Abstract
Risk theories and empirical research indicate that a variety of factors can influence people’s protective decisions for natural hazards. Using data from an online survey that presented coastal U.S. residents with a hypothetical hurricane scenario, this study investigates the relative importance of cognitive risk perceptions, negative affect, efficacy beliefs, and other factors in explaining people’s anticipated evacuation decisions. The analysis finds that multiple factors, including individual and household characteristics, previous experiences, cognitive and affective risk perceptions, and efficacy beliefs, can help predict hurricane evacuation intentions. However, the largest amount of variance in survey participants’ evacuation intentions is explained by their evacuation-related response efficacy (coping appraisals) and their perceived likelihood of getting hurt if they stay home during the storm. Additional analysis explores how risk perceptions and efficacy beliefs interact to influence people’s responses to risk information. Although further investigation in additional situations is needed, these results suggest that persuading people at high risk that evacuating is likely to reduce harm can serve as an important risk communication lever for motivating hurricane evacuation.
Similar content being viewed by others
Introduction
With increasing risk of hurricane impacts along U.S. coastlines, motivating populations at high risk to evacuate when needed remains both important and challenging. In areas where an approaching hurricane poses a threat to life and safety, public officials typically recommend that people move to a safer location before the arrival of hazardous weather. Yet some people in areas at high risk do not or cannot leave, which can have devastating consequences. In 2021 during Hurricane Ian, for example, more than 40 people drowned due to storm surge flooding, and many more experienced physical injuries or traumatic life-threatening situations1,2,3.
Many previous studies have examined why some people evacuate when a hurricane threatens, while other people do not. This body of research finds that multiple factors, including messages received, individual and household characteristics, past experiences, evacuation capacity and barriers, social influences, and risk perceptions, can influence people’s hurricane evacuation decisions (see, e.g., reviews in refs. 4,5,6,7). However, which factors are most influential varies among studies. This variation arises in part because studies examine a variety of real and hypothetical hurricane situations, and people vary in their hurricane-related experiences, vulnerabilities, capacities, and responses to risks4,5,6,7,8,9,10,11,12,13,14,15. In addition, different studies include different sets of possible influences on evacuations and operationalize variables differently. For example, as discussed further below, researchers measure concepts such as hurricane experience and risk perceptions in a variety of ways4,5,6,7,16, which further complicates understanding the primary factors driving hurricane evacuation decisions.
Along with empirical studies, there are multiple theories of how people respond to information about threats and what determines protective behaviors. These include Protection Motivation Theory (PMT17,18), the Extended Parallel Process Model (EPPM19,20), and the Protective Action Decision Model (PADM21,22). Although the specifics vary, each of these theories suggests that protective decision making is influenced by people’s risk perceptions (also called threat appraisals or threat perceptions) and their efficacy beliefs (also called coping appraisals or perceptions of protective actions), as illustrated in Fig. 1. In the PMT and EPPM, risk perceptions include cognitive risk perceptions, conceptualized as perceived likelihood and severity of the threat, and fear, a type of affective risk perceptions (also called affective responses or emotional appraisals). Efficacy beliefs include beliefs about how effective an action is in protecting against a threat (response efficacy) and about one’s ability to perform a protective action (self-efficacy). (See, e.g., refs. 12,15,16,17,19,23,24) These theories also suggest that people facing risks can engage in protective responses, such as deciding to evacuate, and other (non-protective) responses, such as defensive reactions (Fig. 1; see Results).
The top-left box indicates the hurricane risk information that survey respondents received in the experimental module studied here. Solid arrows indicate relationships explored in this article; dashed arrows indicate relationships anticipated based on theory and prior empirical research but not directly investigated here.
The concepts and relationships posited by these theories have been tested extensively in the context of health messaging and behaviors. Meta-analyses of health studies find that both risk and efficacy messaging, as well as the risk perceptions and efficacy beliefs that such messages can elicit, influence people’s protective decisions and other responses to risks25,26,27,28,29. The roles of risk perceptions and efficacy beliefs in decisions have also been investigated in other contexts, including climate adaptation30,31,32,33 and long-term protection from natural, technological, and human-caused hazards34,35,36,37,38,39. Although many of these studies find that risk perceptions are important, some find that efficacy beliefs can have stronger effects on protective decisions25,26,30,31,34,38,39,40,41.
While similar to hurricane evacuation in some respects, the types of decision contexts discussed above involve different considerations and dynamics than near-term decisions for approaching hazards42,43. A number of hurricane studies have examined risk perceptions, with most finding that they are related to people’s evacuation decisions4,5,6,7,8,16,44,45. This has important practical implications because risk perceptions can be influenced by risk communication, which forecasters and public officials can shape. However, because risk perception is a broad construct that includes multiple interrelated attitudes, beliefs, and feelings46,47,48,49, researchers use a variety of hurricane risk perception measures. These include perceptions of storm characteristics and associated wind and flood hazards, perceived likelihood and severity of a hurricane event, perceived impacts and safety, and affective responses such as concern and fear5,7,12,50,51,52,53,54,55.
Research also shows that people’s capacity to take protective actions—including barriers or impediments such as lack of available shelter or transportation, limited funds, work, pets, or household members with disabilities—influences their protective action decisions7,9,52,53,54,56,57. A few studies have investigated the role of efficacy beliefs in the context of near-term hazard threats16,58,59, but less is known about how different types of efficacy influence hurricane evacuation decisions. Thus, improving understanding about how different aspects of risk perceptions, efficacy beliefs, and other factors influence evacuation decisions can help improve hurricane risk communication.
To help address these knowledge gaps, this study investigates the roles of different situation-specific and general hurricane-related factors in explaining evacuation decisions, using data from a survey that presented respondents with information about a hypothetical scenario of an approaching hurricane (see Methods; Fig. 1). The situation-specific variables investigated here measured multiple aspects of respondents’ perceptions, beliefs, and anticipated behaviors related to the specific hurricane situation presented (Table 1). This includes respondents’ cognitive risk perceptions related to the overall hurricane threat (measured here as their perceived likelihood of their home being affected, severity of the hurricane at their home, and likelihood of getting hurt if they stay home) along with their perceptions of whether the hurricane poses a threat to their home from three hurricane-related geophysical hazards (strong winds, storm surge flooding, and rain flooding). It also includes their affective risk perceptions (measured as fear and worry) and efficacy beliefs (measured as evacuation-related self-efficacy and response efficacy). The general hurricane-related variables investigated here measured respondents’ perceptions, preparations, and experiences related to hurricane risks in general, outside the context of the specific hypothetical hurricane situation presented. This includes factors that previous research has found to influence hurricane evacuation decisions, such as perceived residence in a flood or evacuation zone, prior evacuation planning, prior hurricane evacuation, and other related hurricane-related experiences5,6,7,16,45,53.
This article analyzes these data to investigate three research questions:
-
RQ1.
To what extent do different types of situation-specific risk perceptions and efficacy beliefs predict evacuation intentions?
-
RQ2.
How do situation-specific risk perceptions and efficacy beliefs interact in influencing evacuation intentions?
-
RQ3.
How do situation-specific perceptions and beliefs compare with general hurricane-related factors as predictors of evacuation intentions?
Related to RQ2, we also explore whether high hurricane risk perceptions are a prerequisite to high response efficacy and protective responses, and whether these data suggest that people with high risk perceptions and low efficacy tend to engage in non-protective rather than protective responses. In addition, we explore what might underlie respondents’ response efficacy as measured here. Figure 1 depicts a simplified version of the conceptual model that informs our analysis, synthesized from the theories and prior research results discussed above.
By investigating the extent to which different situation-specific and other factors directly influence evacuation intentions, we aim to improve understanding about what can help motivate people at risk from hurricanes to take protective actions. More broadly, by investigating different types of risk perceptions along with efficacy beliefs, we aim to develop new knowledge about how different dimensions of these concepts interact with natural hazard decision making. Our goal is to explore these topics within the scope of the hypothetical hurricane situation and measures used in this survey. The findings presented here can inform future empirical research by illustrating the importance of studying multiple dimensions of risk perceptions along with situation-specific efficacy beliefs. They also help advance understanding about how existing risk theories apply in the context of near-term, approaching natural hazards.
Results
Situation-specific perceptions and beliefs as predictors of hurricane evacuation intentions
To investigate RQ1, Table 2 shows the results from a series of regression analyses investigating which of the situation-specific perceptions and beliefs investigated here help predict evacuation intentions, and how much. As shown in the top-left box of Fig. 1 and described in Methods, all respondents received three types of information about the hurricane risk: (1) an introduction to the scenario, (2) experimentally manipulated message conditions, in which different respondents received different combinations of hurricane risk messages, and (3) information about evacuation. Model 1 shows that, as found in ref. 12, the individual/household characteristics and the three experimentally manipulated message conditions explain only a small amount (2%) of the variance in evacuation intentions in these data. Thus, while we include the experimental message conditions as controls in subsequent regression analyses, we focus on the other variables investigated, and we interpret survey participants’ responses to the hurricane scenario in terms of the information that was received by all respondents.
Models 2–5 test adding different sets of situation-specific perceptions and beliefs as predictors in Model 1. In Model 2, adding the three cognitive risk perception variables related to the overall (cross-hazard) hurricane threat explains an additional 58% of the variance in evacuation intentions—considerable explanatory power for this type of data. Model 5 shows that adding the two efficacy variables explains even more of the variance in evacuation intentions: 67%. Models 3 and 4 indicate that respondents’ perceptions of which hurricane hazards are a threat to their home in this situation can also help explain evacuation intentions, as can their affective risk perceptions (worry and fear). However, these latter two sets of variables have less explanatory power than the other three cognitive risk perceptions or the efficacy beliefs.
Model 6 includes all of the situation-specific variables in Models 2–5 as predictors of evacuation intentions, in the same regression model. In this analysis, the three cognitive risk perception variables related to the overall threat and the two efficacy variables remain statistically significant predictors, and the hurricane hazard and affective risk perception variables are not (p = 0.07–0.94). Respondents’ evacuation-related response efficacy is the strongest predictor of their evacuation intentions, followed by their perceived likelihood of getting hurt, perceived severity of the hurricane at their home, and evacuation-related self-efficacy (see Supplementary Table S1 for standardized regression coefficients). Model 6 explains 73% of the variance in evacuation intentions, a small increase from Model 5, which included the two efficacy variables but none of the risk perception variables.
Together, these results indicate that, consistent with some of the research discussed in the introduction, respondents’ perceptions of what geophysical hazards are a threat to their home from an approaching hurricane can help predict their evacuation intentions, as can their worry and fear about the situation. However, these variables are no longer predictors when overall cognitive risk perceptions and efficacy beliefs are included in the same regression analysis. This suggests shared variance, or possibly mediated relationships, in which respondents’ hazard perceptions and negative affect influence evacuation intentions indirectly through their influence on other cognitive risk perceptions and efficacy beliefs. Given the goals of this article, we do not test these types of more complex relationships explicitly; however, they have been found in other studies54,60.
Interactions among situation-specific risk perceptions and efficacy beliefs
The three risk theories summarized in the introduction also posit more specific relationships among risk perceptions, efficacy beliefs, and decisions. According to the PMT, EPPM, and PADM, when a person receives information about a potential threat, they first appraise the perceived risk. If perceived risk is sufficiently low, then no protective action needs to be considered. If perceived risk is sufficiently high, then the person initiates a second appraisal, this time of protective actions that may alleviate the risk. Although the specifics differ, in all three theories this second appraisal includes constructs related to response efficacy and self-efficacy. The two appraisals then combine to influence if and how the individual responds to the threat.
If a person’s perceived risk and efficacy are both sufficiently high, theory predicts that they will be motivated to take a protective response, which may then lead to engaging in protective behaviors such as evacuation. Some empirical studies in other contexts have found these predicted interactions between risk perceptions and efficacy beliefs in predicting protective responses, while other studies find no or alternative interactions27,28,29,30,34,38,61,62,63,64.
In this section, we investigate how situation-specific risk perceptions and efficacy beliefs interact in influencing evacuation intentions—RQ2—using two approaches. Building on the results in the previous section, the first approach examines these interactions statistically using multiple linear regression models (Table 3). We also examine several interactions in greater detail by analyzing how evacuation intentions vary across a range of combinations of situation-specific variables (Fig. 2). These more in-depth analyses investigate whether these data exhibit the types of non-linear or threshold effects posited by the EPPM, as discussed above. Other (e.g., non-protective) responses to the hurricane risk information presented are discussed in a later section.
Left panels: Matrices depicting mean evacuation intentions for respondents with different combinations of situation-related response efficacy and (a) self-efficacy, (c) perceived likelihood of getting hurt, (e) perceived severity at home, or (g) perceived likelihood of home affected. The background of each cell is colored on a yellow (low) to green (high) scale based on the value of mean evacuation intentions. The font for the numbers in each cell is colored gray (low) to black (high) based on the number of respondents with the variable combination represented by that cell, in other words, based on the N used to calculate that cell’s mean evacuation intentions; cells with N < 5 are left blank. Right panels: Box and whisker plots depicting the same data as in the left panels, to illustrate variability across respondents. Mean and median evacuation intentions, along with the interquartile range (IQR), whiskers representing 1.5 * IQR, and inner and outlier points, are shown for respondents with different combinations of situation-related response efficacy and (b) self-efficacy, (d) perceived likelihood of getting hurt, (f) perceived severity at home, or (h) perceived likelihood of home affected. For clarity, risk perceptions and efficacy beliefs are compacted from 7 to 4 categories.
To investigate interactions statistically, we use a more parsimonious version of Model 6, Model 6a, which is described in Methods. When two-way interactions among the three cognitive risk perception variables are added to Model 6a, along with interactions between the two efficacy variables, two of the four interactions are statistically significant: Severity at home * Likelihood of getting hurt (−0.049, p < 0.001) and Self-efficacy * Response efficacy (0.041, p < 0.001). When two-way interactions between each of the three cognitive risk perception variables and each of the two efficacy variables are added to Model 6a, only one of the six interactions is statistically significant: Likelihood of getting hurt * Response efficacy (−0.061, p < 0.001). When these three interactions are added to Model 6a together, Severity at home * Likelihood of getting hurt is no longer statistically significant (−0.019, p = 0.09). Thus, our final model with interactions is Model 6b, with two interactions as shown in Table 3. Including these interactions does not change the variance explained by the model; however, along with the additional analyses depicted in Fig. 2, it does help elucidate several of the variables’ effects.
The Self-efficacy * Response efficacy interaction in Table 3 is positive. This indicates that the influence of self-efficacy on evacuation intentions tends to be greater for respondents with higher response efficacy, and vice versa. More specifically, Fig. 2a, b shows that among respondents with high to very high response efficacy, higher self-efficacy is associated with higher evacuation intentions. For those with low to moderate response efficacy, however, self-efficacy has limited influence on evacuation intentions. In other words, the analysis in Fig. 2 indicates that if respondents do not believe that evacuating is effective, they are unlikely to evacuate regardless of their belief in their ability to evacuate. In contrast, although few respondents reported low self-efficacy, higher response efficacy is associated with higher evacuation intentions across all levels of self-efficacy.
The Likelihood of getting hurt * Response efficacy interaction in Table 3 is negative. This indicates that the influence of response efficacy on evacuation intentions tends to be smaller for respondents with higher perceived likelihood of getting hurt, and vice versa. More specifically, Fig. 2c, d shows that respondents have low evacuation intentions if their response efficacy and perceived likelihood of getting hurt are both low, and moderate to high evacuation intentions if either is high. Note, however, that few respondents reported high perceived likelihood of getting hurt and low response efficacy.
The analyses in Fig. 2c–h also show an additional result that is not evident in the regression models. For each of the three cognitive risk perception variables related to the overall hurricane threat, the results on the left-hand side of Fig. 2c–d, e–f, and g–h illustrate that even respondents with low risk perceptions can have high response efficacy and high evacuation intentions. This is counter to the predictions of the risk theories discussed above, which suggest that if risk perceptions are low, people will not consider taking protective action. Instead, even among respondents with low risk perceptions, higher response efficacy is associated with higher likelihood of evacuating.
More generally, the yellow to green gradation from bottom to top within Fig. 2a, c, e, and g again illustrates the strong effect of respondents’ evacuation-related response efficacy on their evacuation intentions. Figure 2b, d, f, and h show that some respondents with high response efficacy have low evacuation intentions, but most do not. This dominant role of efficacy across levels of risk perceptions differs from discussions in much of the literature described above, although it has been observed in other contexts30,41.
Comparison of situation-specific with general hurricane-related variables as predictors of evacuation intentions
Next, we investigate RQ3, using Models 7 and 8 in Table 4. Model 7 tests adding to Model 1 six of the general hurricane-related variables that the survey measured outside the context of a specific approaching hurricane scenario. The results show that, as found in ref. 12, these factors—respondents’ perceptions of whether their home is in a flood or evacuation zone, prior evacuation planning, and experiences with home flooding and Hurricane Sandy—can help explain their evacuation intentions in the hurricane scenario studied here. This is consistent with other studies’ findings that these types of perceptions, preparations, and experiences can help predict people’s protective behaviors during natural hazard threats5,6,7,65. Note, however, that the adjusted R2 for Model 7 is 0.14, compared to, e.g., 0.73 in Model 6a in Table 3. In other words, these general hurricane-related factors explain much less of the variance in evacuation intentions than the situation-specific cognitive risk perception and efficacy belief variables investigated above.
To compare the explanatory power of these general hurricane-related factors with situation-specific perceptions and beliefs more directly, Model 8 includes the variables in Models 7 and 6a in the same regression analysis. As in Table 2, adding the situation-specific variables as predictors results in a large increase in the variance explained by the regression model. Of the six general hurricane-related variables that were predictors in Model 7, only two remain statistically significant (p < 0.05): whether respondents thought they were in a flood zone and whether they said they evacuated for Hurricane Sandy before landfall. The other four general hurricane-related variables are no longer direct predictors of evacuation intentions. Again, this suggests that these factors may influence evacuation intentions in other ways, such as indirectly through their influence on situation-specific risk perceptions and efficacy beliefs. Although we do not test these mediated paths explicitly, they are consistent with some other research16,35,39,54.
Note that the two general hurricane-related variables that remain direct predictors in Model 8 are both dichotomous (0 or 1), whereas the cognitive risk perception and efficacy belief variables are on a 1–7 scale. This, along with the standardized coefficients for Model 8 shown in Supplementary Table S2, provides further evidence that the situation-specific cognitive risk perceptions and efficacy beliefs are stronger direct predictors of respondents’ evacuation intentions than the general hurricane-related perceptions, preparations, and experiences measured in this survey.
Other responses to hurricane risks
Along with protective responses such as evacuation, the three theories discussed earlier in this article include the potential for people to engage in other, non-protective responses to risks (Fig. 1). These non-protective responses—which are also referred to as maladaptive responses or emotion-focused coping—include defensive avoidance, denial, or negative reactance, e.g., minimizing the information as “overblown” or perceiving manipulation through “misleading” information60,66. The EPPM posits more specifically that when perceived risk elicits sufficient fear, but efficacy is low, a person will engage in non-protective responses to control their fear—and there may even be a boomerang effect, where they react to the information by engaging in more risky behavior rather than taking a recommended protective action (e.g., refs. 19,67,68).
Although some studies have concluded that people with high risk perceptions and low efficacy tend to exhibit non-protective responses instead of protective responses27,37, others have not28,60,63,64. Following on from previous research investigating the effects of fear-arousing hazardous weather risk information10,69,70, we explore this topic in these survey data using the information perception variables in Table 1. These questions were included in the survey to measure negative reactance to information, a type of non-protective response.
As shown in Table 5, respondents’ perceptions that the information provided about the hazardous weather situation is misleading or overblown are negatively correlated with their evacuation intentions. These information perceptions are also negatively correlated with respondents’ self and response efficacy. Both results are consistent with the EPPM predictions that people with lower efficacy will engage in non-protective rather than protective responses. However, all of these correlations are weak—much weaker than most of the other correlations with evacuation intentions in Table 5.
To investigate these relationships further, we conducted regression analyses similar to those in Table 2, with respondents’ information perceptions as the dependent variable. Self and response efficacy were statistically significant predictors (p < 0.01), but the model’s adjusted R2 was only 0.04. In other words, variables other than risk perceptions and efficacy beliefs explain most of the variability in information perceptions. We also examined how respondents’ information perceptions vary as both risk perceptions and efficacy beliefs change, similar to Fig. 2c–g, and it did not appear that respondents with high risk perceptions but low efficacy tend to perceive the risk information as overblown or misleading. Further investigation revealed that many of the respondents who agreed that the information was misleading and/or overblown reported high evacuation intentions. All of these results are counter to EPPM predictions.
Overall, these results suggest that while a few respondents with low efficacy may be engaging in negative reactance rather than intending to take protective action, such behavior is not common in these data. Instead, respondents’ perceptions that the information is overblown or misleading may be functioning primarily as negative attitudes toward hurricane risk information and not necessarily as emotion-focused coping in lieu of protective responses15. However, these types of information perceptions are only a subset of possible emotion-focused coping responses discussed in the risk literature, and certain subpopulations may be more likely to engage in these types of responses15. Moreover, although Table 1 indicates that the information provided about the approaching hurricane did induce fear and worry among many recipients, only a subset of respondents received high-impact or fear-appeal messages in the experimentally manipulated message conditions12. Thus, further investigation of emotion-focused coping is needed in the context of hurricane risk communication.
Further exploration of efficacy
Across the analyses discussed above, evacuation-related response efficacy is a consistently strong predictor, which suggests that it is a significant driver of respondents’ evacuation intentions in the scenario presented. In addition, as discussed in the introduction, hurricane-related efficacy has been less extensively studied than hurricane risk perceptions. Thus, we close the analysis by further exploring what might underlie the efficacy measures used here and why response efficacy, in particular, might offer so much explanatory power for evacuation decisions in these survey data.
Looking at the response efficacy measure in Table 1, we see that it relates to this specific hazardous weather situation and the specific protective action of interest: evacuating one’s home. The measure also refers to the hurricane’s possible negative impacts that the protective action may help reduce: harm to oneself or one’s family. In addition, as described in Methods, all respondents lived in areas that had recently experienced Hurricane Sandy, and the scenario presented in the survey said that a strong hurricane was approaching, with wind speeds up to 130 miles per hour. The information that all respondents received further stated that people living in evacuation zones should evacuate, and it briefly described options for evacuating. All of these may contribute to the explanatory power of response efficacy in this study.
As another approach to understanding this measure of response efficacy, we examined what other variables measured in the survey are associated with higher or lower response efficacy. As shown in Table 5, of the situation-specific perceptions and beliefs, the three cognitive risk perceptions related to the overall threat have the strongest correlations with response efficacy. Similar to the response efficacy measure, all three of these risk perception measures relate to the personal risks that the hurricane poses to the respondent or their home—and the strongest correlation is with the likelihood of getting hurt measure. This suggests that an important component of this response efficacy measure is the wording related to reducing personal harm.
Table 6 shows that response efficacy is correlated with all of the six general hurricane-related perceptions, preparations, and experiences analyzed in this study, but not as strongly as with the situation-specific measures. Response efficacy is not meaningfully correlated with any of the four individual/household variables included as controls in the analysis. Correlations are also insignificant for other individual/household characteristics measured in the survey, including income, housing type, home ownership, head of household or employment status, household size, and presence of children in home (|r | <0.06, p > 0.10). Together, these results suggest that the response efficacy measure used here is more closely related to the personal risk that respondents perceive in this specific hurricane situation—and their beliefs about the extent to which evacuation can reduce this personal risk—than to the types of general hurricane-related perceptions and experiences measured here or generally available demographic data.
Although the response efficacy variable used in this analysis was measured in the context of the specific hurricane situation presented, it could be partly (or largely) associated with respondents’ general evacuation-related response efficacy, across different hurricane situations. Supporting this idea, of the six general hurricane-related perceptions, preparations, and experiences investigated here, evacuation for Hurricane Sandy has the strongest correlation with response efficacy (Table 6), and it is the strongest predictor of evacuation intentions (Model 7 in Table 4; see also ref. 12). This is also consistent with other research, which finds that prior hurricane evacuation is a predictor of future evacuation, or more generally, that some people tend to be “evacuators” who believe in general that evacuation is likely to reduce harm, while others are “non-evacuators” who will not evacuate in most circumstances6,8,58,71,72,73. Unfortunately, we cannot fully investigate this hypothesis using these data, because the survey did not measure respondents’ general evacuation-related response efficacy. The survey also did not measure response efficacy in the other three experimental modules that were part of the survey. However, the survey did measure respondents’ evacuation intentions in the other three experimental modules, each of which presented a different hurricane scenario (see Methods). Thus, as our best available proxy for respondents’ propensity to evacuate across multiple hurricane situations, we use their average evacuation intentions in the other three experimental modules that were part of the survey.
As shown in Table 6, this average evacuation intention variable is strongly correlated with response efficacy. When this variable is added to the regression analysis in Model 8, Model 9 in Table 4 shows that it is a statistically significant predictor of situation-specific evacuation intentions. In Model 9, the regression coefficient for response efficacy is somewhat smaller than in Model 8, but response efficacy remains a strong predictor of evacuation intentions. This suggests the response efficacy measure used here is partly associated with respondents’ general beliefs that evacuation is effective at reducing personal harm from hurricanes, and partly associated with their beliefs that evacuation is effective in the specific hurricane scenario presented. Further work is needed with additional measures of general and situation-specific efficacy, along with other measures such as prior evacuation experience, to further explicate these relationships.
As discussed earlier, our analyses find that the two efficacy measures used in this survey interact, with self-efficacy influencing evacuation intentions primarily among respondents with moderate to high levels of response efficacy. Note, however, that we cannot infer causality from these data; respondents’ self-efficacy may influence their appraisals of response efficacy, rather than vice versa. Moreover, our ability to investigate the role of self-efficacy using these data is limited by the small number of respondents reporting low self-efficacy: only 6% of the sample reported 1, 2, or 3 on the 7-point self-efficacy scale. We also did not measure evacuation costs and impediments, which are likely related to self-efficacy33,74 and are important components of both the PMT and PADM. Such barriers to evacuation are likely to be more important for actual evacuation behaviors, compared to the evacuation intentions studied here, especially for populations that are likely to experience the most harm. Thus, while this analysis is a step towards understanding the importance of response and self-efficacy for hurricane evacuation decision making, additional research is needed to understand what underlies different types of efficacy and how these influence people’s responses to approaching hazard risks.
Discussion
This article uses data from a hypothetical hurricane situation presented in a survey to examine the roles of different factors in influencing evacuation decisions. Our analysis finds that the strongest predictors of respondents’ evacuation intentions are their beliefs about the effectiveness of evacuation for reducing personal harm (response efficacy) and their perceptions that they could get hurt if they stay home during the hurricane. These types of situation-specific cognitive risk perceptions and response efficacy beliefs explain a much larger amount of the variance in evacuation intentions than respondents’ worry, fear, or perceptions of the hurricane’s wind, storm surge, or rain flooding hazards. Respondents’ beliefs about their ability to evacuate (self-efficacy) are also influential, but primarily for those with moderate to high response efficacy.
Similar to the prior research discussed in the introduction (e.g., refs. 5,6,7,8,16), we also find that variables measured outside the context of the specific hurricane situation, including individual/household characteristics, perceived hurricane-related exposure, and past experiences, can help predict evacuation intentions. However, the situation-specific risk perception and efficacy belief variables explain a larger amount of the variance in evacuation intentions. Together, these findings illustrate the value of including people’s situational perceptions of personal risk and protective action beliefs in studies of natural hazard decision making.
Aspects of these results may be influenced by the hypothetical nature of the situation posed in the survey. For example, affect is likely to be more important when people are facing a real hurricane threat, and evacuation impediments and associated self-efficacy are likely more important when people must actually evacuate. However, even in this hypothetical situation, many respondents reported feeling worry and fear. In addition, our results on the importance of respondents’ perceived likelihood of getting hurt and their beliefs about evacuation reducing harm are consistent with prior analyses that people stay at home as a hurricane approaches when they feel safe in their home, and evacuate when they do not4,8,45,50,51,71,72. Moreover, the variance explained by several of the regression models investigated here is 60% or greater, which is high for analyses with these types of data (see, e.g., refs. 36,37,38,39,47,48,53,54,54,58,72,75). This suggests that some of these results are likely to extend to real hurricane situations, although they may be attenuated.
What might account for the large explanatory power of response efficacy in this study, along with the cognitive risk perceptions related to the overall hurricane threat? One possibility is that the response efficacy measure used here asks about the risk to the respondent in this specific hazardous weather scenario, which is likely to be a strong motivator. The response efficacy measure used here also asks about a specific protective action that can reduce harm in this scenario, evacuating one’s home. These characteristics differ from the risk perception or efficacy measures used in some other studies, which are more general or ask about risk targets or actions that are less directly connected to the respondent and the situation46,52,75. In addition, since the hurricane scenario is hypothetical rather than real, the response efficacy and evacuation intention measures have a similar hypothetical phrasing; however, the cognitive risk perception measures are phrased differently, asking more directly about the threat (Table 1), and still have substantial explanatory power.
Other possible contributors to the large explanatory power of response efficacy include the nature of the hurricane scenario, risk information presented, and survey sample. The survey presented all respondents with a scenario of a strong hurricane, equivalent to a Category 4 storm, approaching their area. For many members of the coastal population sampled here, who lived in areas that experienced Hurricane Sandy several years prior to the survey, this may have been a highly salient risk, prompting concerns about harm and awareness of the need for protective action. The efficacy information presented to all respondents, which said that people in evacuation zones should evacuate and briefly described options for evacuation, may have also influenced the role of response efficacy. These two components of the information about the hurricane scenario were not experimentally manipulated, and so additional research is needed to understand the effects of the type of efficacy information included. Research with other data sets is also needed to further understand what underlies the types of response efficacy measure used in this survey, as well as the extent to which these results generalize to other situations and populations.
Our investigation of how survey respondents’ risk perceptions interacted with efficacy to influence their responses to risks found several results counter to expectations from the risk communication theories discussed above. For example, rather than a positive interaction, we found negative or non-significant interactions between risk perceptions and efficacy in influencing evacuation intentions. Moreover, our more in-depth analysis of interactions found that some respondents with low risk perceptions said they were likely to evacuate, if their response efficacy was high. This suggests that risk perceptions may not be antecedent to response efficacy, at least in this context. One possible explanation is that members of this population are reporting beliefs about the effectiveness of evacuation based on their hurricane experience, even if they perceive low risk in this situation. Or, given the dynamic nature of hurricane threats, respondents may be aware that even if they do not perceive high risk based on the current forecast, their risk may increase as the hurricane approaches and evolves43. These findings can inform applications of existing risk theories for understanding protective decision making and improving risk communication in near-term, approaching hazard situations, including hurricanes. If further research finds that these results extend to other situations, they can also help advance theory by informing models of how risk perceptions, efficacy, and other constructs influence people’s responses to risks.
One limitation of this study, noted above, is that people’s responses in hypothetical situations can differ from their responses during actual hazard threats. Research has found a correspondence between people’s intended and actual behaviors25,62,71,76,77,78, and the hypothetical situation enabled us to explore the concepts and relationships of interest here in a simplified, more controlled context. At the same time, this simplified context focuses on individual decision making, without considering the social processes in evacuation decision making (e.g., refs. 4,5,10,54,56,57,75,79) or more participatory approaches to risk management (e.g., ref. 80). In addition, people’s decision-making processes and responses can also vary across regions and populations with different characteristics and experiences, and across hazard situations. Thus, it is important to investigate the topics studied here in other populations and hazards contexts, to understand the generalizability and potential implications of these results.
Along with the areas mentioned above, our study suggests several additional topics for further research, using data sets with additional measures. One is understanding what influences and explains the type of response efficacy measure used here, to build further understanding about the underlying drivers of evacuation decision making. This includes investigating how much the type of response efficacy measure used here is determined by people’s general beliefs about the effectiveness of a protective action, and how much it is influenced by the specific situation. Another topic for further research is improving understanding about how risk perceptions and efficacy beliefs interact in hazard decision making. More specifically, what are the pathways from risk information to protective and other responses, and what are the roles of different types of pre-existing and situation-specific characteristics, perceptions and beliefs? Although these topics can be investigated using cross-sectional studies, a limitation of such work (including the study discussed here) is that one cannot explicitly test causal relationships. Thus, longitudinal studies that measure how individuals’ information use, perceptions, beliefs, and behaviors evolve over time may be especially valuable81,82. The analysis presented in this article provides valuable insight for informing these types of follow-on work.
Finally, in each of these areas, it is important to investigate these topics for different populations. For example, most of our respondents reported moderate to high self-efficacy, but capacities and constraints are key factors limiting protective behaviors for some populations. Thus, although response efficacy may be highly influential for many people, removing evacuation barriers or otherwise enabling capacity (both generally and in specific situations) can be critical for others. As another example, our results together with other research suggest that some people would typically evacuate or not across a variety of hurricane threats, while others decide based on the situation. Which factors most influence evacuation decisions may vary across these populations, leading to different strategies for effective risk communication.
Even with these potential limitations and future research needs in mind, this study suggests that persuading people at high risk that they or their families may be harmed if they stay home during a hurricane—and that evacuating can reduce the risk of personal harm—may be important levers for using risk communication to motivate hurricane evacuation. Our results further suggest that, as discussed by ref. 28, arousing fear may not be as important as effectively conveying possible negative consequences and strategies for reducing them (see also, e.g., refs. 25,34,37,40,64,83). A corresponding implication is the need for testing communication strategies that can increase response efficacy, during specific hurricane situations and more generally over time. Since believing that evacuation will reduce harm will not enable evacuation for people who do not have the ability to evacuate, or who do not believe that they can, it is also important to advance interventions that increase capacity and self-efficacy for diverse populations. Additional research on these topics can inform the design of hazard risk communications that help a variety of people at risk reduce harm when natural hazards threaten.
Methods
Survey data collection and sample
This study builds on previous research that used data from the same survey to investigate other topics12,13,14. Here we provide an overview of the survey data collection and sample; additional details are provided in refs. 12,13,14.
The survey data were collected online in 2015 from 1716 residents of coastal areas in three U.S. states (Connecticut, New York, and New Jersey) that were affected by Hurricane Sandy in 2012. Near the time of landfall, Sandy transitioned from a hurricane to a post-tropical cyclone, and so the storm is also called Post-Tropical Cyclone Sandy84 or Superstorm Sandy85. Thus, the survey referred to the storm as “Sandy”. For simplicity, however, in this article we refer to the storm as “Hurricane Sandy”.
Survey data collection and sampling was managed by GfK Custom Research, with financial incentives provided to participants. The research protocol was approved by the Institutional Review Board at Rutgers University, and written informed consent was obtained from all participants. The sample was recruited using GfK’s probability-based online panel, along with non-probability opt-in online recruitment to obtain additional respondents in the targeted geographic areas. Respondents were recruited using the ZIP code of their primary residence, with ZIP codes selected for sampling based on U.S. National Weather Service (NWS) risk assessments of hurricane storm surge flooding86. More specifically, using NWS MOM [Maximum of MEOW (Maximum Envelopes of Water)] data to represent areas with potential for storm surge flooding from a category 2 hurricane, respondents were recruited from ZIP codes with 40% or more of the landmass in those areas in New Jersey and New York, and 1% or more of the landmass in those areas in Connecticut. Although the sample was limited to people who lived in these areas at the time of the survey, 9.3% reported living in a different home during Hurricane Sandy.
The survey sample included respondents with a mix of sociodemographic characteristics such as age, gender, education, income, and race/ethnicity13. At the time of the survey, 61.7% of respondents lived in areas at risk from hurricane storm surge inundation, and 19.5% of respondents lived in an officially designated 100-year floodplain13; 30.9% lived outside both risk areas (percentages do not sum to 100% due to overlap in the risk areas). Across the sample, 12.9% of respondents said they had evacuated prior to Hurricane Sandy, and 58.6% reported preparing their residence12. When asked how much property damage and emotional distress they experienced due to Hurricane Sandy, respondents reported a median value of 2 for each, on a 1–4 scale12. In other words, even though we sampled respondents from areas that were at risk during Hurricane Sandy, respondents lived in areas with varying levels of hurricane risk, and they had a range of experiences related to Hurricane Sandy.
Survey instrument and measures
The survey began with a set of questions to screen participants, provide data for fields used later in the survey, and measure potentially relevant variables outside the context of the specific hazardous weather situations presented. This included questions about whether respondents thought they lived in an officially designated flood zone or hurricane evacuation zone, referred to as their perceived residence in a flood or evacuation zone (response options: Yes, No, Don’t Know). It also included questions about whether respondents had an evacuation plan (response options: Yes, No), whether their home had previously flooded (response options: Yes, No, Don’t Know), whether they had evacuated for Hurricane Sandy before landfall (Yes or No, coded as in ref. 12), and how much emotional distress they had experienced due to Hurricane Sandy (response options: 1 = None to 4 = A lot). Measures for these variables and their summary statistics are provided in Table 1 in ref. 12.
This set of questions was followed by a series of four separate experimental modules, each of which presented information about a different hypothetical scenario of an approaching hurricane or other coastal storm and then asked a set of questions related to that scenario. All respondents received all four modules, presented in random order. This article focuses on one of the four experimental modules that were part of the survey.
In the survey module studied here, all respondents were presented with the same introduction to the scenario: “Imagine that a hurricane is approaching [insert state]. You receive the following information from the National Weather Service: A hurricane is predicted to make landfall two days from now, with wind speeds of up to 130 miles per hour.” Each respondent then received additional information about the threat, which included an embedded experimental design that randomly assigned respondents to receive different combinations of three message conditions: Hazard, Impact, and Fear (shown in Figure 3 of ref. 12). All respondents were then presented the same information about recommended behavioral responses: “You should evacuate if you live in an evacuation zone. Options for evacuation include a hotel or the home of family or friends located outside the evacuation area, or an emergency evacuation shelter.” This information about evacuation was included based on prior work indicating that effective hazard risk communications convey information about recommended protective actions along with threat information (e.g., refs. 27,64,87).
After receiving this combination of information within the survey module, each respondent was asked a set of questions related to the hurricane situation, including those shown in Table 1. These measures of intended protective behaviors, cognitive and affective risk perceptions, efficacy beliefs, and perceptions of the information presented were developed based on the previous research discussed in the introduction. Three of the cognitive risk perception measures asked about respondents’ perceptions of the overall threat posed by the hurricane: perceived likelihood and severity at their home, and, given prior research suggesting that people take protective actions if they feel personally vulnerable or unsafe4,8,28,45,50,65,72, perceived likelihood of getting hurt if they stay home. The other three cognitive risk perception measures asked whether or not respondents thought that each of three major hurricane-related hazards—strong winds, storm surge flooding, and rain flooding—were a threat to their home from the hurricane. The two affective risk perception measures asked about respondents’ fear and worry, and the two efficacy belief measures asked about respondents’ evacuation-related self and response efficacy. Respondents were also asked whether they thought the information presented was overblown or misleading as measures of reactance, a type of non-protective response discussed in the fear appeals literature10,53,60,66.
In addition to the situation-specific variables measured in this experimental module, this article also uses data on respondents’ evacuation intentions from the other three experimental modules that were part of the survey. These data represent respondents’ evacuation intentions in the scenarios presented in each of those modules, which were measured using the same survey question shown in Table 1 in this article. We calculated the average of each respondents’ evacuation intentions across the other three modules and used this variable (mean = 4.66, SD = 1.64) in the analyses as an indication of respondents’ propensity to evacuate or not across a variety of hurricane scenarios.
Tables 5 and 6 present Pearson correlations for the variables investigated in this paper, with 2-tailed significance tests. Spearman correlations were also calculated, with similar results. As shown in Table 5, many of the situation-specific perceptions and beliefs examined here are correlated, several with moderate to strong correlations. However, given our interest in investigating the relative importance of different predictors of evacuation intention, we decided to leave each of these variables separate, rather than conducting factor analysis or forming scales related to the broader concepts. This approach is consistent with several studies that have emphasized the value of investigating the distinct roles of different components of risk perceptions and efficacy beliefs34,38,62. It also allows us to interpret our results in the context of other work that examines these variables separately, as components of or responses to risk messages.
Data analysis
We conducted the statistical analyses in this article using IBM SPSS Statistics for Windows, version 28.0.1.1. The primary analyses are multiple linear regression models with evacuation intentions as the dependent variable. The number of respondents, N, varies slightly across analyses due to missing data. Although the conceptual model in Fig. 1 and other work suggests that some of the variables investigated here may have more complex relationships (e.g., refs. 16,35,36,39,40,54,79), here we focus on testing which variables are the strongest predictors of evacuation intentions. Thus, we test direct effects, which also provides a starting point for investigating more complex relationships in future work.
All of the regression models include four individual/household characteristics as predictors: age, gender (coded as 0 = male, 1 = female), race/ethnicity (recoded into two categories: 0 = white non-Hispanic, 1 = other), and education (recoded into three categories: high school graduate or less, some college, Bachelor’s degree or higher). All of the regressions also include as predictors the three experimentally manipulated message conditions embedded within the survey module examined here. The Impact and Fear message conditions are dichotomous, and the four Hazard message conditions were recoded into a dichotomous variable (0 = wind only, 1 = any added flood message) for this analysis. As a starting point for comparing subsequent results, Model 1 (Table 2) includes only these variables as predictors of evacuation intentions.
To investigate RQ1, Models 2–6 (Table 2) compare different types of situation-specific perceptions and beliefs as predictors of evacuation intentions. Models 2–5 add the situation-specific predictor variables to Model 1 in four conceptual sets: cognitive risk perceptions related to the overall (cross-hazard) hurricane threat and specific hurricane hazards, affective risk perceptions, and efficacy beliefs. Model 6 includes all of the predictors in Models 2–5 in the same regression analysis.
For use in subsequent analyses, we developed a more parsimonious regression model, Model 6a (Table 3), by removing the five situation-specific variables that were not statistically significant predictors of evacuation intentions in Model 6; this leaves three cognitive risk perception and two efficacy variables. Comparing Table 3 with Table 2 shows that the regression coefficients and adjusted R2 in Model 6a and Model 6 are nearly identical. As part of investigating RQ2, we then tested interactions among the five situation-specific variables in Model 6a, as described in Results. All of the models with interactions had similar adjusted R2 (0.73).
To investigate RQ3, Models 7–8 (Table 4) compare respondents’ general hurricane-related perceptions, preparations, and experiences to their situation-specific perceptions and beliefs as predictors of evacuation intentions. Starting with Model 1, Model 7 adds as predictors six general hurricane-related factors measured in the first part of the survey, outside the context of the hurricane scenarios. We selected these variables based on previous findings that they are predictors of evacuation intentions, using the same survey data set analyzed here12. Model 8 adds to Model 7 the five situation-specific perceptions and beliefs that were predictors of evacuation intentions in Model 6a.
The last regression model, Model 9 (Table 4), is included as part of our additional exploration of response efficacy. It adds to Model 8 respondents’ average evacuation intentions across the other three experimental modules that were part of the survey, as a proxy for their general propensity to evacuate.
Results from the regression analyses are presented in tables using unstandardized coefficients, along with standard errors. To support results on the relative importance of different predictors of evacuation intentions, standardized regression coefficients for Models 1–9 are presented in supplementary information. Because the primary independent variables of interest are measured on the same 7-point scale and have similar standard deviations (Table 1), comparing unstandardized and standardized coefficients yields similar interpretations.
To test for collinearity in the regression analyses, we examined Variance Inflation Factors (VIFs). In Model 6 (Table 1), the largest VIFs are 3.0–3.1, for fear and worry, and most of the other VIFs are less than 2.5. In Models 8 and 9, the largest VIFs are 2.8–3.0 and most other VIFs are less than 2.5. Removing the variables with the highest VIFs produces little change in the results. This, along with the similarity in coefficients for key variables across Models 6, 6a, 8, and 9, indicates that our analysis approach is robust.
To investigate additional aspects of RQ2, Fig. 2 analyzes in greater depth how several of the situation-specific perceptions and beliefs interact to influence evacuation intentions. These results allow us to explore the more complex, non-linear relationships between risk perceptions and efficacy beliefs posited in the PMT and EPPM. Many published PMT and EPPM analyses compare low vs. high risk perceptions and efficacy by manipulating these aspects of messages, analyzing risk perception and efficacy data using median splits, or segmenting risk perception * efficacy interactions into four quadrants27,29,60,61. Here, we use an approach that enables us to examine interactions between risk perception and efficacy across a broader range of both constructs.
Data availability
The datasets used during the current study are available from the corresponding author on reasonable request.
References
Schuppe, J., Chuck, E., Chan, M., Kamb, L. & Chiwaya, N. Ian was one of the most lethal hurricanes in decades. Many of the deaths were preventable. NBC News https://www.nbcnews.com/news/us-news/hurricane-ian-florida-death-toll-rcna54069 (2022).
Fawcett, E., Smith, M., Sasani, A., Robles, F. & Weingart, E. Vulnerable and trapped: A look at those lost in Hurricane Ian. New York Times https://www.nytimes.com/2022/10/21/us/hurricane-ian-victims.html (2022).
Bucci, L., Alaka, L., Hagen, A., Delgado, S. & Beven, J. National Hurricane Center Tropical Cyclone Report: Hurricane Ian (AL092022). https://www.nhc.noaa.gov/data/tcr/AL092022_Ian.pdf (2023).
Dash, N. & Gladwin, H. Evacuation decision making and behavioral responses: Individual and household. Nat. Hazards Rev. 8, 69–77 (2007).
Huang, S.-K., Lindell, M. K. & Prater, C. S. Who leaves and who stays? A review and statistical meta-analysis of hurricane evacuation studies. Environ. Behav. 48, 991–1029 (2016).
Thompson, R. R., Garfin, D. R. & Silver, R. C. Evacuation from natural disasters: a systematic review of the literature. Risk Anal 37, 812–839 (2017).
Tanim, S. H., Wiernik, B. M., Reader, S. & Hu, Y. Predictors of hurricane evacuation decisions: a meta-analysis. J. Env. Psych. 79, 101742 (2022).
Gladwin, C. H., Gladwin, H. & Peacock, W. G. Modeling hurricane evacuation decisions with ethnographic methods. Int. J. Mass Emerg. Disasters 19, 117–143 (2001).
Phillips, B. D. & Morrow, B. H. Social science research needs: focus on vulnerable populations, forecasting, and warnings. Nat. Hazards Rev. 8, 61–68 (2007).
Morss, R. E. & Hayden, M. H. Storm surge and “certain death”: interviews with Texas coastal residents following Hurricane Ike. Wea. Clim. Soc. 2, 174–189 (2010).
Dixon, D. S., Mozumder, P., Vásquez, W. F. & Gladwin, H. Heterogeneity within and across households in hurricane evacuation response. Netw. Spat. Econ. 17, 645–680 (2017).
Morss, R. E., Cuite, C. L., Demuth, J. L., Hallman, W. K. & Shwom. R. L. Is storm surge scary? The influence of hazard, impact, and fear-based messages and individual differences on responses to hurricane risks. Int. J. Disaster Risk Reduct. 30A, 44–58 (2018).
Cuite, C. L., Shwom, R. L., Hallman, W. K., Morss, R. E. & Demuth, J. L. Improving coastal storm evacuation messages. Wea. Clim. Soc. 9, 155–170 (2017).
Cuite, C. L., Morss, R. E., Demuth, J. L. & Hallman, W. K. Hurricanes vs Nor’easters: the effects of storm type on perceived severity and protective actions. B. Am. Meteorol. Soc. 102, E1306–E1316 (2021).
Morss, R. E., Lazrus, H., Bostrom, A. & Demuth, J. L. The influence of cultural worldviews on people’s responses to hurricane risks and threat information. J. Risk Res. 23, 1620–1649 (2020).
Demuth, J. L., Morss, R. E., Lazo, J. K. & Trumbo, C. The effects of past hurricane experiences on evacuation intentions through risk perception and efficacy beliefs: a mediation analysis. Weather Clim. Soc. 8, 327–344 (2016).
Rogers, R. W. Cognitive and physiological processes in fear appeals and attitude change: a revised theory of protection motivation. in Social Psychophysiology (ed. Cacioppo, J. & Petty, R. E.) 153–176 (Guilford, 1983).
Rogers, R. W. & Prentice-Dunn, S. Protection motivation theory. in Handbook of Health Behavior Research 1: Personal and Social Determinants, 1st edition (ed. Gochman, D. S.) 113–132 (Springer, 1997).
Witte, K. Putting the fear back into fear appeal: The extended parallel process model. Commun. Monogr. 59, 329–349 (1992).
Witte, K. Fear control and danger control: a test of the extended parallel process model (EPPM). Commun. Monogr. 61, 113–134 (1994).
Lindell, M. K. & Perry, R. W. Communicating environmental risk in multiethnic communities (Sage, 2004).
Lindell, M. K. & Perry, R. W. The protective action decision model: theoretical modifications and additional evidence. Risk Anal. 32, 616–632 (2012).
Loewenstein, G. F., Weber, E. U., Hsee, C. K. & Welch, N. Risk as feelings. Psychol. Bull. 127, 267–286 (2001).
Slovic, P., Finucane, M. L., Peters, E. & MacGregor, D. G. Risk as analysis and risk as feelings: some thoughts about affect, reason, risk, and rationality. Risk Anal. 24, 311–322 (2004).
Milne, S., Sheeran, P. & Orbell, S. Prediction and intervention in health-related behavior: a meta-analytic review of protection motivation theory. J. Appl. Soc. Psychol. 30, 106–143 (2000).
Floyd, D. L., Prentice-Dunn, S. & Rogers, R. W. A meta-analysis of research on protection motivation theory. J. Appl. Soc. Psychol. 30, 407–429 (2000).
Witte, K. & Allen, M. A meta-analysis of fear appeals: implications for effective public health campaigns. Health Educ. Behav. 27, 591–615 (2000).
de Hoog, N., Stroebe, W. & de Wit, J. B. F. The impact of vulnerability to and severity of a health risk on processing and acceptance of fear-arousing communications: a meta-analysis. Rev. Gen. Psychol. 11, 258–285 (2007).
Peters, G.-J. Y., Ruiter, R. A. C. & Kok, G. Threatening communication: a critical re-analysis and a revised metaanalytic test of fear appeal theory. Health Psychol. Rev. 7, S8–S31 (2013).
van Valkengoed, A. M., Perlaviciute, G. & Steg, L. From believing in climate change to adapting to climate change: the role of risk perception and efficacy beliefs. Risk Anal. 44, 553–565 (2024).
van Valkengoed, A. M. & Steg, L. Meta-analyses of factors motivating climate change adaptation behaviour. Nat. Clim. Change 9, 158–163 (2019).
Chu, H. & Yang, J. Z. Risk or efficacy? How psychological distance influences climate change engagement. Risk Anal. 40, 758–770 (2020).
Grothmann, T. & Patt, A. Adaptive capacity and human cognition: the process of individual adaptation to climate change. Global Environ. Change 15, 199–213 (2005).
Bubeck, P., Botzen, W. J. W., Kreibich, H. & Aerts, J. C. J. H. Detailed insights into the influence of flood-coping appraisals on mitigation behaviour. Global Environ. Change 23, 1327–1338 (2013).
Zaalberg, R., Midden, C., Meijnders, A. & McCalley, T. Prevention, adaptation, and threat denial: flooding experiences in the Netherlands. Risk Anal 29, 1759–1778 (2009).
Borque, L. B. et al. 2013: An examination of the effect of perceived risk on preparedness behavior. Environ. Behav. 45, 615–649 (2013).
Grothmann, T. & Reusswig, F. People at risk of flooding: Why some residents take precautionary action while others do not. Nat. Hazards 38, 101–120 (2006).
Terpstra, T. & Lindell, M. K. Citizens’ perceptions of flood hazard adjustments: an application of the Protective Action Decision Model. Env. Behav. 45, 993–1018 (2013).
Li, Y., Greer, A. & Wu, H.-C. Modeling household earthquake hazard adjustment intentions: an extension of the protection motivation theory. Nat. Hazard Rev. 24, 04022051 (2022).
Ruiter, R. A. C., Abraham, C. & Kok, G. Scary warnings and rational precautions: a review of the psychology of fear appeals. Psychol. Health 16, 613–630 (2001).
Koebele, E. A. et al. Perceptions of efficacy are key determinants of mask-wearing behavior during the COVID-19 pandemic. Nat. Hazards Rev. 18, 06021002 (2021).
Spence, P. R., Lachlan, K. A., Lin, X. & del Greco, M. Variability in Twitter content across the stages of a natural disaster: Implications for crisis communication. Commun. Quart. 63, 171–186 (2015).
Morss, R. E. et al. Hazardous weather prediction and communication in the modern information environment. Bull. Amer. Meteor. Soc. 98, 2653–2674 (2017).
Baker, E. J. Hurricane evacuation behavior. Int. J. Mass Emerg. Disasters 9, 287–310 (1991).
Zhang, F. et al. An in-person survey investigating public perceptions of and response to Hurricane Rita forecasts along the Texas Coast. Wea. Forecast. 22, 1177–1190 (2007).
Sjoberg, L. Factors in risk perception. Risk Anal 20, 1–12 (2000).
Kellens, W., Terpstra, T. & Maeyer, P. D. Perception and communication of flood risks: a systematic review of empirical research. Risk Anal. 33, 24–49 (2013).
Wilson, R. S., Zwickle, A. & Walpole, H. Developing a broadly applicable measure of risk perception. Risk Anal. 39, 777–791 (2019).
Walpole, H. D. & Wilson, R. S. Extending a broadly applicable measure of risk perception: the case for susceptibility. J. Risk Res. 24, 135–147 (2021).
Stein, R., Buzcu-Guven, B., Dueñas-Osorio, L., Subramanian, D. & Kahle, D. How risk perceptions influence evacuations from hurricanes and compliance with government directives. Pol. Stud. J. 41, 319–342 (2013).
Meyer, R. J., Baker, J., Broad, K., Czajkowski, J. & Orlove, B. The dynamics of hurricane risk perception: Real-time evidence from the 2012 Atlantic hurricane season. Bull. Amer. Meteor. Soc. 95, 1389–1404 (2014).
Lazo, J. K., Bostrom, A., Morss, R. E., Demuth, J. L. & Lazrus, H. Factors affecting hurricane evacuation intentions. Risk Anal. 35, 1837–1857 (2015).
Morss, R. E. et al. Understanding public hurricane evacuation decisions and responses to forecast and warning messages. Wea. Forecast. 31, 395–417 (2016).
Huang, S.-K., Lindell, M. K. & Prater, C. S. Multistage model of hurricane evacuation decision: Empirical study of Hurricanes Katrina and Rita. Nat. Hazards Rev. 18, 05016008 (2017).
Senkbeil, J., Collins, J. & Reed, J. Evacuee perception of geophysical hazards for Hurricane Irma. Wea. Climate Soc. 11, 217–227 (2019).
Eisenman, D. P., Cordasco, K. M., Asch, S., Golden, J. F. & Glik, D. Disaster planning and risk communication with vulnerable communities: Lessons from Hurricane Katrina. Amer. J. Public Health 97, S109–S115 (2007).
Lazrus, H., Wilhelmi, O., Morss, R., Henderson, J. & Dietrich, A. Information as intervention: How hurricane risk communication interacted with vulnerability and capacities in Hurricane Sandy. Int. J. Mass Emerg. Disasters 38, 89–120 (2020).
McCaffrey, S., Wilson, R. & Konar, A. Should I stay or should I go now? Or should I wait and see? Influences on wildfire evacuation decisions. Risk Anal 38, 1390–1404 (2018).
Liu, B. F., Egnoto, M. & Lim, J. R. How mobile home residents understand and respond to tornado warnings. Wea. Clim. Soc. 11, 521–534 (2019).
Popova, L. The extended parallel process model: illuminating the gaps in research. Health Educ. Behav. 39, 455–473 (2012).
Rimal, R. N. Perceived risk and self-efficacy as motivators: understanding individuals’ long-term use of health information. J. Commun. 51, 633–654 (2001).
Sheeran, P., Harris, P. R. & Epton, T. Does heightening risk appraisals change people’s intentions and behavior? A meta-analysis of experimental studies. Psychol. Bull. 140, 511–543 (2014).
Babcicky, P. & Seebauer, S. Unpacking protection motivation theory: evidence for a separate protective and non-protective route in private flood mitigation behavior. J. Risk Res. 22, 1503–1521 (2019).
Tannenbaum, M. B. et al. Appealing to fear: a meta-analysis of fear appeal effectiveness and theories. Psychol. Bull. 141, 1178–1204 (2015).
Morss, R. E., Mulder, K. J., Lazo, J. K. & Demuth, J. L. How do people perceive, understand, and anticipate responding to flash flood risks and warnings? Results from a public survey in Boulder, Colorado, USA. J. Hydrol. 541, 649–664 (2016).
Witte, K., Cameron, K. A., McKeon, J. & Berkowitz, J. Predicting risk behaviors: development and validation of a diagnostic scale. J. Health Commun. 1, 317–341 (1996).
Janis, I. L. & Feshbach, S. Effects of fear-arousing communications. J. Abnorm. Psychol. 48, 78–92 (1953).
Janis, I. L. Effects of fear arousal on attitude change: Recent developments in theory and experimental research. Adv. Exp. Soc. Psychol. 3, 166–224 (1967).
Perreault, M. F., Houston, J. B. & Wilkins, L. Does scary matter?: Testing the effectiveness of new National Weather Service tornado warning messages. Commun. Stud. 65, 484–499 (2014).
Wei, H.-L., Lindell, M. K. & Prater, C. S. “Certain Death” from storm surge: a comparative study of household responses to warnings about Hurricanes Rita and Ike. Weather Clim. Soc. 6, 425–433 (2014).
Dow, K. & Cutter, S. L. Public orders and personal opinions: Household strategies for hurricane risk assessment. Global Environ. Change 2, 143–155 (2000).
Meyer, M. A., Mitchell, B., Purdum, J. C., Breen, K. & Iles, R. L. Previous hurricane evacuation decisions and future evacuation intentions among residents of southeast Louisiana. Int. J. Disaster Risk Reduct. 31, 1231–1244 (2018).
Murray-Tuite, P., Yin, W., Ukkusuri, S. V. & Gladwin, H. Changes in evacuation decisions between Hurricanes Ivan and Katrina. Transport. Res. Rec. 2312, 98–107 (2012).
Newnham, E. A. et al. Self-efficacy and barriers to disaster evacuation in Hong Kong. Int. J. Pub. Health 62, 1051–1058 (2017).
Wong-Parodi, G. & Feygina, I. Factors influencing (mal)adaptive responses to natural disasters: the case of Hurricane Matthew. Wea. Clim. Soc. 10, 747–768 (2018).
Whitehead, J. C. Environmental risk and averting behavior: predictive validity of jointly estimated revealed and stated behavior data. J. Environ. Resour. Econ. 32, 301–316 (2005).
Kang, J. E., Lindell, M. K. & Prater, C. S. Hurricane evacuation expectations and actual behavior in Hurricane Lili. J. Appl. Soc. Psychol. 37, 887–903 (2007).
Gudishala, R. & Wilmot, C. Development of a time-dependent, audiovisual, stated-choice method of data collection of hurricane evacuation behavior. J. Transp. Saf. Secur. 2, 171–183 (2010).
Horney, J. A., MacDonald, P. D. M., Van Willigen, M., Berke, P. R. & Kaufman, J. S. Individual actual or perceived property flood risk: Did it predict evacuation from hurricane Isabel in North Carolina, 2003? Risk Anal. 30, 501–511 (2010).
Rufat, S. et al. Swimming alone? Why linking flood risk perception and behavior requires more than “it’s the individual, stupid”. WIREs Water 7, e1462 (2020).
Siegrist, M. Longitudinal studies on risk research. Risk Anal. 34, 1376–1377 (2014).
Demuth, J. L. et al. Longitudinal studies of risk perceptions and behavioral responses for natural hazards. in Handbook of Risk, Crisis, and Disaster Communication (ed. Liu, B. F. & Mehta, A. M.) (Routledge, 2024).
Ruiter, R. A. C., Kessels, L. T. E., Peters, G.-J. Y. & Kok, G. Sixty years of fear appeal research: current state of the evidence. Int. J. Psychol. 49, 63–70 (2014).
National Weather Service. Service Assessment: Hurricane/Post-Tropical Cyclone Sandy, October 22–29, 2012. https://www.weather.gov/media/publications/assessments/Sandy13.pdf (2013).
Halverson, J. B. & Rabenhorst, T. Hurricane Sandy: the science and impacts of a superstorm. Weatherwise 66, 14–23 (2013).
Zachry, B. C., Booth, W. J., Rhome, J. R. & Sharon, T. M. A national view of storm surge risk and inundation. Wea. Clim. Soc. 7, 109–117 (2015).
Mileti, D. S. & Sorensen, J. H. Communication of Emergency Public Warnings: A Social Science Perspective and State-of-the-art Assessment. Oak Ridge National Laboratory Manuscript #ORNL-6609. http://emc.ed.ornl.gov/publications/PDF/CommunicationFinal.pdf (1990).
Acknowledgements
The authors thank our collaborators on the survey design and data collection, William Hallman and Rachael Shwom. This material is based upon work supported by the National Center for Atmospheric Research, which is a major facility sponsored by the National Science Foundation under Cooperative Agreement 1852977. This material is also based upon work supported by (while the lead author was serving at) the National Science Foundation. Data collection for this study was funded by NJ Sea Grant Coastal Storm Awareness Program, Grant R/CSAP-1-NJ. Any opinion, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation. The funders played no role in study design, data collection, analysis and interpretation of data, or the writing of this manuscript.
Author information
Authors and Affiliations
Contributions
R.M. contributed to data collection, performed the data analyses shown in this article, and led interpretation of results and writing the manuscript. C.C. led data collection and contributed to the design of data analyses, interpretation of results, and writing. J.D. contributed to data collection, design of data analyses, interpretation of results, and writing. All authors read and approved the final manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Morss, R.E., Cuite, C.L. & Demuth, J.L. What predicts hurricane evacuation decisions? The importance of efficacy beliefs, risk perceptions, and other factors. npj Nat. Hazards 1, 24 (2024). https://doi.org/10.1038/s44304-024-00025-8
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s44304-024-00025-8
