Move Out or Dig in? Risk Awareness and Mobility Plans in Disaster‐Affected Communities

Post&#8208;disaster migration patterns have been thoroughly studied from a demographic standpoint, but affected community residents&#8217; perceptions of ongoing risks and their willingness to remain in an affected community remain under&#8208;researched. Using data generated by 407 surveys and 40 interviews with residents impacted by the 2013 Calgary flood, this study analyses the effects of flood experience on residents&#8217; worry about future floods and their ensuing short&#8208;term and medium&#8208;term mobility plans. The results indicate that home flooding and evacuation orders are both predictive of worry about future floods. In turn, worry about future floods as well as age, homeownership, and place attachment are all predictive of post&#8208;disaster mobility plans. Residents discuss how the flood either strengthened or weakened their place attachment. The paper concludes by discussing the implications for social science research and for public policy that aims to mitigate disaster risk.


| INTRODUC TI ON
Although these questions have been addressed by demographic research at the community and neighbourhood level, they have rarely been examined at the individual and household levels.
Especially uncommon is a focus on wealthy urban areas, where residents' migration decisions are expected to be based not only upon immediate subsistence needs, but also on non-economic and non-material factors.
The present study addresses this gap through the use of innovative mixed-method data, including both a representative survey and qualitative interviews, allowing for both context and depth, collected after the costliest disaster in Canadian history, the 2013 Southern Alberta flood. This mixed-method design provides an understanding of which disaster-affected residents intend to remain and which intend to move, as well as why and how they make their respective decisions. The analysis reveals not only the covariates and predictors of risk perception and mobility plans, but also the indicators of how residents' worry over future flooding transforms into mobility plans. DOI: 10.1111DOI: 10. /1468 location decisions among relatively affluent and privileged populations" (387).
The existing work on decision-making in the context of a disaster tends to examine how households make risk-management (Peacock, Brody, & Highfield, 2005;Taylor, Priest, Sisco, Banning, & Campbell, 2009) and evacuation decisions (Bateman & Edwards, 2002;Bowser & Cutter, 2015). In particular, we know much about risk awareness, worry, and how those factors translate into preparedness. First, we know that a host of demographic variables are predictive of flood risk awareness including age (Lazo, Bostrom, Morss, Demuth, & Lazrus, 2015), income or social class (Burningham, Fielding, & Thrush, 2008), and education (Karanci, Aksit, & Dirik, 2005), as are actual objective levels of flood risk (Kellens, Zaalberg, Neutens, Vanneuville, & De Maeyer, 2011;O'Neill, Brereton, Shahumyan, & Clinch, 2016;Wallace, Poole, & Horney, 2016). Second, we know that experience in prior events is a robust and important predictor of risk awareness (Burningham et al., 2008) and of protective action, such as evacuation (Paul, Stimers, & Caldas, 2015). On the one hand, prior experience provides knowledge which should be positively related to awareness of ongoing risk (Scolobig, De Marchi, & Borga, 2012). But on the other hand, research indicates that this awareness often does not translate into worry or preparedness (Hopkins & Warburton, 2015;Knocke & Kolivras, 2007), owing in large part to the role of emotions as an influencer of understandings and responses to risk (Bubeck, Botzen, & Aerts, 2012), although a few studies have shown that past experience does lead to greater concern (Thistlethwaite, Henstra, Brown, & Scott, 2017). When it occurs, the failure to translate experience and awareness into concern happens because individuals misunderstand the possible consequences of a flood event and cannot properly envision how it will unfold (Siegrist & Gutscher, 2008). Third, we also know that social capital and place-based social networks are predictive of risk awareness (Aldrich & Meyer, 2015), preparedness (Levac, Toal-Sullivan, & O'Sullivan, 2012;Poussin, Botzen, & Aerts, 2014), and of the willingness and ability to evacuate in advance of a flood event (Buckland & Rahman, 1999;Litt, 2008). With all this said, research in Canada indicates that residents in flood-prone areas have surprisingly low overall levels of risk awareness (Thistlethwaite et al., 2017;Thistlewaite, Henstra, Peddle, & Scott, 2017). But how does this lack of awareness prior to disasters translate into risk awareness and mobility decisions following a flood?
Existing work provides useful insight into how households balance risk-related decisions with their financial resources (Elder et al., 2007), considerations about protecting their property investments (Riad, Norris, & Ruback, 1999), and concerns about family togetherness (Haney, Elliott, & Fussell, 2007;Litt, 2008), but again, it is mostly based on pre-disaster or pre-evacuation surveys of or interviews with at-risk residents. Research on how the already-affected residents make longer-term mobility decisions is missing, although this approach would echo Hunter et al.'s (2015) call for more understanding of the temporal complexities of the environmental migration process (391).
Many analyses of post-disaster migration take an explicitly demographic perspective, analysing how patterns of destruction intersect with local demographics to trigger various types of in-andout-migration. For instance, Elliott and Pais (2010) demonstrate that longer-term recovery displaces socially disadvantaged residents (see also Schultz & Elliott, 2013) and racial and ethnic minorities from the most impacted areas (Elliott, 2015). It has also been established that post-disaster out-migration is impacted by several macro-level variables, including the degree of house damage, housing density, and larger populations of disadvantaged residents (Collenteur, de Moel, Jongman, & Di Baldassarre, 2015;Myers, Slack, & Singelmann, 2008). In terms of new residents' in-migration, Fussell, Sastry, and Vanlandingham (2010) demonstrate that racial and ethnic inequalities influence these patterns, as in the case of post-Katrina New Orleans.
Largely missing from the existing demographic work, however, are household-level analyses of individual decision-making in the aftermath of a disaster. Some research provides a theoretical framework for understanding how vulnerable people make migration decisions in the face of environmental change, examining such factors as the state of the environment and individuals' coping capacities (Reynaud et al., 2013). For instance, one study finds that, despite the inherently emotional nature of disasters, residents employ a rational-actor, cost-benefit framework when deciding whether to return to an affected region. For many of these affected residents, disaster transforms a physical structure from a "home" into a "house," decreasing emotional investment in the dwelling (Henry, 2013). Along those lines, Koslov (2016) demonstrates how Staten Island residents affected by Hurricane Sandy made a case for official buyout programmes, an approach called "managed retreat." Asad (2015), on the other hand, finds that economic concerns alone cannot explain postdisaster return decisions as many displaced residents return even if that entails paying an economic price. In other words, place attachment may trump-or at least compete with-purely economic concerns for many residents making post-disaster mobility decisions. In broad strokes, experiencing one disaster has a tangible, long-term impact on individual assessment of future risk and disasters. For instance, Siegel, Shoaf, Afifi, and Bourque (2003) find that, irrespective of the degree of property damage, those who have experienced emotional injury from a prior disaster feel more impacted by a second disaster than otherwise similar people without such prior experience.
When we move beyond rational-actor models, research demonstrates that place attachment can explain how people understand and engage with climate change (Scannell & Gifford, 2013). In the post-disaster context, Morrice (2013) highlights a "powerful emotional quality associated with how people relate to place, recognizing that return decisions are emotionally driven and not necessarily based on material constraints" (33). This is particularly true for homeowners who report a larger social and emotional place connection than renters (Windsong, 2010). Additionally, such factors as political trust (Reinhardt, 2015), emotion and nostalgia (Morrice, 2013), and connection to place (Landry, Bin, Hindsley, Whitehead, & Wilson, 2007) also play a role in return decision-making by encouraging return migration. Still, just as most work in the field, these studies rely upon the analyses of displaced evacuees' return decisions rather than on returned residents' longer-term mobility decisions. Thus, while disaster-affected residents often experience reinvigorated place attachment, even when economic and political pressures prevent them from returning and rebuilding (Chamlee-Wright & Storr, 2009), very few studies look at residents who have already returned and are making subsequent decisions about longer-term mobility.
One notable exception is Adams' (2016) work which demonstrates that non-migration after a disaster is related to "high levels of satisfaction, resource, barriers and low mobility potential" (429).
An emergent area of inquiry involves decision-making in the wake of government-sponsored buyout programmes. One compelling finding is that community-level resilience and social capital appear to temper residents' willingness to accept buyout offers and relocate (Binder, Baker, & Barile, 2015). However, like most extant work, this research draws upon interviews with evacuees, assuming that evacuated residents choose to either return or relocate, and ignoring the residents who return only to begin making plans for longer-term relocation to less vulnerable places. As a result, little is known about how different disaster experiences, for example, home flooding or prolonged evacuation, affect urban residents' future orientations, which are expected to impact residents' decision to stay or to relocate following a catastrophic event. One exception to this knowledge gap is the study of the effects of Hurricane Katrina by Myers et al. (2008), who find that the degree of house damage is predictive of eventual relocation.
Researchers and policymakers are divided as to whether officials should encourage displaced residents' return migration. The debate hinges on whether displacement is a long-term, secondary disaster, or a propitious opportunity to move residents from physically vulnerable places (Fernando, Warner, & Birkmann, 2010; see also Graif, 2015). In the Alberta context, while the Institute for Catastrophic Loss Reduction argues that the government should encourage more out-migration from vulnerable places (Kovacs & Sandink, 2013), such a move would be hindered by Calgary's substantial oil wealth, since wealthier locales, which can often afford to fund their own recoveries, typically see less out-migration after a disaster (Cross, 2014).
If we hope to advance ongoing debates about whether long-term mobility mitigates risks or destroys an established community, it is important to determine what factors influence residents' plans to leave. The understanding of longer-term mobility plans is particularly important as disaster-affected residents try to re-establish a sense of normalcy, often called "ontological security" (Hawkins & Maurer, 2011) or a "new normal" (Gotham, Blum, & Campanella, 2014;Tierney, 2013), a state that Turner (1976) calls "a stage of culture readjustment… in which prolonged analyses are not undertaken, but only the minimal recognition of changed circumstances necessary to deal with the immediately pressing problems" inherent in the post-disaster milieu (763; see also Dynes, 1993). Research on Florida's Hurricane Andrew indicates that residential relocation is associated with higher levels of medium-term stress, isolation and social disruption (Riad & Norris, 1996), and increased economic hardship (Hori & Schafer, 2010). The importance of these relocation decisions is intensified in a neoliberal state, where the responsibility for recovery is "placed on the shoulders of neighbourhoods and citizens, without providing them the means to achieve their goals" (Kroll-Smith, Baxter, & Jenkins, 2015, 101).
If municipalities hope to retain the affected residents and attract new in-migrants, we should first examine how households and families make migration decisions after a disaster. This knowledge can help social scientists to better understand how disaster-affected residents translate risk awareness into concrete mobility plans and to create a more empirically informed theory of social action and practice in the post-disaster setting. Given the literature in the field, this article hypothesizes that: 1. Having one's home flood and evacuating will be positively associated with worry about future flooding; 2. More worry about future flooding will be associated with less intention to remain in residents' pre-flood neighbourhoods; 3. Place attachment will figure at least as prominently as flood impact and economic circumstances in residents' post-disaster mobility plans; 4. Residents planning to remain will experience an enhanced attachment to their neighbourhoods, rationalizing their decisions to stay | 227 HANEY by either the low likelihood of a future flood or the structural mitigation activities believed to make their neighbourhoods safer.

| DATA AND ME THODS
The data used in the following analyses are derived from a survey of 407 Calgary residents living in the city's 26 flooded and/ or evacuated neighbourhoods, as well as qualitative interviews with 40 flood-affected residents. In May 2014, our research team randomly selected 1,500 households from these 26 neighbourhoods for the survey, proportionally to each neighbourhood's population. The households were selected by first numbering each block within each of the 26 neighbourhoods, then numbering each house or apartment, including all four sides of a block, and finally using a random number generator to select the required number of households on each block. For example, if 50 households had to be selected from a 25-block neighbourhood to maintain representativeness, two households per block were selected. Thus, each household in the neighbourhood had an equal chance of inclusion in the final sample, and each neighbourhood was proportionally represented.
After the sampling was complete, a survey, along with an information sheet and an envelope with return postage, was mailed to each selected household. The participants were also given a form allowing them to claim a $25 gift card to RONA, a Canadian-owned home improvement chain, as a token of appreciation for their participation. The survey contained more than 100 items, including questions about evacuation, the use of social networks during the disaster, and future plans for returning and rebuilding, as well as the demographic information about household members. Ninety-six envelopes came back marked "return to sender," which is very common in disaster-affected areas where residents are no longer living in the house, the house is slated for demolition, or current construction is underway (Haney & Elliott, 2013). Therefore, I assume that 1,404 surveys were received by the residents, though this is surely a highend estimate. It is likely that many envelopes reached vacant homes or apartments, but were never officially returned to sender. As a matter of fact, a few surveys were returned as undeliverable even 2 years after the initial mailing.
In June 2014, the research team began visiting the sampled households on-foot, asking residents to participate. Many residents of the sampled households completed the survey upon these visits, substantially increasing the response rate. On-the-ground data col- The logistic models below regress several dependent variables related to post-disaster flood worry and future mobility plans on a host of independent variables. The models use demographic variables including age (and age-squared), race/ethnicity (1 = White; 0 = Not white), female (1 = Yes; 0 = No), marital status (1 = Married; 0 = Not married), parenthood (1 = Parent; 0 = Not a parent), and whether the participant has a Bachelor's degree or higher (1 = Yes; 0 = No). The models also include two dummy variables for income, with an excluded reference category. The original survey used 20 ordinal categories for household income, with each category representing a $10,000 increment: $0-9,999 coded as 0, $10,000-19,999 coded as 1, and above $200,000, which is the highest category, top-coded as 20. Since only 86% of participants provided a valid income, the income variable was imputed by the multiple imputation command available in Stata ("impute"). 1 This variable was then recoded into dummy variables representing income tertiles, including $0-89,999 (the lowest income tertile), $90,000-129,999 (middle-income) and above $130,000 (the highest income tertile). The "high" and "low" variables are included in the regression models, and "middle" is excluded as the reference category. To measure pre-disaster connectedness to community, the models include variables for  Table 3, which models mobility plans, also utilizes independent variables measuring the respondents' worry about future floods affecting the Calgary and their neighbourhood (1 = Yes; 0 = No). Table 1 includes the descriptive statistics and metrics for all the variables used in the models.
Using data from the City of Calgary's Community Profiles (City of Calgary, 2015), which are derived from 2011 Canadian census data, I calculated descriptive statistics for the population of residents living in these neighbourhoods. I will briefly compare them to the sample data provided in Table 1. As an example of the sample's representativeness, 66.43% of residents in the neighbourhoods used for the study have a bachelor's degree or higher, while 63% of the sample holds a bachelor's degree or higher, meaning that the sample is highly representative of the neighbourhood population in terms of educational attainment. The mean age of participants in the sample is 48 years old, compared to 39 years old in the river communities as a whole. While that appears like an oversampling of older residents, it is important to remember that the census data contain children, whereas the survey excluded children. Thus, the 8 year difference between the sample and population means likely reflects the decision to exclude children. Higher income residents are oversampled; the median household income in the 26 affected Electronic copy available at: https://ssrn.com/abstract=3605643 communities was $83,645 at the 2011 census, yet the median income in the sample was between $100,000 and $109,999. The survey also contains a disproportionate number of women (65% of the sample compared to 50% of the neighbourhood population).
In short, there is an overrepresentation of both women and higher earning residents, a common finding in research done in post-disaster communities. Haney and Elliott (2013) suggest this occurs because of the comparatively easier time wealthier residents have returning after a disaster and, therefore, receiving the invitation to participate.
To augment the regression results, the article utilizes qualitative data gleaned from 40 in-depth interviews with the affected residents, whose names have been changed to ensure confidentiality. I recruited participants through the community associations in the 26 flood-affected neighbourhoods. All community associations were sent an email describing the interview, and giving the criteria for inclusion (must have been an adult neighbourhood res-

| Risk perception and worry
Presumably, experiencing a disaster encourages residents to think more critically about environmental and climate-related issues, particularly whether climate change will translate into more events   (Boudet et al., 2014;Marshall, 2004; see also Nagel, 2015).
The most notable finding, however, is the effect of flooding on worries over the future: the residents whose homes flooded in 2013 are over five times more likely to worry about their neighbourhoods' TA B L E 2 Logistic regression models of future flood concern future flooding than the residents who did not flood (odds ratio = 5.594). This figure suggests that direct disaster experience and material losses trigger an increased feeling of vulnerability to future disasters (see Burningham et al., 2008). Similarly, those ordered to evacuate are some 68% more likely to worry about their neighbourhoods' future flooding than those who did not evacuate (odds ratio = 1.680). This result is likely due to both the evacuees' residence in the most vulnerable neighbourhood areas and their evacuation experience making a future flood seem more plausible or imminent.
Income also matters, as both those who are higher income and those who are lower income are less likely than the middle-income group to worry about future flooding affecting their neighbourhood.
Those who rated their pre-disaster neighbourhoods as "excellent" places to live are 94% more likely to fear future flood eventsalmost twice as likely as those who rated their neighbourhoods as less than excellent (odds ratio = suggests not only a lack of worry, but perhaps also a sense of fatalism. The most common reason for the lack of worry was the doubt that another disaster like the 2013 flood would occur again, a finding that echoes the existing research on residents' misinformation about actual flood risk (Lata & Nunn, 2012) and their misunderstanding of the "1-in-100-year" calculation of flood frequency (Ludy & Kondolf, 2012). Even though this term indicates a 1% chance of flooding each year, many residents nevertheless believe that it indicates a flood happens only every 100 years. As Ludy and Kondolf (2012) find, even most high-income, well educated people living in flood-prone areas are unfamiliar with this form of risk analysis and communication prior to a flood event. This leads to flawed analyses of risks during and after an event, as well. Bell and Tobin (2007) find that the 1% language of risk is more effective than the 1-in-100 year parlance for communicating risk, although surprisingly ineffective for motivating protective or mitigative action. Thus, Irene claimed, I'm not too worried anymore. I don't think it's going to ever happen again here. And I might be blind but I don't think it will. I think we learned enough. The province has learned enough and I think the city's learned enough, that it will never happen again. And  HANEY receded into history. Despite self-reminders about the infrequency of catastrophic flooding, explicit attempts to ignore flood risk, and the acknowledgement of seasonal variation in worry, most of the interviewees reported being "always worried" or "very concerned."

| Mobility plans
While the above models capture risk perception and worry, it is also prudent to ask how flood-affected residents plan to adapt or respond to perceived flood risk. In particular, the factors that predict who plans to remain in their neighbourhoods and who plans to move elsewhere are important. Table 3  The descriptive statistics indicate that 81.8%, or the vast majority of participants, intended to remain in their neighbourhoods in the next year, though this number drops to 56% for the next 5 years.
Even among those who flooded in 2013, 79% intended to remain in their neighbourhoods in the ensuing year, while 52% intended to remain for at least 5 years. Table 3 models resident plans to remain in their pre-flood neighbourhoods 1 and 5 years after the survey (respectively, 2 and 6 years after the flood).
The results in Table 3 indicate that age is a significant predictor of mobility plans. The age-squared predictor tests for a curvilinear effect and, indeed, both age and age-squared are significant. The positive effect of age indicates that as people age, their odds of intending to remain in their neighbourhoods increase, but the negative effect of age-squared indicates that the magnitude of this effect decreases at higher ages. Homeownership decreases mobility plans, as homeowners are 2.5 times more likely than renters to intend to remain in their neighbourhood 1 year after the survey (odds ratio = 2.485), and 2.4 times more likely to be there 5 years later (odds ratio = 2.389).
This finding probably relates to the affected properties' decrease in value (Bin & Polasky, 2004), which renders them harder to sell, as well as to homeowners' increased emotional and social investments in the neighbourhoods where they have purchased homes (Windsong, 2010). The perception of one's neighbourhood being an excellent place to live increases the odds of planning to stay in one's neighbourhood, both in the short and medium term (odds ratios of 2.764 and 2.473). However, the worry about future flood affecting one's neighbourhood significantly decreases these odds; those who harbour such worries are 67% less likely to see themselves in their neighbourhood after 1 year (odds ratio = 0.328) and 50% less likely to see themselves in their neighbourhood after 5 years (odds ratio = 0.503). These results indicate that risk perception and worry are significant and robust predictors of post-disaster mobility plans; worrying about future flooding, even when controlling for place attachment and other relevant predictors, is highly indicative of intentions to leave a flood-affected area. In contrast to work indicating the importance of variables such as gender, parenthood, and social class or income (Haney et al., 2007) for taking non-structural mitigative actions (such as evacuating or moving), there is no evidence that these demographic variables matter for mobility intentions in this sample of Calgarians living in flood-affected areas, a finding similar to Binder et al. (2015).
The qualitative findings reveal that while flood experience prompted some residents to ponder avenues for mobility, it entrenched other residents into their communities and enhanced their desire to reinvest themselves in their neighbourhoods. Out-migration desires were often fuelled by economic preservation. As one survey participant told us, he wanted to sell while his house still had some value. Decisions to dig in and persist, however, were often explained by increased place attachment and connection to one's neighbours.
As expected, many flooded residents said that they wished to move to less risky locations and never again live so close to a river.
As Scott stated, We would never have bought our house had we known that. Never. And we will never again buy next to a river. I think it is a mistake. The river needs room and the developers should not … the city should not allow it first off. It is the government's fault. I believe it is the city government; city government allowed that development to occur and the developers did it, so I am very, very unhappy about that.
Eager to blame developers and the city government for allow- Although some affected residents wanted to remain in their neighbourhood, they perceived older age as a hindrance to this decision. According to Christian, "You get tougher until you turn… Fifty-Five-ish and then most people begin to get less tough. And so that, that's happened to me…So we're moving out…I would say the people who cannot handle stress should not live…in more, in risky communities." For others, the material impact of the flood provoked a sense of dread over future floods and, therefore, a desire to move.

| CON CLUS ION
The findings of the study provide support for all four of the hypotheses provided earlier in the article. They indicate that gender and direct material flood experience, specifically home flooding and being ordered to evacuate, are significant predictors of worry over future flooding (consistent with Hypothesis #1). This worry, in turn, is a significant predictor of 1-and 5-year mobility plans (Hypothesis #2).
Plans to leave one's neighbourhood are somewhat buffered by place attachment and homeownership, however (Hypothesis #3). Similar to the study by Binder et al. (2015), this study shows that the demographic predictors such as gender and parenthood are unrelated to post-disaster mobility plans.
The qualitative data from the study indicate that the residents who worry about future flooding also plan residential relocations into less hazardous locations. By contrast, a non-trivial segment of the interviewees believe that future floods are unlikely and, should they occur, either the floods will not be as extensive and destructive as the 2013 flood or the participants will be better equipped to handle them. For those residents who do not worry about future flooding, residential relocation would logically seem unnecessary.
Additionally, for many, their new-found or reinvigorated place connection prompts them to stay, regardless of potential future flood risk (Hypothesis #4).
These findings resonate with the call by Hunter et al. (2015) for a more temporally nuanced understanding of environmentally induced migration decisions, as they illuminate how residents returning after an environmental disaster make longer-term mobility plans. Mobility decisions cannot be explained simply by cost-benefit, rational-actor models of behaviour (i.e., Henry, 2013;Riad et al., 1999). Indeed, several common variables related to economic concerns, such as income and home flooding, are not statistically related to 1-or 5-year mobility plans. Instead, place attachment, expressed through perceptions of one's neighbourhood as an excellent place to live, the strength of local social networks, and local civic engagement, are significant predictors. The models also reveal that those who believe flooding will become more common in the future exhibit higher levels of worry about future floods. However, only worry about one's neighbourhood flood risk-not municipal flood risk-is associated with mobility plans. These results emphasize the centrality of neighbourhood place attachment and neighbourly networks not only for bringing disaster-affected residents home, but also for encouraging them to make longer-term commitments to their flood-affected neighbourhoods. While research has focused on climate change and migration in the Global South and in subsistence farming communities (Nawyn, 2016), the present article demonstrates how residents in wealthier, urban centres of the Global North may respond to environmental hazards and change. Whereas many people around the world are prompted to migrate by considerations of income or subsistence, the residents of wealthier places like Calgary are primarily motivated by non-economic, cultural, and network-related factors.
Just as all studies, the present study carries some limitations and caveats. First of all, there is the issue of intention vs. action: people do not always move when they say they will, and many move unexpectedly, without prior planning (Lu, 1989). Therefore, the present study can only gauge residents' intended mobility plans, not their actual mobility decisions. Secondly, the article utilizes a modest sample of survey data (n = 407), and I would welcome a much larger study on residents' post-disaster mobility plans. It is highly probable that many of the non-significant coefficients in this study would be significant in a larger sample. The article only interprets the significant findings, despite recent cautions from the American Statistical Association about the limits of significance (Wasserstein & Lazar, 2016). Thirdly, although the survey participants are representative of the flood-affected neighbourhoods, the interview participants may not be. They were recruited through community associations, so it is unlikely that civically disengaged residents would have received the recruitment information. Given this selection method, it is likely that the interviews oversampled for residents who are civically engaged. But, since the interviews did not focus specifically on civic engagement (but rather, on mobility plans and worry, as well as environmental concerns and practices-which will be the focus of future papers), resultant bias is somewhat minimized. Fourth, it is important to note that the flood-affected Calgary neighbourhoods have been largely untouched by home buyout programmes, which distinguishes this case from the case of Hurricane Sandy (Koslov, 2016) and other disasters. Although the Alberta government has bought a handful of the most vulnerable properties, the vast majority of flood-affected residents have not been offered a chance to participate in a buyout programme and must, therefore, make relocation decisions that involve selling their homes at post-flood market value.
These findings are particularly important in the Albertan and Canadian contexts. Just as in many other places, floods are posited to become much more common, with the Institute for Catastrophic Loss Reduction and the Insurance Bureau of Canada predicting that Alberta will see a 10% increase in severe weather events as temperatures rise by as much as 4°C by 2050 (Sandford & Freek, 2014), twice the global rate of increase. The present study is an important step in helping social scientists determine how and why larger demographic trends such as out-migration occur in a post-disaster context. Although the findings are drawn from a sample of floodaffected residents, it should also help us understand residential mobility decisions in many types of disasters, albeit probably only in wealthier cities of the Global North where local social networks and place attachment can afford to trump purely economic considerations. With that said, the findings may not be as useful for understanding mobility decisions when hazards linger (such as chronic contamination) long after the acute disaster stage and residents understandably worry about the human health effects of remaining (Edelstein, 2004). The findings can also help researchers understand how the combination of individual and neighbourhood factors play a role in facilitating or preventing post-disaster migration. In particular, this should be of interest to researchers developing mechanisms that could encourage residents to return, rebuild, and reinvest themselves in their pre-disaster communities, thereby strengthening place-based social capital. The study can also assist applied practitioners and government officials involved in buyout programmes in better understanding residents' needs and concerns after a catastrophic event. Sinclair, and Daran Gray-Scholz) for their hard work on data collection and entry. Thanks are owed to Pam MacQuarrie for assistance on GIS and mapping. Gratitude also goes to the flood-affected residents who took the time to share their experiences and views with us.

E N D N OTE S
1 Sensitivity analyses, available upon request, indicate that these missing cases are mostly missing at random and whether a participant provided a valid income is not statistically related to any of the demographic variables (except marital status), nor is it predictive of any of the dependent variables in the models.
2 In Calgary, the locus of this activity is the neighbourhood. In fact, each of Calgary's approximately 150 neighbourhoods maintains its own community association, many of which have their own community centre, hosts events and workshops, has its own hockey rink, children's soccer league, and so on. These associations require volunteers as board members, event organizers, coaches, maintenance people, and so on. The City of Calgary maintains an office to coordinate and oversee the activities of these community associations. They are also joined by the non-profit Federation of Calgary Communities. Given this structure, community (neighbourhood) associations are an important facet of life in Calgary that involve many residents. Therefore, to most Calgarians, being active in the neighbourhood equates to being active in the association and its activities.