. This relation is used to define the moment at which to call for an evacuation; planning documents that support decision-making are based mainly on deterministic assumptions about a linear relation between threat/ decision/execution, which is applied to all possible events. Almost never addressed is the role of the decision-making process itself in strategy choice and the moment of decision-making and the impact on evacuation planning. Owing to the effect of ambiguity and current risk perception, a deterministic approach to decision-making processes might not result in minimisation of loss of life or social consequences.

This paper focuses on the impact of ambiguity and uncertainties in decision- making on mass evacuation strategies in the event of flood risk. This is called ‘top strategic decision-making’ and it centres on preventive or vertical evacuation and the moment at which to initiate such a strategy. The literature review below shows that evacuation planning for flooding concentrates on preventive evacuation because this is the best-case strategy. In a best-case strategy optimistic assumptions are used to define the threat and all operational measures. There is no attention to other evac- uation strategies even though they might be more effective in reducing potential loss of life. A better understanding of the role of the decision-maker and the crisis manager in the decision-making process in relation to evacuation strategy choices can result in more realistic evacuation planning and improve the effectiveness of evacu- ation. Consequently, this study concentrated on the:

• concept of top strategic decision-making and the transition phase;
• primary information for top strategic decision-making for evacuation;
• key factors in the decision-making process that affect strategy choices; and
• willingness to act according to the decision

The results are discussed within the context of evacuation planning.

The challenges for the Dutch decision-maker

The people of the Netherlands live in a delta that is largely below sea level. Historically, the country has concentrated principally on flood prevention, resulting in a flood defence system with the highest safety standards in the world. This paper examines flooding in the Netherlands caused by extreme discharges from rivers and/or storm

surges. Forecasting models and early warnings are used to alert crisis organisations and citizens to the need to implement protective measures. However, the time avail- able for evacuation based on forecast and failure mechanisms and the time needed for decision-making are uncertain (Barendregt et al., 2005). Flooding, as a natural hazard, can happen under many different scenarios (ten Brinke et al., 2010; Kolen et al., 2011).

Preparing for flood disasters on a national scale means preparing for extreme— but very unlikely—events (ten Brinke et al., 2010) that involve multiple decision- makers. Critical (forecasted) water levels in the Netherlands that could initiate a decision-making process are foreseen to occur less than once in a lifetime (Ministry of Transport, Public Works, and Water Management, 2008). Hence, the Dutch lack frequent experience of these events as well as other comparable threats with a short lead time in which to initiate a mass evacuation. Most relevant is the evacuation of some 250,000 persons in 1995 owing to the level of river waters. The main reason to select preventive evacuation in 1995 was the message from the water boards that they could not guarantee any longer the strength of the dikes (van Duin et al., 1995). In the days and hours before this announcement, there was a sense of urgency among the authorities and members of the general public because of the rising water levels in the rivers. Subsequently, discussions took place about the need for mandatory evacu- ation (Meurs, 1996).

An evaluation of the water safety policy in 2004 showed that the Netherlands is not prepared for extreme flooding (RIVM, 2004; see also ten Brinke, Bannink, and Ligtvoet, 2008a). As a result, the Government of the Netherlands (Ministry of the Interior and Kingdom Relations and Ministry of Transport, Public Works, and Water Management, 2005, 2006) sought to address the need for improved preparation. National (BZK, 2007; Kolen et al., 2007; Kolen, Vermeulen, and van Bokkum, 2008; LOCC, 2008; Ministry of Transport, Public Works, and Water Management, 2008; Wegh, 2008) and regional (Brabant, Veiligheidsregio Midden- en West- Brabant, 2008; Hulpverleningsregio Haaglanden, 2008; Veiligheidsregio Zeeland, 2008; TMO, 2009b) authorities prepared emergency plans for flood prevention and large-scale preventive evacuation. Drafts and first-generation plans were tested in a nationwide exercise entitled ‘Waterproef’ between 3 and 7 November 2008 (TMO, 2009b). The Government of the Netherlands (Ministry of the Interior and Kingdom Relations and Ministry of Transport, Public Works, and Water Management, 2009) concluded that, owing to this planning, research, and testing, the country was better prepared for flooding, although improvements still could be made, such as in evac- uation planning.

A complete preventive evacuation of large coastal areas generally is not possible. Road capacity is not appropriate for a preventive evacuation within a realistic early warning time frame. The provinces of North and South Holland, the most valuable part of the Netherlands in terms of economic processes, need the most time for preventive evacuation. Other provinces, such as Zeeland, require less time because they are less populated. In most circumstances, however, they still need more than

one day. Other evacuation strategies are highlighted in planning documents as pos- sible alternatives strategy, but they are not (yet) taken into account. A preventive evacuation owing to river flooding in polder areas takes less time than one for coastal flooding in highly populated areas, such as the province of South Holland, and fore- casts of water levels are more uncertain in the case of a storm surge than in relation to high discharges on rivers. In general, these areas can evacuate in time, although exceptions can manifest themselves because of unexpected events.

Earlier research by the authors (Kolen and Helsloot, 2012) underlined the need to consider three alternative strategies:

• Preventive evacuation: the organisation and movement of people from a poten- tially exposed area to a safe location outside of this area before the start of a
• Vertical evacuation: the organisation and movement of people inside the area under threat to shelters or safe havens before the start of a
• Shelter in place (or hide): the organisation and movement of people to the upper levels of residential buildings at the location before the start of a

In addition to preventive evacuation, safe havens, shelters, vertical evacuation, and support to augment the self-reliance of citizens should be taken into account in order to reduce loss of life and the impacts of evacuation (Haynes et al., 2009b; Kolen and Helsloot, 2012; Kolen et al., 2013). Decision-makers, though, have to make choices about evacuation strategy. These are only relevant when they can be implemented before people in the threatened area initiate an evacuation on their own. Research shows that an increase in the time available for evacuation is more critical for its effectiveness than further improvements in how organisations connect their plan- ning, how people behave during an evacuation, or how infrastructure is used during an evacuation (Kolen et al., 2013).

The challenge facing decision-makers is how to deal with the positive (reduction in loss of life in a flood) and negative (economic and social disruption) consequences of evacuation, which are related to the uncertainty of a flood and the time available. The size of a flood (for example, 20 per cent of the Netherlands might be flooded in one event), the magnitude of the evacuation, and the possible autonomous response of the public increase complexity.

To gain more insight into the decision-making process and the role of decision- makers and crisis managers during an evacuation, this paper evaluates the ‘Waterproef’ large-scale exercise of 2008. In addition, a survey of all 431 Dutch mayors and 95 international crisis managers who advise decision-makers was conducted between January and March 2010.

The study focused on the situation in the Netherlands for two key reasons:

• complete preventive evacuation is not always possible so there is a need to consider different evacuation strategies; and
• the magnitude of the threat in the Netherlands means that a large number of local and national decision-makers and crisis managers are

Dutch mayors of municipalities are responsible for emergency planning in their community and safety region. When their communities are under threat they have a role to play in the evacuation decision-making process, or they can be confronted by the consequences of evacuation in their community or safety region or the deci- sions of the national authorities.

The 95 crisis managers who participated in the survey all have a role in crisis centres and inform and advise decision-makers. The survey of crisis managers took place during the visiting programme of the international exercise ‘EU Floodex’ on 23–24 September 2009. All respondents were given some background information on mass evacuation in the Netherlands and a description of the different types of evacuation described in this paper. Furthermore, it was assumed that all respond- ents were aware of the possibility of flood risk (because of their involvement in the exercise and their profession).

The research endeavour aimed to gain more insight into the impact of decision- makers and crisis managers. Although one can question whether decision-makers in the Netherlands will respond to a crisis in the same way as they answered the ques- tions in the survey, the results are important for emergency planning since they clarify the way in which they think.

Experience and knowledge of decision-making for flood-related mass evacuation in the Netherlands is scarce. Flooding is a (very) low-frequency event in the coun- try, and the public perception is limited (Terpstra, 2009). However, authorities have paid increasing attention to the consequences of flooding since Hurricane Katrina in the US in August 2005 (ten Brinke, Bannink, and Ligtvoet, 2008a).

Future preparation for mass evacuation (such as emergency planning, exercises, and research) could challenge the results of the survey. Moreover, according to the literature, a lack of knowledge of heuristics and biases in dealing with uncertainties in these situations can influence the decision-making processes negatively (Tversky and Kahneman, 1974). Nevertheless, the survey serves to reveal the current percep- tions of decision-makers and crisis managers of mass evacuation in the Netherlands. For emergency planning, this range of perceptions is particularly important.

Literature review of different mass evacuation strategies Much attention has been paid to preventive evacuation in the event of a possible flood, yet almost none has been devoted to other strategies. In New Orleans, for

instance, ‘a complete evacuation of the city has been the cornerstone of hurricane preparedness planning for the region’ (Wolshon, 2006, p. 28). During Hurricane Katrina it was clear that not all citizens could, or wanted to, leave the area in time. Therefore ‘shelters of last resort’, such as the Ernest N. Morial Convention Center and the Mercedes-Benz Superdome, were opened (CNN, 2005). Post Katrina, New Orleans Mayor Ray Nagin declared that shelters of last resort would not be used again in the future (CNN, 2006). None were opened during Hurricane Gustav of August–September 2008. Those who did not leave the area in time arranged their own shelter, such as building on high areas (in the French Quarter).

Wolshon (2006, p. 28) describes hurricane-related evacuation in the US as an ini- tiative to ‘move people away from danger’, but he notes that it might not be possible to evacuate everyone preventively. Hence, there is an implicit need for other strategies. The need for strategies other than preventive evacuation has been addressed in contemporary literature. Following an analysis of 50-year flash-flooding in Australia, Haynes et al. (2009a, p. 9) concluded that: ‘in cases where evacuation may lead to increased exposure to danger and a suitable refuge exists for suitable occupants, shel- tering in place may be a better option’. In addition, they pointed out that: ‘At the moment, the literature cannot unequivocally support one option over another, in part due to the fact that because evacuation is such a well-established emergency management strategy, literature about policy alternatives is relatively thin on the ground. What the literature does show is that neither strategy is without risk and more research is needed to guide decision-making by emergency managers. In the end, emergency managers and the people directly at risk need to be able to assess the

relative risks of alternative strategies’.

Emergency planners in the Netherlands concentrate only on preventive evacu- ation (as illustrated by the Waterproef exercise) (TMO, 2009a). Jonkman (2007) mentions the possibility of vertical evacuation, but he focuses only on preventive evacuation. What is more, emergency planning for coastal areas of the provinces of North and South Holland assumes that coordination could solve the problems associ- ated with the limited capacity of the infrastructure and restricted lead time. The Coordination plan dikering area 14 states that: ‘as long as there is no national operation evacuation plan a large scale preventive mass evacuation seems not possible in case of coastal flooding’ (South-Holland, Province, et al., 2010, p. 24). Earlier research, however, reveals that, even in a perfect situation, a complete preventive evacuation is not possible (Barendregt et al., 2005; BZK, 2008b; Maaskant et al., 2009).

The success of an evacuation strategy depends on the relation between the time available and the time required to execute the strategy and emergency measures. Time available depends on a combination of availability of forecasts and their use by decision-makers and experts (van Zuilekom, van Maarseveen, and van der Doef, 2005; Jonkman, 2007; Kolen and Helsloot, 2012). Meanwhile, analyses show that the time required for an evacuation can be reduced by introducing several extra emergency measures, such as mass communication and traffic management. The time needed to evacuate the ‘Islands of Zealand and South Holland’ in the event of a possible storm surge, for example, can be decreased from a worst-case scenario of approximately 55 hours to some 25 hours. Although a complete preventive evacu- ation for the entire Dutch coast still remains impossible, national traffic management can lower the time needed to evacuate 50 per cent of the population by almost a full day (from 44 to 27 hours). A theoretical mathematical solution that optimises the use of roads and utilises the behaviour of people as a variable predicts that the time required can be reduced by up to 48 hours (Kolen and Helsloot, 2012). Furthermore, the contra-flow system of New Orleans shows that it is possible to decrease the time required for evacuation (DHS, 2006; Wolshon, 2006),

Other strategies, such as vertical evacuation, might result in less loss of life because shelters are relatively safe areas. By comparing lead and required time and deter- mining loss of life in a preventive and vertical evacuation, it has been shown that a vertical evacuation in the coastal areas of the provinces of North and South Holland will result in less loss of life than a preventive evacuation, except in the very opti- mistic (best) case of an exceptional (very uncertain) lead time and a perfect logistical operation (Kolen et al., 2013). For other less-populated areas, whether a preventive or a vertical evacuation results in minimum of loss of life depends on the time available and the implemented measures (Kolen and Helsloot, 2012).

Case study: ‘Waterproef’

Scope of the exercise

The Ministry of the Interior and Kingdom Relations, the Ministry of Transport, Public Works, and Water Management, and Taskforce Management Flooding organised ‘Waterproef’ between 3 and 7 November 2008. It was the first national exercise on flooding and mass evacuation and was held after a two-year programme of improve- ments for flood preparedness (TMO, 2009b). This study assesses the part of the exer- cise related to decision-making on evacuation during coastal flooding.1

Several national (crisis centres and Rijkswaterstaat) and regional (safety regions and water boards) organisations took part in the exercise. A public panel also partici- pated, providing feedback on the communications of the authorities and the media (de Jong and Helsloot, 2010). Waterproef focused on the choice of decision-making strategy for evacuation four days before an expected dike failure and possible flood. A scenario was developed that described a possible storm surge that could cause large- scale flooding along the coast of the Netherlands—a best-case scenario was used to reduce complexity and to present preventive evacuation of coastal areas as a serious option. The development of the threat and of all decisions between the first warning (eight days before the possible flood) and the day of the exercise was described in a start document based on existing emergency planning (Kolen, Vermeulen, and van Bokkum, 2008). Decisions concerning evacuation were foreseen four days before the expected moment of failure of levees. During the exercise regional decision- makers called for an evacuation, but this was not supported at the national level.

Even though an exercise is a constructed situation and is not developed for scientific research, and it is so large that it cannot be controlled (Helsloot, Scholtes, and Warners, 2010), lessons can be learned that are applicable to top strategic decision-making for mass evacuation by taking its circumstances into account. An evaluation of the exercise offers a unique (because of the presence of observers) view of national crisis management.

Lessons learned during preparations for Waterproef

Waterproef illustrates the difficulties that decision-makers face in dealing with uncer- tainties and in employing an integrated approach. Three alternatives for mass evac- uation presented to the national decision-makers sparked debate among them. This

resulted in a decision to implement a totally new strategy (not prepared in advance)— involving the evacuation of non-self-reliant individuals and the families of first responders; others had to wait—that delayed the preventive evacuation by at least a day. One can also question whether the decision made was realistic or even counterproductive. At the same time, regional crisis centres advocated complete preventive evacua- tion because they were not aware of the other options. In addition, the decision- makers of one region decided to call for an evacuation on their own (based on their own risk perceptions and responsibilities), even though they were aware of the national

decision-making process.

The combined impact of all decisions on an evacuation (in terms of a reduction in loss of life or risk) has not been defined in planning and was not addressed dur- ing the exercise. Consequently, they were not taken into account in the study. The decision-making process was dominated by perceptions and expectations regarding the effectiveness of an evacuation and the cooperation of others.

Waterproef also showed that the personal opinions of crisis managers vary strongly and are influenced by available information over time. Analysis of the development of top strategic decision-making, as recorded in the starting document, which was the input for the exercise, reveals substantial differences in when and how to inform others. Although this starting document (BZK and VenW, 2008) was based on avail- able emergency planning, several crisis managers still questioned the point at which to inform decision-makers and when to call for certain emergency measures. This produced several extreme opinions, such as ‘directly after the detection of a possible storm surge’ up to the ‘moment to call for a mass evacuation’. Arguments presented included ‘decision-makers are too busy and not willing to spend time’, ‘it is not serious enough’, ‘media pressure will force them to meet directly’, and ‘because of the lack of resources and the possible consequences’. The matter has a serious political dimension because the most logical move, ‘to wait and see’, can become a dramatic decision, as less time is available for evacuation, resulting in greater loss of life.

Top strategic decision-making: key factors and parameters

The concept and the transition phase

Top strategic decision-making for mass evacuation deals with (i) when to initiate an evacuation and (ii) the type of evacuation (preventive, vertical evacuation, use of shelters, and the creation of optimal (or better) circumstances for evacuation). The top strategic decision-maker is at the apex of the decision-making tree—the Ministerial Policy Team vis-à-vis flooding in the Netherlands. Top strategic decision-makers will be confronted automatically with many choices, great uncertainties, and myriad consequences in all circumstances. Uncertainties occur, for example, in predicting flooding (size and probability of occurrence) (ten Brinke et al., 2010; Kolen et al., 2011), the effectiveness of emergency measures (Kolen and Helsloot, 2012), and the responses of other stakeholders (local authorities, first responders, and citizens).

When forecasts become clearer and uncertainties decline (see Figure 1), people and decision-makers start to act. The autonomous response of citizens can lead to overload or the inefficient use of road capacity and available equipment and can place limitations on authorities in implementing further mitigating measures. Several models describe the possible responses of citizens to a natural hazard based on the interaction of environmental, individual, and social processes (Lindell and Perry, 1992; Sorensen, 2000; Grothmann and Reusswig, 2006; Kolen et al., 2013). It is known that not all people act directly after receiving a flood warning and that it takes time before people start to evacuate (Lindell et al., 2002). Since floods do not respect administrative boundaries multiple decision-makers are involved. The autono- mous response of these decision-makers can result in counterproductive measures as well as the less optimal use of available resources and infrastructure.

Emergency measures have to be implemented before the combined consequences of the autonomous responses of others (citizens, organisations) create boundary con- ditions for evacuation. The impact of these ‘top strategic decisions’ depends on the possibility of establishing circumstances that facilitate a future response by citizens and several stakeholders. Rasmussen, Brehmer, and Leplat (1991) describe this pro- cess as reflective decision-making: the decision has to be made in relation to the decisions of others. These top strategic decisions have to be made based on informa- tion on forecasts and scenarios for evacuation and before people start to act. These decisions involve a transition from normal life to a mass evacuation mode. This period is called the transition phase (see Figure 1). Future decisions are made during an operation in the context of the evacuation mode.

During the top strategic decision-making phase, one can already speak of a ‘crisis’. This is defined as the moment when policymakers experience ‘a serious threat to the basic structures or the fundamental values and norms of a system, which under time pressure and highly uncertain circumstance necessitates making vital decisions’ (Rosenthal, Charles, and Hart, 1989, p. 10). Preparations for a future mass evacuation have the objective of transporting as many people, animals, and movable goods to the

Figure 1. Conceptual illustration of the transition phase for top strategic decision-making

Source: authors.

safest possible place before the event (in this case, a flood). In addition to flooding, which is considered to be a national crisis in the Netherlands (Helsloot and Scholtens, 2007), and hurricanes in the US (Cole, 2008; Parker et al., 2009), mass evacuation itself also has to be seen as a crisis.

To reduce the consequences of a possible flood, decision-makers can opt to initiate another crisis: the evacuation itself. An example is a ‘shadow evacuation’ (a non- authorised evacuation), as seen in the US during a chlorine spill in Graniteville, South Carolina (Mitchell, Cutter, and Edmonds, 2007), and during Hurricane Rita in Houston, Texas (DHS, 2006), in 2005.

The transition phase reveals the role of top strategic decision-making for mass evacuation: to create the conditions needed for the taking of other decisions and for the implementation of emergency measures in the near future by authorities, emer- gency services, and citizens. The following are examples of top strategic decisions with regard to communication policy and operational measures:

• Communicate with the public about the risk, the impacts, possible emergency measures, and
• Take policy decisions to influence other authorities. Warn the relevant national and regional authorities (if not warned already). Define the go/no-go decision and strategy for evacuation (preventive, vertical, shelter in place, or a combination). Inform other authorities about the risks and consequences (and timelines) of the threat, as well as about possible emergency measures, the impact of uncertainties, how to call for assistance, juridical arrangements, and international
• Implement operational emergency measures to adapt the environment, including implementing national traffic management, identifying the availability of routes, assigning regions that will offer public shelters, and prioritising the use of limited available (national)

For evacuation planning, authorities in surrounding areas have an important part to play in supporting evacuation operations, such as traffic management, providing shelter, and delivering equipment and services (Wolshon, 2006; Ministry of Transport, Public Works, and Water Management, 2008; Wegh, 2008). Emergency planning (BZK, 2008a; Ministry of Transport, Public Works, and Water Management, 2008, 2009), research post evacuation (Jonkman, 2007; Kolen and Helsloot, 2012), experience of Hurricane Katrina (Parker et al., 2009), and exercises (TMO, 2009a) in the Netherlands indicate that proactive and direct involvement at the national level is necessary to increase the effectiveness of emergency measures following a national disaster.

The mass evacuation of Rivierenland in 1995 and the response to Hurricane Katrina in 2005 underscore the importance of and the difficulties associated with top stra- tegic decision-making, including involving relevant partners in time and when to call for a preventive evacuation. During Katrina, some people did not want to evac- uate because they hoped or they assumed that, as in the past, the hurricane would not hit their area (Parker et al., 2009). Earlier involvement at the national level, such as by the Federal Emergency Management Agency or the Red Cross, might have reduced

some of the consequences (Parker et al., 2009). The top strategic decision to involve the national level early in creating better conditions for a response was a lesson learned in New Orleans, as well as in the Netherlands (van Duin et al., 1995). During Hurricane Gustav of 2008, the national level made a concerted effort to be on top of the situ- ation and to show its concern (Cole, 2008). Although this is not clear evidence of an increase in the effectiveness of top strategic decision-making, such action surely affects the perceptions of professionals and members of the general public. Another key factor that contributed to the Gustav response was recent experience of Katrina. When time is limited, other strategies for mass evacuation, such as vertical evacu- ation and implementation of national traffic management (if implemented in time), might be more attractive (van Noortwijk and Barendregt, 2004; Wolshon, 2006; Jonkman, 2007). Whether to implement them is down to the decision-makers involved.

Primary information for top strategic decision-making for evacuation

The literature shows that, when relevant stakeholders have contact with one another, an optimal decision-making process can be implemented in which the right people work on the right objective at the right moment with the right information (Aldunate, Pena-Mora, and Robinson, 2005). One can question whether or not information can or will ever be completely available. At the moment the information is analysed, new information is, by definition, available because of the ongoing nature of the disaster or threat. It is also impossible to know whether all relevant information is available when taking the number of stakeholders (such as citizens, crisis centres, and first responders) into account.

In a Western society, some tasks of government are spread across several national and regional (semi-) governmental organisations; others are privatised. In normal day-to-day life, these organisations implement their own emergency measures based on their policies. The theory of ‘Distributed Decision Making’—defined as the design and coordination of connected decisions (Schneeweiss, 2003)—describes the optimi- sation of multiple decisions in a situation involving multiple interests of organisations. The theory assumes that society is differentiated in such a way that a central body cannot control it via a hierarchical relation. The theory becomes more relevant when more stakeholders make decisions. Thus, decision-makers should take ‘other decisions’ into account so they do not frustrate other decisions.

In the case of the threat of flooding, time is available to share information and to discuss the decision-making process. Therefore, the theory of ‘Natural Decision Making’, which describes how people act in a disaster (Fjellman, 1976), applies only to top strategic decision-makers and their crisis managers. First responders are not confronted immediately with the need to act: during this phase, the flood has not occurred yet and evacuation has still to commence. Aside from top strategic decision- makers, no one else faces the consequences directly, so they are not required to make any immediate decisions. This might result in calls for further information gathering and in a delay in decision-making. Critical moments could pass, meaning that some emergency measures, such as a preventive evacuation, can no longer be introduced.

A major flood event, involving all stakeholders, as noted above, is a national crisis in the Netherlands (Helsloot and Scholtens, 2007; BZK, 2008c). In the best case, stakeholders in the dynamic organisation are completely aware of the available infor- mation and execute the (centrally) chosen strategy perfectly. However, decision- makers have to deal with imperfect and uncertain information and are confronted by the decisions of others. The key question is: what kind of information do they need during the transition phase? Another fundamental matter concerns the prioritising of sources of information (see Table 1).

The results show that the survey participants (mayors and crisis managers) all tend to prefer a risk-based approach, according most value to the probability, impact, and effectiveness of possible strategies. In addition, all placed great importance on expert advice. Public pressure and the economic and social consequences of decisions were considered to be less important. This means that decision-makers tend towards a rational approach and rely on experts for advice to support decision-making.

In a potential mass evacuation, however, one can conclude that the capacities of the emergency services are far outweighed by the population that needs to be served. Evacuation might reduce loss of life and the cost of lost movable goods, but it cannot decrease the expected damage to fixed goods, such as agricultural land and houses. Hence, authorities and emergency services have to prioritise and deal with limited resources.

Table 1. ‘Determine the importance of each item to the decision-making process about mass evacuation on a scale of 1 to 5 (1 = no importance, 2 = less important, 3 = important 4 = very important 5 = most important)’

Parameter	Decision-makers		Crisis managers
Response rate of decision-makers: 38% Response rate of crisis managers: 59%	Expected value	Standard deviation	Expected value	Standard deviation
Probability of flooding	4.2	0.8	3.9	0.8
Size of the threatened area	3.5	0.8	3.6	0.8
Time available until failure of defence system	4.0	0.7	4.1	0.7
Public pressure	2.9	0.6	2.9	0.7
Effectiveness of a strategy	3.8	0.8	3.8	0.8
Economic impact of an unnecessary evacuation	3.0	0.7	2.9	0.8
Social impact of an unnecessary evacuation	3.3	0.7	2.9	0.8
Accountability for decisions made	3.4	0.8	3.2	0.8
Required leadership	3.5	0.8	4.0	0.8
Expert advice	3.9	0.7	3.8	0.7

Hindsight permits an examination of the best decisions in a flood event or in a situ- ation when a flood did not occur. The aftermath also influences public opinion on the response of decision-makers. Thus the second question in the survey focused on the parameters expected to be most important following an evacuation when a flood did or did not occur (see Table 2).

The results show that a reduction in loss of life is seen as a more important param- eter than a decrease in damage. Evacuation happens more frequently than flooding in the Netherlands (HKV Consultants, 2010). In the case of an evacuation that is not followed by a flood event, more attention is paid to accountability of decision-makers and cooperation between authorities than when the flood does happen, although pre- vention of loss of life remains important (instead of prevention of damage).

Table 2. ‘What are the 3 factors that contribute most to whether an evacuation decision was “right” in (1) a situation after a flood and (2) after a false alarm?’

Parameter	Decision-makers				Crisis managers
Response rate of decision-makers: 23% Responserate of crisis managers: 48%	Contributionto top three in case of a flood	Contributionto top three in case of a false alarm (no flood)	Contributionto top one after a flood	Contributionto top one in case of a false alarm (no flood)	Contributionto top three in case of a flood	Contributionto top three in case of a false alarm (no flood)	Contributionto top one after a flood	Contributionto top one in case of a false alarm (no flood)
Prevention of casualties (loss of life)	97%	47%	92%	40%	87%	48%	87%	44%
Prevention of damage	48%	31%	0%	3%	46%	26%	0%	0%
Availability of public shelters and care	63%	22%	1%	0%	48%	17%	0%	3%
Cooperation between authorities and emergency response units	24%	40%	0%	14%	28%	35%	6%	6%
Support of self-reliance	17%	13%	3%	3%	15%	13%	0%	3%
Accountability of authorities	21%	61%	2%	25%	24%	48%	6%	9%
Public perception	12%	47%	1%	6%	22%	41%	0%	16%
Perception of media	5%	31%	0%	6%	13%	52%	0%	16%
Impact of consequences of flooding outside the flood zone	12%	6%	0%	1%	17%	15%	0%	0%

Key factors in the decision-making process that affect strategy choices

By definition, top strategic decision-making for mass evacuation in the case of a threat of flooding is a low-frequency event for citizens and decision-makers. Most (devel- oped) deltas in the world already have a combination of prevention and emergency management (based on available emergency management for other threats) that influ- ences flood risk. Hence, a basic level of protection is already available.

Decision-makers also have to decide which information to use, and they have to assign a value to information (Boin et al., 2005). Decision-makers (in multiple teams) and crisis managers can provide simultaneously multiple frames of reference about a certain phenomenon. This is called ‘ambiguity’. Some literature describes it as uncertainty (Dewulf et al., 2005, Brugnach et al., 2008), whereas other works state that ambiguity is not a part of uncertainty but is ‘removed on the level of words by linguistic conventions’ (Bedford and Cooke, 2001, p. 19). The risk of linguistic prob- lems increases when risk perception or awareness is limited. Given the continuing struggle to raise awareness of flood risk management among decision-makers (ten Brinke et al., 2008b), and the low perception of risk among the general public (Terpstra, 2009), ambiguity might affect evacuation-related decision-making. Above all, these decision-makers are trained daily in a normal situation in how to take decisions on their own, and they focus on measures that are known and that are common to them. These measures, though, might be less effective in reducing loss of life in a flood and a mass evacuation. Because of ambiguity, therefore, it cannot be guaranteed that all decision-makers will execute a strategy as foreseen. In addition, it cannot be guaranteed that all relevant stakeholders will cooperate with the decision- making process. As a result measures can be counterproductive.

Table3. ‘What is the impact of an external issue on the outcome of the decision-making process for mass evacuation?’

Parameter	Decision-makers				Crisis managers
Response rate of decision-makers: 30% Responserate of crisis managers: 59%	No effect	Delay in decision-making process	Speed up decision-making process	Change of strategy	No effect	Delay in decision-making process	Speed up decision-making process	Change of strategy
Large-scale flu	70%	7%	7%	16%	38%	21%	20%	21%
Pandemic flu	38%	13%	18%	32%	18%	14%	20%	48%
Animal diseases (such as foot-and-mouth disease)	41%	13%	11%	34%	30%	25%	25%	20%
Economic crisis	85%	7%	1%	7%	48%	34%	9%	9%
False alarm in previous year	49%	32%	3%	16%	27%	50%	13%	11%

Table 3 shows the impacts of external issues (such as events in the past or actual circumstances in society) on the chosen evacuation strategy of decision-makers. (Tables 4 and 5, moreover, highlight the influence of different perceptions of risk infor- mation by decision-makers and how they influence top strategic decision-making.) Table 3 clearly reveals that actual circumstances have a strong bearing on top strategic decision-making. Most of the circumstances presented in Table 3 cause a delay in decision-making, and so less time is available to execute an evacuation. These circum- stances also result in the consideration of alternative strategies. The actual circum- stances in a society cannot be gauged in advance in planning documents, meaning that decision-makers have to be able to take them into account in real time. Circumstances that affect human well-being and trust in the government (with regard to false alarms) seem to influence the decision-making process more than economic circumstances.

Table 4. ‘In a situation when the forecast models show the first indications of a possible flood 4 days in advance and the time required for a successful preventive evacuation is approximately one day: When (1 = Certainly, 2 = Probably, 3 = Probably not, 4 = Not at all) should you decide to (A) start to develop several alternatives for evacuation for later decision-making, (B) advise the public to evacuate and (C) call for a mandatory evacuation?’

Parameter	Start planning process				Advised evacuation				Mandatory evacuation
Response rate of decision-makers: 23% Responserate of crisis managers: 54%	Decision-makers		Crisis managers		Decision-makers		Crisis managers		Decision-makers		Crisis managers
	Expected value	Standard deviation	Expected value	Standard deviation	Expected value	Standard deviation	Expected value	Standard deviation	Expected value	Standard deviation	Expected value	Standard deviation
Directly after first signals from forecast models	1.9	0.8	1.9	0.9	3.1	0.9	3.1	0.9	3.6	0.6	3.3	0.9
Later, when experts address the threat as ‘serious’	1.7	1.0	1.7	0.8	2.4	0.9	2.1	0.7	2.8	1.0	2.6	0.8
Later, when public opinion addresses the threat as ‘serious’	2.5	1.1	2.2	1.0	2.7	0.9	2.4	0.7	3.1	0.7	2.9	0.9
Later, when the risk increases to a low probability (10%)	2.4	1.2	2.4	1.1	2.8	0.8	2.6	0.8	3.1	0.7	3.0	0.8
Later, when the risk increases to an average probability (10%–25%)	2.2	1.1	2.1	1.0	2.3	0.9	2.2	0.9	2.7	0.8	2.5	0.8
Later, when the risk increases to a large probability (25%–50%)	2.0	1.2	1.7	1.1	1.7	0.8	1.7	0.9	1.9	0.9	2.0	0.8
Later, when the flood is almost certain (>50%)	1.9	1.2	2.0	1.3	1.3	0.7	1.6	1.0	1.4	0.8	1.6	1.0

Key factors in the decision-making process that affect strategy

Table 4 contains the survey results for when decision-makers and their crisis manag- ers, given enough time, would start emergency planning for an evacuation and call for a mandatory or advised evacuation. An interesting outcome is that decision- makers tend to initiate emergency planning directly when information about a threat is available. This means that the warnings of experts should be presented to them and not be kept from them. Given the information, the decision-makers can decide to prepare (and how to prepare) for a possible nearby event (top strategic decision- making). Table 4 shows that decision-makers only call for an evacuation when the risk increases because of a rise in the probability of flooding. In addition, it reveals that planning, and therefore crisis management structures, are activated more quickly than evacuation decisions are made. In general, a mandatory evacuation requires a higher probability of flooding than an advised evacuation. ‘Probability’ influences the decision to call for an evacuation, yet it also depends strongly, or even more so, on

Table5. ‘What probability of flooding is necessary to be able to choose a certain type of evacuation in a situation 1.5 days before the possible flooding with the knowledge that 1 day is required at minimum?’

Parameter	No opinion	1 (Very low probability: 1–5%)	2 (Low probability: 5–10%)	3 (Average probability: 20–30%)	4 (High probability: 30–40%)	5 (Very High probability: 40–50%)	6 (Almost certain: > 50%)	Expected value	Standard deviation
	Response rate of decision-makers: 28%
A preventive evacuation, instead of a vertical evacuation or shelter in place?	1%	5%	8%	16%	19%	20%	31%	4.4	2.3
A vertical evacuation to a safe haven inside the threatened area, instead of a preventive evacuation or shelter in place?	2%	3%	11%	31%	20%	20%	13%	3.8	1.8
Shelter in place instead of other forms of evacuation	4%	37%	43%	7%	8%	0%	1%	1.9	0.9
	Response rate of crisis managers: 53%
A preventive evacuation, instead of a vertical evacuation or shelter in place?	11%	7%	5%	25%	13%	13%	27%	4.1	2.6
A vertical evacuation to a safe haven inside the threatened area, instead of a preventive evacuation or shelter in place?	9%	0%	5%	13%	23%	34%	16%	4.5	1.2
Shelter in place instead of other forms of evacuation	14%	9%	14%	18%	13%	16%	16%	3.7	2.7

‘expert advice’ (and less on public opinion). Consequently, it is recommended that experts, as well as crisis managers, be included in decision-making.

In relation to the transition phase, this gives the decision-maker the opportunity to influence the people who have to evacuate. This can occur in three key ways: he/she can demonstrate involvement by activating planning, by recommending an evacu- ation when the risk increases, and by calling for a mandatory evacuation.

An extra element of uncertainty in the preparation for flooding and mass evac- uation pertains to how decision-makers and crisis managers deal with risks and uncertainties. Table 5 presents the chosen strategy in relation to the probability of the occurrence of a disaster when just enough time is available for a preventive evacuation; other strategies (with a lower economic impact) could be considered as well, though. It illustrates clearly that the outcome and the speed of a decision-making process depends on the actual probability of flooding and the risk perception of those involved. Based on the same risk, decision-makers tend to choose a variety of strategies, such as pre- ventive evacuation, vertical evacuation, or shelter in place. When more stakeholders on different levels are involved (plus decision-makers and crisis managers), this auto- matically creates a climate for time-consuming discussions or delayed or contradictory decisions. Thus, different timelines for decision-making have to be taken into account, as well as different strategies because of the possible behaviour of the decision-makers and the consequences of decisions.

Expectations of decision-makers to act as decided

Decision-makers and crisis managers were asked about their willingness to cooper- ate as a citizen with the chosen strategy and how they expected their neighbours to behave. Table 6 shows that approximately 25 per cent of respondents would not react to the evacuation call of the government. Expectations concerning the behaviour of neighbours were more pessimistic. Crisis managers assumed that 48 per cent of the people would not pursue the expected strategy of the government; decision-makers were less pessimistic, expecting 64 per cent of them to respond. The literature also

Table 6. ‘How should you respond as a citizen, as a member of a family, to a call for evacuation by the authorities in a situation when the possibility to evacuate preventively exists but you are ordered to respond alternatively?’

Parameter	Decision-makers		Crisis managers
Response rate of decision-makers: 29% Response rate of crisis managers: 59%	Yes	No	Yes	No
Delay moment of departure for preventive evacuation in favour of other strategies	77%	23%	75%	25%
Shelter in place and prepare yourself in your own house so other, more threatened people can evacuate preventively	76%	24%	79%	21%
How would your neighbour respond? Do you expect him to make the same choices as you and your family?	64%	36%	48%	52%

demonstrates that not all people act as advised by the government (the non-compliance rate). During Hurricane Katrina, for instance, some 20 per cent of the people did not leave the New Orleans area (Wolshon, 2006). Furthermore, the evacuation of the Rivierenland area in 1995 (van Duin et al., 1995; Meurs, 1996) did not lead to the removal of all people. Significant non-compliance rates are also known to exist for hurricanes in the US (between 35 and 64 per cent) (Lindell et al., 2002). Planning has to take into account, therefore, that not all people will comply with the chosen strategy. Realistic planning considers possible scenarios for people who do not comply with an evacuation instruction.

Discussion: how to structure evacuation planning to cope with the possible outcome of decision-making

Time span of the transition phase

During the transition phase the authorities can consider adapting the infrastructure, reallocating means and rescue workers, and informing the public. The created cir- cumstance increase the later effectiveness of emergency measures. The time available for top strategic decision-making and the period of the transition phase cannot be defined in advance because such moments depend on the availability of forecasts and the speed of sense-making by decision-makers and the public and the capacity of the infrastructure. Hence, a delay in deciding to evacuate is also a decision with a poten- tial major impact: emergency measures taken at a later stage could be less effective, not effective, or even counterproductive. Using the same line of argument, Boin et al. (2005) state that a non-decision is equal to a decision. Given the lead time for flood- ing and the slow onset of the event, the alternative strategy of ‘delaying the decision’ should be made explicit to decision-makers, and the consequences should be reviewed. Owing to the different risk perceptions of decision-makers and crisis managers, the uncertainties, and the lack of time, the creation of better circumstances for decision- making is recommended. When the situation is considered directly as a national crisis and crisis management structures are activated to connect initiatives and to identify

realistic measures, the performance rate of an evacuation increases.

The involvement of decision-makers and crisis managers

Another factor that might influence top strategic decision-making is the busy agen- das of decision-makers and policymakers. This was highlighted before Waterproef in preparatory discussions about when to involve decision-makers. Following the detection and recognition of a low probability but high impact threat, experts have to be able to put the warning on the agenda to create the boundary conditions for the start of top strategic decision-making. Of course, this will clash directly with other issues; debates will arise about the need for the action. The uncertainty of the threat and low risk perception mean that there is a risk of delaying or ignoring the warning.

Accountability and the need for probabilistic preparation

Public leaders are expected to take care of citizens in liberal democracies (in contrast to non-democratic societies) (Boin et al., 2005). Although it is clear to all stake- holders that the capacities of the authorities are limited in the case of a mass event such as a flood (BZK, 2008b), the public expects to be warned and it expects emergency measures to be taken to reduce the possible impact. Emergency management, therefore, has to maximise the use of available means and infrastructure to prevent casualties, damage, and the capability to return to a normal situation (resilience). Emergency management has to be able to adapt the environment to an evacuation mode in time.

During the survey, decision-makers and crisis managers addressed the importance of the parameters of leadership and accountability in a false alarm and the impor- tance of risk reduction in a flood. Leadership and accountability are also key drivers (in addition to probabilities, costs, and benefits) in the sphere of preparation. For the Netherlands, as well as parts of Australia and the US, the focus is on preventive evacuation, based on the assumption of a certain window of opportunity. Emergency planning focuses on one deterministic strategy based on a pre-defined event: using a chosen flood event, and after identifying the moment when measures to evacuate (all preventive) have been taken, and by whom, a timeline can be defined to con- nect all decisions at the strategic, tactical, and operational level by all stakeholders (including citizen response). This is deterministic planning, often based on best- case scenarios, which have a lead time that is sufficient to implement all measures. Other realistic events, such as a shorter lead time (Barendregt et al., 2005), a larger threatened area (ten Brinke et al., 2010), and disruptions in the decision-making process, are almost not foreseen in emergency preparation, yet they have to be taken into account. Thus, planning does not cover all possible strategies to minimise loss of life and damage.

Public leaders are also responsible, therefore, for realistic planning. The possible consequences of decisions, as well as uncertainties, should be considered. A leader has to initiate probabilistic planning instead of deterministic planning when uncer- tainties and ambiguity can result in different choices during top strategic decision- making to minimise the risk of loss of life and damage.

Owing to a lack of experience of all possible situations, preparations for mass evacuation should be based on relevant international knowledge and data. This can be combined with expertise in local characteristics to avoid the production of ‘fan- tasy’ documents.

Just as training and education are required to develop a culture of emergency management (Alexander et al., 2009), it is also important for the culture that train- ing and education are realistic. When all decision-makers are trained and educated in preventative evacuation in best-case circumstances, other evacuation strategies, such as vertical evacuation, or even no evacuation, are unlikely because they are unknown to the decision-makers and their crisis managers. This is true even when these other strategies might result in less loss of life in a real-world situation.

The decision taken during Waterproef to delay the evacuation of self-reliant indi- viduals while evacuating non-self-reliant individuals and the families of first responders could be seen as unrealistic planning because the effect was not considered. After the evacuation of 1995, it was concluded that a time-phased evacuation strategy of a threatened area (first this group, then the next) does not work in practice because most people will act primarily based on their own assessment (van Duin et al., 1995). The decision taken during Waterproef, combined with the press conference, in reality might be the signal to start a spontaneous evacuation directly. The result could be a preventive evacuation but without the support of mitigating emergency measures. One can question, therefore, whether the level of preparedness of decision-makers and organisations increased because they were not confronted by the consequences of their decision during the exercise.

The role of decision-makers, crisis managers, and experts

The survey shows that the crisis manager has almost the same perception as the decision-maker. Both support the need for experts to explain the threat and possible responses and strongly depend on them for decision-making. Some differences can be identified, though: decision-makers tend to accord more value to risk-based data (probabilities, consequences in terms of casualties and damage) combined with leadership, whereas crisis managers also attach importance to the perceptions of the public and the media. This might introduce extra elements to the decision-making process, increasing the likelihood of establishing meaning and framing and reduc- ing the possibility of risk-based decision-making.

Many crisis teams from different authorities will be involved in a mass evacuation. It is recommended that the decision-making process be simplified to prevent loss of time owing to interference. Experts could advise decision-makers directly, whereas crisis managers could support this process and implement decisions. Their role is less to inform decision-makers and more to make things work.

Conclusions and recommendations

A traditional view of decision-making is that, given a probable but uncertain threat, there is a deterministic way of defining the best decision. This approach is not suf- ficient for mass evacuation due to potential flooding. Decision-making for mass evacuation cannot be seen as a deterministic process. Different timelines for decision- making have to be taken into account, as do different strategies. Owing to uncer- tainties, including the decision-making process itself, and ambiguity, a probabilistic approach (and the development of event trees) is required. This results in realistic planning. A probabilistic approach connects the measures of several stakeholders.

It is unrealistic to assume that all involved stakeholders (and local decision-makers in this network) will act in accordance with the strategy chosen by the top strategic decision-maker. In addition, the time needed for the decision-making process cannot

be determined in advance, and it depends on the involved decision-makers. That time is limited and it might exclude some evacuation strategies (as forms of preventive evacuation) or make them less effective, leading to an increase in loss of life or damage. When ambiguity is considered as a linguistic problem it can be prevented. However, because of low risk perception, the struggle to keep decision-makers involved and the low frequency of event ambiguity have to be taken into account in emergency planning. These variables could influence the effectiveness of strategies or result in

different strategies.

This does not mean that all deterministic planning documents are not realistic. Different strategies could be chosen based on the same information, depending on the decision-makers involved. In addition, local circumstances in a society also influ- ence the decision-making process. When uncertainties are considered, one can conclude that a deterministic approach covers all possible events. As shown in the survey, though, the impact of uncertainties, including the role of decision-makers, necessitates a probabilistic approach to evacuation planning when the objective is to minimise loss of life and damage.

The Waterproef case study demonstrates a lack of attention to the consequences of top strategic decisions and the impact of evacuation on effectiveness and relations with other stakeholders. One can question, therefore, whether such exercises increase or decrease preparedness when no feedback is forthcoming about the comparison between the chosen strategy and actual or planned consequences.

Top strategic decision-making focuses on creating optimal conditions for a response by all stakeholders to a future threat. These emergency measures have to be imple- mented in the ‘transition phase’, to create boundary conditions for future responses. To support top strategic decision-making, the situation has to be considered directly as a crisis, even in the case of a low probability threat. The top strategic decision- maker has to initiate crisis management mechanisms and involve other local decision- makers to enhance the possibility of connected planning. This is a top-down approach that differs from more classic ways of initiating crisis management mechanisms (using a bottom-up approach). When clear visual signals of nearby events are unavailable and public awareness is minimal, the decision has to be taken by decision-makers who are informed by experts. A delay in decision-making might be realistic, but it

can also result in a less effective response.

The study shows that the most important information variables for top strategic decision-making on evacuation are the probability of flooding, the consequences in terms of loss of life and possible economic damage, and leadership and accountability. Accountability with regard to taking care of citizens when things go wrong is also one of the requirements of realistic evacuation planning in a democratic country. Preparations should aim to maximise available means and infrastructure to reduce the risk of loss of life and damage.

This research illustrates that crisis managers have almost the same perceptions as decision-makers. It is recommended that the decision-making process for a mass evac- uation be simplified to prevent loss of time and to address properly risks, costs and

benefits, and uncertainties. It is suggested that experts be added directly to teams of decision-makers, whereas crisis managers should focus on supporting this process and implementing decisions.

Correspondence

Bas Kolen, HKV Consultants, PO Box 2120, 8203 AC Lelystad, The Netherlands. Telephone: +31 320 294231/+31 6 10962684; e-mail: Kolen@hkv.nl

Endnotes

¹ For more background information, see the official evaluation of the exercise (Cappelleveen and Ven, 2009; TMO, 2009c) and the response of the Cabinet on flood preparedness (Ministry of the Interior and Kingdom Relations, 2009).

References

Aldunate, R.G., F. Pena-Mora, and G.E. Robinson (2005) ‘Collaborative distributed decision making for large scale disaster relief operations: drawing analogies from robust natural systems’. Complexity. 11(2). pp. 28–28.

Alexander, D. (1993) Natural Disasters. UCL Press, London.

Alexander, D. (2000) ‘Scenario methodology for teaching principles of emergency management’.

Disaster Prevention and Management. 9(2). pp. 89–97.

Alexander, D. (2004) ‘Vulnerability to landslides’. In T. Glade, M.G. Anderson, and M.J. Crozier (eds.) Landslide Hazard and Risk. Wiley, Chichester. pp. 175–198.

Alexander, D., L. Bramati, and M. Simonetta (2009) ‘Emergency preparedness training and education in Lombardy region, Italy: survey of supply and demand’. Natural Hazards Review. 10(3). pp. 77–83. Baker, E.J. (1991) ‘Hurricane evacuation behavior’. International Journal of Mass Emergencies and Disasters.

9(2). pp. 287–310.

Barendregt, A., J.M. van Noortwijk, M. van der Doef, and S.R. Holterman (2005) Determining the time available for evacuation of a dike-ring area by expert judgement. ISSH – Stochastic Hydraulics 2005. 23 and 24 May. http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.65.6439&rep=

rep1&type=pdf (accessed on 18 March 2014).

Bedford, T. and R. Cooke (2001) ProbabilisticRiskAnalyses:FoundationsandMethods. Cambridge University Press, Cambridge.

Boin, A., P. ’t Hart, E. Stern, and B. Sundelius (2005) The Politics of Crisis Management: Public Leadership under Pressure. Cambridge University Press, Cambridge.

Bourque, L.B., J.M. Siegel, M. Kano, and M.M. Wood (2006) ‘Weathering the storm: the impact of hurricanes on physical and mental health’. The Annals of the American Academy of Political and Social Science. 604(1). pp. 121–159.

Brugnach, M., A. Dewulf, C. Pahl-Wostl, and T. Taillieu (2008) ‘Toward a relational concept of uncertainty: about knowing too little, knowing too differently, and accepting not to know’. Ecology and Society. 13(2). http://www.ecologyandsociety.org/vol13/iss2/art30/ (accessed on 18 March 2014). BZK (Ministry of the Interior and Kingdom Relations) (2007) National Crisis Plan: Extreme Water

Levels and Flooding (in Dutch). BZK, The Hague.

BZK (2008a) National Crisis Plan: Extreme Water Levels and Flooding. Communication Strategy (in Dutch). BZK, The Hague.

BZK (2008b) Kamerstukken II, 2007–2008, 30821 nr. 6. BZK, The Hague.

BZK (2008c) Nationaal Crisisplan Hoogwater en Overstromingen. BZK, The Hague.

BZK and VenW (Ministry of Transport, Public Works, and Water Management (2008) Exercise Waterproef 3 November: Coastal Flooding. Start Information. BZK, The Hague.

Cappelleveen, E. and J. Ven (2009) Evaluation Exercise Waterproef: Summary (in Dutch). Twynstra and Gudde, Amersfoort.

CNN (Cable News Network) (2005) ‘New Orleans braces for monster hurricane’. 29 August. http:// www.cnn.com/2005/WEATHER/08/28/hurricane.katrina/ (accessed on 18 March 2014).

CNN (2006) ‘New Orleans disaster plan: get out of town’. http://www.cnn.com/2006/US/05/02/ hurricane.plan/index.html (accessed on 18 May 2014).

Cole, J. (2008) Hurricane Gustav – Testing the Lessons Learned from Katrina. RUSI Analysis. 3 September. https://www.rusi.org/analysis/commentary/ref:C48BEB8D3402BE/ (accessed on 18 May 2014). Dewulf, A., M. Craps, R. Bouwen, T. Taillieu, and C. Pahl-Wostl (2005) ‘Integrated management of natural resources: dealing with ambiguous issues, multiple actors and diverging frames’. Water,

Science and Technology. 52(6). pp. 115–124.

DHS (Department of Homeland Security) (2006) Mass Evacuation: Developing a Contraflow Plan. DHS, Washington, DC

Dow, K. and C.L. Cutter (2000) ‘Public orders and personal opinions: household strategies for hurri- cane risk assessment’. Environmental Hazards. 2. pp. 143–155.

Fjellman, S.M. (2976) ‘Natural and unnatural decision-making: a critique of decision theory’. Ethos.

4(1). pp. 73–94.

Grothmann, T. and F. Reusswig (2006) ‘People at risk of flooding: why some residents take precau- tionary action while others do not’. Natural Hazards. 38(1–2). pp. 101–120.

Gruntfest, E.C. and K. Carsell (2000) The Warning Process: Toward an Understanding of False Alarms. University of Colorado at Colorado Springs, Colorado Springs, CO.

Hulpverleningsregio Haaglanden (2008) Coördinatieplan Overstromingen Haaglanden, Versie 1.3, 24 April

2008 (in Dutch). Hulpverleningsregio Haaglanden, The Hague.

Haynes, K. et al. (2009a) ‘Shelter-in-place versus evacuation in flash flood environments’. Paper presented at the ‘49th Annual floodplain management authorities conference and 6th biennial Victorian flood conference’, Albury, Australia, 17–20 February 2009. http://floods.org.au/wp- content/uploads/Session-7-paper-23-Coates.pdf (accessed on 18 March 2014).

Haynes, K. et al. (2009b) ‘Shelter-in-place versus evacuation in flash floods’. Environmental Hazards.

8(4). pp. 291–303.

Helsloot, I. and A.C.J. Scholtens (2007) Description of National Crisis Management (in Dutch). BZK, The Hague.

Helsloot, I., A.C.J. Scholtens, and E. Warners (2009) ‘Leerpunten van nationale crisisoefeningen’ (in Dutch). Recht, bestuur en organisatie van hulpdiensten. 6(4). pp. 154–165.

HKV Consultants (2010) Figures for Evacuation Cost–Benefit Analyses. Memorandum PR1919-10 (in Dutch). HKV Consultants, Lelystad.

Jong de, M. and I. Helsloot (2010) ‘The effects of information and evacuation plans on civilian response during the national Dutch flooding exercise “Waterproef”’. In S.P. Hoogendoorn et al. (eds.) First International Conference on Evacuation Modelling and Management. Elsevier, The Hague. pp. 153–162. Jonkman, S.N. (2007) Loss of Life Estimation in Flood Risk Assessment. Theory and Applications. PhD thesis.

Delft University of Technology, Delft.

Jonkman, S.N. and I. Kelman (2005) ‘An analysis of the causes and circumstances of flood disaster deaths’. Disasters. 29(1). pp. 75−97.

Kolen, B. and I. Helsloot (2012) ‘Time needed to evacuate the Netherlands in the event of large- scale flooding: strategies and consequences’. Disasters. 36(4). pp. 700–722.

Kolen, B., C. Vermeulen, and Y. van Bokkum (2008) Interdepartmental Operational Guideline Waterproef, Version 1.0. (in Dutch). Project 13. HKV lijn in water, Lelystad.

Kolen, B., J.K. Leenders, J. Van Der Schaaf, and D.M. Van Dijk (2007) Roles of National Governmental Organisations in Case of Mass Evacuation in the Netherlands. Guidelines for Evacuation and an Emergency Plan (Coordination and Decision-making) for the National Level (in Dutch). HKV, Oranjewoud Save, Lelystad. Kolen, B., M. Bos, J.M. Zoethout, S. Nieuwenhuis, and K. Van Ruiten (2011) ‘The use of different flood scenarios including worst cases for emergency planning and flood preparedness’. Paper presented at the Fifth International Conference on Flood Management (ICFM5), Tokyo, Japan,

27–29 September 2011.

Kolen, B., M. Kok, I. Helsloot, and B. Maaskant (2013) ‘EvacuAid: a probabilistic model to determine the expected loss of life for different mass evacuation strategies during flood threats’. Risk Analysis. 33(7). pp. 1312–1333.

Kunreuther, H. et al. (2002) ‘High stakes decision making: normative, descriptive and prescriptive considerations’. Marketing Letters. 13(3). pp. 259–268.

Lindell, M.K. and R. Perry (1992) Behavioral Foundations of Community Emergency Planning. Hemisphere, Washington, DC.

Lindell, M.K., C.S. Prater, R.W. Perry, and J.Y. Wu (2002) EMBLEM: An Empirically Based Large-scale Evacuation Time Estimate Model. Hazard Reduction and Recovery Center, Texas A&M University, College Station, TX.

Lindell, M.K., J. Lu, and C. Prater (2005) ‘Household decision making and evacuation in response to Hurricane Lili’. Natural Hazards Review. 6(4). pp. 171–179.

LOCC (National Operational Centre) (2008) National Operational Plan: Evacuation for Exercise Waterproef

(in Dutch). LOCC, Driebergen.

Maaskant, B., B. Kolen, R. Jongejan, S.N. Jonkman, and M. Kok (2009) Fractions of Preventive Evacuation in Case of Flooding in the Netherlands (in Dutch). HKV Consultants, Lelystad.

Meurs, R. van (1996) Extreme Water Levels. The Power of Farmer Organisations in River Delta (in Dutch). Scheffers, Herwijnen.

Ministry of the Interior and Kingdom Relations and Ministry of Transport, Public Works, and Water Management (2005) Strategies for Flood Disaster Control along the Rivers Rhine and Meuse (in Dutch). Ministry of the Interior and Kingdom Relations and Ministry of Transport, Public Works, and Water Management, The Hague.

Ministry of the Interior and Kingdom Relations and Ministry of Transport, Public Works, and Water Management (2006) Cabinet Standpoint Disaster Management Floods (in Dutch). Ministry of the Interior and Kingdom Relations and Ministry of Transport, Public Works, and Water Management, The Hague.

Ministry of the Interior and Kingdom Relations and Ministry of Transport, Public Works, and Water Management (2009) Reaction of the Dutch Cabinet on the Results of Taskforce Management Flooding (in Dutch). Ministry of the Interior and Kingdom Relations and Ministry of Transport, Public Works, and Water Management, The Hague.

Ministry of Transport, Public Works, and Water Management (2008) Emergency Plan: ‘Extreme Water Level and Storm Surge’. Guidelines for a National Approach (in Dutch). Ministry of Transport, Public Works, and Water Management, The Hague.

Ministry of Transport, Public Works, and Water Management (2009) National Situational Report. Flooding: Roles and Responsibilities of the National Commission of Flooding. HKV lijn in water, Lelystad. Mitchell, J.T., S.L. Cutter, and A.S. Edmonds (2007) ‘Improving shadow evacuation management: case study of the Graniteville, South Carolina chlorine spill’. Journal of Emergency Management. 5(1).

24. 28–24.

Parker, C., E. Stern, E. Paglia, and C. Brown (2009) ‘Preventable catastrophe? The Hurricane Katrina disaster revisited’. Journal of Contingencies and Crisis Management. 17(14). pp. 206–220.

Rasmussen, J., B. Brehmer, and J. Leplat (1991) Distributed Decision Making: Cognitive Models for Cooperative Work. John Wiley and Sons, Chichester.

RIVM (National Institute for Public Health and the Environment) (2004) Dutch Dikes and Risk Hikes. A Thematic Policy Evaluation of Risks of Flooding in the Netherlands. RIVM, Bilthoven.

Rosenthal, U., M.T. Charles, and P.T. Hart (1989) Coping with Crisis: The Management of Disasters, Riots and Terrorism. Charles C. Thomas, Springfield, IL.

Schneeweiss, C. (2003) ‘Distributed Decision Making – A unified approach’. European Journal of Operational Research. 150(2). pp. 237–252.

Sorensen, J.H. (2000) ‘Hazard warning systems: review of 20 years of progress’. Natural Hazards Review. 1(2). pp. 119–125.

South-Holland, Province et al. (2010) Coordination plan dikering area 14, February 2010 (in Dutch). ten Brinke, W.B.M., B.A. Bannink, and W. Ligtvoet (2008a) ‘The evaluation of flood risk policy in

the Netherlands’. Proceedings of the ICE – Water Management. 161(4). pp. 181–188.

ten Brinke, W.B.M., G.E.M. Saeijs, I. Helsloot, and J. van Alphen (2008b) ‘Safety chain approach in flood risk management’. Municipal Engineer.161(4). pp. 181–188.

ten Brinke, W.B.M., B. Kolen, A. Dollee, H. Van Waveren, and K. Wouters (2010) ‘Contingency planning for large-scale floods in the Netherlands’. Journal of Contingencies and Crisis Management. 18(1). pp. 55–69.

Terpstra, T. (2009) Flood Preparedness: Thoughts, Feelings and Intentions of the Dutch Public. University of Twente, Enschede.

TMO (Taskforce Management Flooding) (2009a) First Impression: Exercise Waterproef. TMO, The Hague. TMO (2009b) Final Report (in Dutch). TMO, The Hague.

TMO (2009c) First Impression Exercise Waterproef (Task Force Management Flooding). The Hague. Tversky, A. and D. Kahneman (1974) ‘Judgment under uncertainty: heuristics and biases’. Science.

185(4157). pp. 1124–1131.

Urbina, E. and B. Wolshon (2003) ‘National review of hurricane evacuation plans and policies: a comparison and contrast of state practices’. Transportation Research Part A: Policy and Practice. 37(1). pp. 257–275.

van Duin, M.J., M.J. Bezuyen, and U. Rosenthal (1995) Evacuation in Case of Extreme Water Levels: Self-reliance and Care of Authorities. COT, University of Leiden andErasmus University Rotterdam, Leiden.

van Noortwijk, J. and A. Barendregt (2004) Available and Necessary Time for Evacuation in Case of Possible Flooding (in Dutch). HKV lijn in water, Lelystad.

van Zuilekom, K.M., M.F.A.M. van Maarseveen, and M.R. van der Doef (2005) ‘A decision support system for preventive evacuation of people’. In P. Van Oosterom, S. Zlatanova, and E.M. Fendel (eds.) Geo-information for Disaster Management. Springer Berlin, Heidelberg. pp. 229–253.

Veiligheidsregio Midden- en West-Brabant (2008) Regionaal Basisplan Overstromingen, Samen voorbereiden op overstromingen, versie 1.0 juni 2008 (in Dutch).

Veiligheidsregio Zeeland (2008) Strategisch Plan Deel 1:Strategisch beeld Fase 1, Dag -5 tot en met dag 0. Veiligheidsregio Zeeland, Middelburg.

Vrijling, J.K. (2009) ‘The lesson of New Orleans’. In M.J. Kallen and S.P. Kuniewski (eds.) Risk and Decision Analysis in Maintenance Optimization and Flood Management. IOS Press, Amsterdam. pp. 57–69. Wegh, E. (2008) National Concept. Traffic Management for Exercise Waterproef (in Dutch). Ministry of

Transport, Public Works, and Water Management, Utrecht.

Wolshon, B. (2006) ‘Evacuation planning and engineering for Hurricane Katrina’. The Bridge. 36(1). pp. 27–34.

Wolshon, B., E. Urbina, C. Wilmot, and M. Levitan (2005) ‘Review of policies and practices for hurricane evacuation. I: transportation planning, preparedness, and responses’. Natural Hazards Review. 6(3). pp. 129–142.

ersie 1.0 juni 2008 (in Dutch).

Veiligheidsregio Zeeland (2008) Strategisch Plan Deel 1:Strategisch beeld Fase 1, Dag -5 tot en met dag 0. Veiligheidsregio Zeeland, Middelburg.

Vrijling, J.K. (2009) ‘The lesson of New Orleans’. In M.J. Kallen and S.P. Kuniewski (eds.) Risk and Decision Analysis in Maintenance Optimization and Flood Management. IOS Press, Amsterdam. pp. 57–69. Wegh, E. (2008) National Concept. Traffic Management for Exercise Waterproef (in Dutch). Ministry of

Transport, Public Works, and Water Management, Utrecht.

Wolshon, B. (2006) ‘Evacuation planning and engineering for Hurricane Katrina’. The Bridge. 36(1). pp. 27–34.

Wolshon, B., E. Urbina, C. Wilmot, and M. Levitan (2005) ‘Review of policies and practices for hurricane evacuation. I: transportation planning, preparedness, and responses’. Natural Hazards Review. 6(3). pp. 129–142.