Anticipating sea‐level rise and human migration: A review of empirical evidence and avenues for future research
Duijndam, Botzen, Hagedoorn, Aerts · 2021
공개 학술지 Wiley Interdisciplinary Reviews. Climate Change 에 실린 총설 논문입니다(2021년). CC BY 4.0 으로 공개된 글이고, 그림 · 표 · 참고 문헌 표시를 뺀 본문을 실었습니다.
출처 Duijndam et al. (2021), Wiley Interdisciplinary Reviews. Climate Change · 라이선스 CC BY 4.0 · 원문
이 장은 원문만 있습니다. 문장별 해석은 앞 장부터 차례로 붙입니다.
2장 · Anticipating sea‐level rise and human migration: A review of empirical evidence and avenues for future research — CONCEPTUAL FRAMEWORK
Migration can be defined as the permanent or semi-permanent change of residence (Lee, 1966). Studying the impacts of environmental change on migration has become of increasing relevance due to climate change and widespread environmental degradation (McLeman & Gemenne, 2018). While initially studied predominantly by scholars from the broader natural sciences, the environmental migration literature is increasingly integrated with concepts, theories, and methodologies from the social sciences, and it is acknowledged that environmental factors are just one of many factors that influence the migration decision (Adger et al., 2018, 2021; Hunter et al., 2015; Nicholls et al., 2020; Safra de Campos et al., 2020).
Figure 1 presents a conceptual framework of migration under SLR-related hazards. This framework is modified from the original work of Black et al. (2011), and the later adaptation of Hauer et al. (2020) who applied the framework to SLR-induced migration. The main benefit of this framework is that it recognizes the complexity of the migration decision-making process and the associated drivers of migration (Hunter et al., 2015). In our adapted version, we take advantage of the richness of non-environmental factors proposed by Black et al. (2011) and the application to SLR by Hauer et al. (2020). The framework depicts five macro drivers of migration in the context of SLR: economic, political, social, demographic, and SLR-related hazards as a subset of environmental drivers. SLR-related hazards in the conceptual framework and in the remainder of this article refer to hazards that are expected to be exacerbated by climate-induced SLR, being flooding, salinization, and erosion. However, these hazards may also have other (non-climatic) causes such as subsidence, dam construction, and industrial (dredging) activities (Dunn et al., 2018, 2019; Nicholls, Lincke, et al., 2021; Vasilopoulos et al., 2021). Environmental drivers of migration refer to the hazard people face (Black et al., 2011). Hence, studying people's migration response to (perceptions of) SLR-related hazards, whether or not caused by climate-induced SLR, provides us with important lessons on how people may respond to similar hazards under increased future SLR. SLR-related hazards in the framework can influence migration directly or indirectly through the other four drivers. Within the pentagon, spatiotemporal variabilities in the migration response are shown. SLR-related hazards can drive migration gradually (e.g., salinization) or suddenly (e.g., flooding from storm surges), and migration can be in anticipation or in reaction to SLR-related hazards. Furthermore, both actual and perceived levels of risk can play a role. Whether drivers indeed translate to migration behavior is mediated by personal/household characteristics such as education, wealth, and preferences, and by intervening obstacles and facilitators such as the costs of migration, protection or retreat policies, and social networks. The framework reflects that both structural forces and human agency shape migration decisions, and it can be applied to different spatiotemporal types of migration. This conceptualization helps to classify and interpret the findings of our systematic literature review, and to identify knowledge gaps and fruitful avenues for future research.
METHODS
To synthesize current knowledge on the impact of SLR-related hazards on migration, this study conducts a systematic review of the empirical literature. Appendix S1 provides a detailed description of the article inclusion criteria and the search and screening procedure of the review. To be included in the review, the following article inclusion criteria had to be met. First, the research should analyze migration intentions or actual migration as the dependent variable. Second, the research should study the effects of SLR-related hazards, as an independent variable, on actual migration or migration intentions. Both studies looking at coastal and riverine flooding are included, as SLR through backwater effects can also increase the incidence and extent of upstream river flooding (Ikeuchi et al., 2015). Third, the study should utilize individual-level or household-level data. Fourth, data should be analyzed using multivariate regression analysis.
We conducted the systematic literature search over the period November and December 2020 to identify peer-reviewed scientific articles meeting all the inclusion criteria. A search term including keywords on SLR, migration, and empirical research methods was used and yielded 5353 articles. These articles were subsequently screened and filtered for eligibility following a three-step procedure: title analysis, abstract and keyword analysis, and finally a thorough full-text analysis. This resulted in 13 studies meeting all the inclusion criteria. We then screened the references of these 13 studies and the references of a selection of review articles dealing with migration and the environment, which yielded an additional two studies. The article selection process is summarized in Figure 2.
LESSONS FROM EMPIRICAL RESEARCH ON SLR-RELATED HAZARDS AND HUMAN MIGRATION
Characteristics of articles included in the review
Table 1 provides an overview of the 15 articles included in the review, incorporating the following characteristics: (a) author(s) and year, (b) study area, (c) SLR-related hazards studied, (d) how migration is measured, (e) type of migration studied in terms of migration destination and whether migration is temporary or permanent, and (f) survey or census data characteristics. The publication years of the articles included in this review lie between 2008 and 2020, with the majority of studies being published in the period 2017–2020. Geographically, the articles cover seven countries, being Bangladesh (five studies), the United States (four studies), Indonesia (two studies), Ghana (two studies), Australia (one study), Canada (one study), and Pakistan (one study). Most case studies comprise (at least partly) the coastal zone, except two. Regarding socio-economic context, six studies are situated in developed countries, and nine studies in developing countries. In terms of SLR-related hazards, the effect of flooding on migration is investigated in all included studies, and to a lesser extent salinization (three studies) and erosion (two studies) are also researched. For flood type, six studies focus on coastal flooding, three studies on riverine flooding, while six studies do not distinguish between flood types.
The studies either use actual migration data (nine studies) or gauge migration intentions in survey questionnaires (six studies). Six out of the fifteen studies specify the migration destination (i.e., whether migration is characterized as short-distance or long-distance, internal, or international). Nine studies do not make this specification in their regression analyses, among which are all studies that investigate migration intentions. Six studies also specify the temporal dimension of migration (i.e., whether migration is temporary or permanent). Lastly, studies differ in terms of whether they use cross-sectional data (nine) or panel data (six) for their regression analyses.
SLR-related hazards and human migration
Table 2 presents the results of the 15 studies included in the review. The results are separately presented for the impacts of flooding, salinization, and erosion. Because of the heterogeneity in research approaches, the specific measurements of the independent variables are also described in the table. For each finding, the sign of the effect is shown, where a “+” indicates a positive impact on migration, and a “–” a negative impact. In addition, the significance level of the result is shown (p < 0.10, p < 0.05, or p < 0.01). Sometimes a range of significance levels is provided in case a study presents results of multiple model estimations. Non-significant results are denoted by NS. Results are separated by migration destination, if applicable.
Flooding
All studies included in this review studied the effect of flooding on migration. Different types of flood indicators are studied: flood experience, flood severity or frequency, economic damages from flooding, geographical indicators of flood risk, and flood risk perceptions. Of the eight flood experience indicators five are reported as insignificant, two as having a positive impact on migration, and one as having a negative impact on migration. Positive impacts of flood experience on migration are found in US studies by Paxson and Rouse (2008) and Schwaller et al. (2020), for experience with home flooding specifically. Home flooding can cause serious economic damage and experiencing it first-hand can strongly increase risk perceptions of flooding, as explained by the availability heuristic (Botzen et al., 2021), which can promote migration behavior. However, for the other studies we find insignificant or even negative impacts of flood experience on migration. This discrepancy might be due to the fact that these studies were conducted in developing countries (except for Haney, 2019, who finds a positive, albeit insignificant, effect on migration intentions in a Canadian study). People in developing countries are often more dependent on coastal and riverine resources for their livelihoods, and also face higher financial constraints, which may limit their ability to migrate. These financial constraints can be further exacerbated by flood events. This is exemplified by Call et al. (2017) who find that a river flood event in the month of occurrence significantly reduces temporary migration in Bangladesh, with the odds of migration being 17% lower in a month of flooding compared with a month without flooding. Besides financial constraints, the flood type studied could also explain the divergence in results. River flooding, as opposed to coastal flooding, provides freshwater that can boost agricultural productivity and may reduce migration in agricultural households (Banerjee, 2010; Chen & Mueller, 2019). As will be explained in Section 5 in greater detail, more research evidence is needed to validate and clarify this possible multidirectional relationship between flooding and migration.
A similar discrepancy between developed and developing country case studies can be observed when we look at indicators of flood severity or frequency. Studies focusing on Bangladesh, Indonesia, and Pakistan find that more severe floods in terms of deaths, destroyed houses, or exposed households significantly decrease long-term migration (Bohra-Mishra et al., 2014; Chen & Mueller, 2019; Mueller et al., 2014). Contrastingly, studies in Australia and New York find that more severe or frequent flooding increases intentions to migrate (Boon, 2014; Buchanan et al., 2019). However, the measurement of migration could also play a role here. Namely, the studies that focused on developing countries looked at actual migration flows, while those focusing on developed countries studied migration intentions. Flood type (coastal or riverine), which in some of these studies is unspecified, could also moderate the effects found. Nonetheless, this does not undermine the finding that severe flooding tends to restrict long-term migration in developing country contexts. Although severe flooding can generate large temporary population displacements in developing countries (IDMC, 2017), displacement may often not result in permanent migration because people prefer to return to their place of residence for social or economic reasons, or because severe floods deplete household resources to finance migration (McLeman, 2014). Looking at less severe “moderate” flood events, Gray and Mueller (2012) find that this increases short-distance migration and decreases long-distance migration in Bangladesh. In contrast to severe flood events, moderate flood events might leave enough household resources to allow for short-distance migration (Kaczan & Orgill-Meyer, 2020). In light of the discussion on the possibly important role of financial constraints in developing countries, it is surprising that Bohra-Mishra et al. (2014) and Gray and Mueller (2012) find that indicators of economic losses from flooding are insignificantly or weakly positively associated with migration. However, both studies use measures of absolute economic losses. Absolute economic losses from flooding may be higher for richer households, for whom migration may still be affordable.
Risk measured in terms of the distance people live from rivers and coasts is significantly associated with migration in three out of four studies looking at this geographical indicator. Goldbach (2017) and Bernzen et al. (2019) find in, respectively, Bangladesh and Indonesian case studies that people living closer to a coast or river are more likely to migrate. In contrast, Call et al. (2017) in a Bangladesh case study find that people living closer to a river are less likely to migrate. The choice of living close to coastal or riverine waterbodies reflects a trade-off between greater economic and livelihood opportunities and increased flooding and other SLR-related hazards. However, riverine flooding also comes with benefits by irrigating the land with freshwater (Banerjee, 2010). These benefits of riverine flooding may explain the positive effect found by Call et al. (2017), who focus exclusively on the distance to rivers.
Two studies in our review investigate the relationship between migration and self-reported risk perception. Haney (2019) finds that people in the Canadian inland city of Calgary who believe that flooding will become more frequent in the future do not report higher migration intentions, while people who worry about future flooding affecting their neighborhood are significantly more likely to intend to migrate. In fact, worried individuals report a 67% lower likelihood of staying in their neighborhood in the coming year, and a 50% lower likelihood of staying for the coming 5 years, compared with non-worried individuals. Song and Peng (2017) in a US study find that both a belief that SLR is occurring and a belief that SLR will intensify extreme weather events is significantly positively associated with migration intentions. Both findings illustrate the important role that risk perceptions can play alongside actual risk, as is also pointed out in Figure 1. The influence of risk perceptions is particularly relevant for SLR, which main impacts lie in the future and can be gauged by self-reported risk perceptions.
Chen and Mueller (2019), Gray and Mueller (2012), and Mueller et al. (2014) compare the migration impacts of flooding by gender. Chen and Mueller (2019) find that in Bangladesh more severe flooding decreases migration among men, but not among women. Gray and Mueller (2012) in another Bangladesh study find that indicators of flood severity are negatively associated with migration among men, albeit insignificantly, while they tend to be positively associated with migration among women, and significantly so for moderate flood severity. Mueller et al. (2014) find that flooding in Pakistan significantly deters short-distance migration for both males and females but find no significant effects for long-distance migration. The finding of the two Bangladesh studies that flooding has more negative effects on migration for men than for women is somewhat unexpected, given that men generally have more migration decision-making power than women in developing countries, and men are also more likely than women to move for economic purposes (Kaczan & Orgill-Meyer, 2020). It could be that migration decreases more among men than among women when communities have to be rebuild after a severe flood event. In addition, migration for marriage, one of the most important reasons for migration among Bangladeshi women, might be little affected or even increase after flooding as it can be seen as a risk diversification strategy for resource-constrained households (Rosenzweig & Stark, 1989).
Whether studies focus on migration intention or on actual migration also seems to matter for the study outcomes. In studies that investigate migration intention, the effects of the flooding indicators are either insignificant or positive. Studies that investigate actual migration find that flooding can both decrease and increase migration, depending on the type of flood indicator and study context. Besides variations in study context, the intention–behavior gap could be a reason for the observed differences, as intentions do not always lead to action (Echegaray & Hansstein, 2017). Although people might intend to leave because of flooding, they may change their minds over time or find themselves physically, financially, or in any other way constrained in their migration decision. These diverging findings could also disclose that flooding impacts migration over longer time horizons than shortly after flooding events; flooding can influence migration intentions or migration decisions, but this might not yet be revealed through actual migration behavior. Reduced (financial) capability to migrate shortly after flood events can be one reason for this (Kaczan & Orgill-Meyer, 2020).
Salinization and erosion
In contrast to flooding, much less research has been done on the impact of salinization on migration, and the studies included in this review are confined to Bangladesh. However, the results that are available provide a more consistent picture of the relationship between migration and salinization than is the case for flooding. In most cases, experience with or being geographically at risk of salinization increases migration (Bernzen et al., 2019; Chen & Mueller, 2018, 2019). Although Chen and Mueller (2018) find that salinization deters international migration overall, salinization still has a significant positive influence on international migration to nearby countries in South Asia. Perhaps negative economic effects as a result of salinization can obstruct long-distance international moves. Chen and Mueller (2019) find that a one standard deviation increase in soil salinity increases migration from coastal Bangladesh to India and other countries in the region by a factor of 18, leading to an expected yearly increase in outmigration of 17,874 people to India alone (based on 2011 data). Chen and Mueller (2019) also examine whether the impact of salinization differs by gender, but find that salinization increases migration among both men and women. Salinization in Bangladesh greatly reduces yields of rice and other important crops, and as a result threatens agricultural production and people's livelihoods (Clarke et al., 2015; Hussain et al., 2018). There is now a widespread conversion from paddy cultivation to aquaculture in saline-affected areas in the country, increasing rural unemployment and poverty, and stimulating out-migration (Lázár et al., 2020; Nicholls et al., 2018). This can all explain the strong positive effect of salinization on migration found in our review. In addition, while sudden financial losses from extreme flooding events can deter people's capability to migrate, salinization is more of a gradual process where people have the time to plan migration and collect the necessary resources (Chen & Mueller, 2019; Kaczan & Orgill-Meyer, 2020). These findings on the impact of salinization in coastal Bangladesh correspond to Li et al. (2014) who in a case study in inland China find that individuals who experienced problematic salinization as a result of excessive groundwater extraction were much more inclined to migrate than individuals who did not experience these difficulties.
Only two studies in this review looked at how erosion in coastal areas influences migration. Goldbach (2017) finds that coastal erosion does not affect migration in both Ghana and Indonesia. A reason for this insignificant effect could be that coastal erosion only impacts migration for people living directly adjacent to the coast, while in the study of Goldbach (2017) most respondents lived a few kilometers away from the shoreline. In another study, Bernzen et al. (2019) focused on loss of arable land in Bangladesh and find that this does affect migration. Over 93% of respondents reported that riverbank erosion was the reason for the arable land loss. Future SLR, by intensifying the hydrological cycle and river flooding, is expected to enhance the loss of arable land as a result of riverbank erosion (Alam et al., 2016). Loss of arable land impedes agricultural and housing opportunities and, consequently, may force people to migrate (Islam & Rashid, 2011). In fact, in Bernzen et al. (2019) those who experienced arable land loss were 2.5 times more likely to migrate than those who did not. In the scientific and in the public discourse flooding as a result of SLR captures most attention (Nicholls, 2011), but the previous findings illustrate that salinization and erosion and their relationship with migration should be seriously considered as well.
Other factors that influence migration
Table A1 in the Appendix presents an overview of non-environmental factors and their effects on migration, as obtained from the results of the 15 studies included in our review. These factors cover many, but certainly not all, elements shown in the conceptual framework in Figure 1. Especially information on the impact of macrolevel factors, such as government policies, is still lacking. We only discuss factors that have been investigated in at least two study settings, so that the patterns found are more reliable.
We identify several factors that play a role in explaining migration behavior next to SLR-related hazards. First, individuals with higher income/wealth are in most cases more likely to migrate than poorer individuals. Although the poor and marginalized are often hit hardest by natural disasters, a lack of economic resources can inhibit their capability to migrate, which can be further exacerbated by natural hazard events (Black et al., 2011; Kaczan & Orgill-Meyer, 2020). However, Gray and Mueller (2012) find that coming from a richer household significantly increases migration among men but not among women in Bangladesh, reflecting potential gender inequalities in the capability to capitalize on higher household resources. Second, homeowners are less likely to migrate. Homeownership indicates a substantial economic investment that ties people to their place of residence, and may also indicate place attachment (Haney, 2019). Third, and related to the previous point, place attachment is associated with a lower intention to migrate, which is a commonly held view in the literature (Adams & Kay, 2019; Swapan & Sadeque, 2021). Fourth, two US studies find that black people are more likely to migrate than people of other ethnicities (Paxson & Rouse, 2008; Song & Peng, 2017). This could reflect that black communities in the United States are disproportionally located in areas most vulnerable to climate change (Shepard & Corbin-Mark, 2009), as well as that black people are more likely to be affected by the phenomenon of climate gentrification (Shokry et al., 2020). This raises an important issue, and policy should prevent that already existing inequalities will be aggravated by SLR. Fifth, people with social networks outside of their place of residence are more inclined to migrate. External social networks reduce information costs and make it easier for migrants to integrate into their new destination, and are central in network theories of migration (Haug, 2008; McLeman, 2014; Ryan, 2011). Finally, individual preferences including risk aversion and impatience have been found to lower the likelihood of migration. This can be explained by the fact that migration involves a high level of uncertainty, and it might take time before one can reap the benefits of a migration investment (Goldbach, 2017). We observe no clear influence of age, education, or gender. Although younger, male, and higher educated individuals are often depicted as being more likely to take part in environmental migration, especially internationally (Kaczan & Orgill-Meyer, 2020; Obokata et al., 2014), this general pattern is not observed for the studies included in this review.
TOEFL® 읽기 문제 5문항. 이 장의 한 대목을 그대로 옮기고 문항을 붙였습니다. 대목을 읽고 문제를 풉니다.
이 글은 위 출처의 원문을 그 라이선스(CC BY 4.0)에 따라 옮겼습니다. 참고문헌 목록과 인용 번호, 그림 · 표 · 각주 · 수식 · 부록은 빼고 절별로 나누어 실었으며, 그 밖의 문장은 원문 그대로입니다. 원 저작자와 발행 기관은 이 화면을 만들거나 검토하지 않았고, 이곳의 해석과 낱말 풀이는 공식 번역이 아닙니다.
