Studying Displacement After a Disaster Using Large-Scale Survey Methods: Sumatra After the 2004 Tsunami

Abstract

Understanding of human vulnerability to environmental change has advanced in recent years, but measuring vulnerability and interpreting mobility across many sites differentially affected by change remains a significant challenge. Drawing on longitudinal data collected on the same respondents who were living in coastal areas of Indonesia before the 2004 Indian Ocean tsunami and were reinterviewed after the tsunami, this article illustrates how the combination of population-based survey methods, satellite imagery and multivariate statistical analyses has the potential to provide new insights into vulnerability, mobility, and impacts of major disasters on population well-being. The data are used to map and analyze vulnerability to post-tsunami displacement across the provinces of Aceh and North Sumatra and to compare patterns of migration after the tsunami between damaged areas and areas not directly affected by the tsunami. The comparison reveals that migration after a disaster is less selective overall than migration in other contexts. Gender and age, for example, are strong predictors of moving from undamaged areas but are not related to displacement in areas experiencing damage. In our analyses, traditional predictors of vulnerability do not always operate in expected directions. Low levels of socioeconomic status and education were not predictive of moving after the tsunami, although for those who did move, they were predictive of displacement to a camp rather than a private home. This survey-based approach, although not without difficulties, is broadly applicable to many topics in human–environment research and potentially opens the door to rigorous testing of new hypotheses in this literature.

理解人类之于环境变迁的脆弱性, 在近年内不断推进, 但测量脆弱性, 以及对诸多受到变迁不等影响的场域之流动进行诠释, 却仍然是个重大挑战。本文运用 2004 年印度洋海啸发生前居住于印尼沿海地区、并于海啸之后重新访谈的同一批受访者的纵向数据, 阐明结合根据人口的调查方法、卫星影像与多变量统计分析, 如何有潜力对脆弱性、流动, 以及主要灾害对于人口福祉的冲击提出崭新的洞见。研究数据用来製图并分析亚齐与北苏门答腊各省在海啸之后迁移的脆弱性, 以及比较海啸过后受创地区与未受直接影响的地区的迁徙模式。此般比较, 显示出灾害后的迁徙, 总体而言较其它脉络中的迁徙更不具选择性。例如, 性别与年龄是预测未受创地区中迁徙的有力指标, 但却无关乎受创地区的迁移。在我们的分析中, 传统的脆弱性预测指标, 并不永远依照预期的方向进行。低度的社经地位与教育, 并无法预测海啸过后的迁徙, 即便这些指标对真正迁徙的人而言, 可以预测低度社经地位与教育者将迁徙至难民营, 而非迁徙至私人家户。此一以调查为基础的方法, 儘管仍面临困难, 但仍可广泛地应用至诸多人类—环境的研究主题中, 且有潜力开放给本类文献对新假说的严格测试。

La comprensión de la vulnerabilidad humana frente al cambio ambiental ha progresado en años recientes, aunque medir la vulnerabilidad e interpretar la movilidad a través de muchos sitios afectados de manera diferente por el cambio siguen siendo retos significativos. Con base en datos longitudinales obtenidos de los mismos encuestados que residían en áreas litorales de Indonesia antes del tsunami del 2004 en el Océano Índico, y que volvieron a ser entrevistados después del desastre, este artículo ilustra el potencial que tiene la combinación de métodos de estudio basados en población, imaginería satelital y análisis estadísticos multivariados para generar nuevas apreciaciones valiosas sobre vulnerabilidad, movilidad e impactos de grandes desastres sobre el bienestar de la población. Los datos se utilizaron para mapear y analizar la vulnerabilidad sobre los desplazamientos posteriores al tsunami, a través de las provincias de Aceh y Sumatra del Norte, y para comparar los patrones de migración postsunami entre las áreas dañadas y las que no fueron afectadas directamente por la catástrofe. La comparación permite ver que la migración que ocurre después de un desastre es en general menos selectiva que la migración desarrollada en otros contextos. Por ejemplo, el género y la edad son fuertes predictores para el movimiento desde áreas no dañadas, pero no están relacionados con los desplazamientos desde áreas que sufren daños. En nuestro análisis, los tradicionales predictores de vulnerabilidad no siempre operan en las direcciones anticipadas. Bajos niveles de estatus socioeconómico y educación no fueron predictivos del movimiento después del tsunami, aunque para aquellos que en realidad se desplazaron, fueron predictivos del desplazamiento hacia un campamento y no hacia una casa privada. Este enfoque basado en estudio de campo, así no esté libre de dificultades, es de amplia aplicabilidad a muchos tópicos de la investigación humano-ambiental y potencialmente abre la puerta a la prueba rigurosa de nuevas hipótesis en esta literatura.

Key Words:

• disaster
• displacement
• Indonesia
• migration
• tsunami

关键词:

• 灾害
• 迁移
• 印尼
• 迁徙
• 海啸。

Palabras clave:

• desastres
• desplazamiento
• Indonesia
• migración
• tsunami

Notes

¹Following recent reviews (National Research Council 2006; Intergovernmental Panel on Climate Change 2012), we refer to biophysical events that place humans at risk as natural hazards and to cases in which hazards overwhelm societal coping mechanisms as disasters.

²In the literature, a change in a person's place of residence is typically referred to as mobility. When this change is either forced by or reflects a voluntary response to a hazard or disaster, it is referred to as displacement (Deng 1998). Migration commonly refers to a change in residence that crosses some minimum threshold of distance. In this article we examine change of residence and refer to it as mobility because we are analyzing individuals across a continuum of disaster-related destruction (see Hugo 1996). We recognize that for many individuals the change reflects displacement in that it occurred as a direct result of the tsunami.

³The 2004 SUSENAS survey occurred before the events that brought peace to Aceh in 2005, which complicated fieldwork in both 2004 and 2005, particularly in rural areas along Aceh's north coast. We implemented a number of procedures to address possible data quality issues as a result of the security situation (they are described in more detail later).

⁴Information about ownership of assets and networks is missing for a very small number of cases (<0.5 percent). To account for these missing data, we include indicator variables that identify these cases in the regression models.

⁵Our desa-level questionnaire included a history section, in which leaders were asked to report (since 2000) any of various events of significance, including issues of political security. Of the 981 event reports, just over 10 percent involved political security, of which only 26 (or less than 3 percent) were described as conflict (typically between GAM and the Indonesian government).

⁶The reported differences in the fifth column are odds ratios calculated using the coefficients on interactions between a binary indicator for tsunami damage and each of the covariates included in the model and estimated with the full sample and all the covariates.

⁷The estimates in the final three columns of Table 3 can be interpreted as the outcome of a two-step or nested process in which an individual chooses whether or not to move and, conditional on that choice, the individual chooses the destination. The first step is reflected in the sixth column and the second step in the seventh column (for distance) and eighth column (for type of destination). We have also estimated the model using a multinomial logit specification that imposes the assumption of the independence of irrelevant alternatives. The coefficient estimates for these models and the models reported in the table are very similar for both the choice between staying in the desa or going outside the desa and for moving to a camp or to a private home. Because the multinomial logit estimates impose more structure, the standard errors are smaller and some of the odds ratios in those models are significant but not in the models reported in the table.