Natural disasters can have devastating long-term impacts because they can impede the accumulation of physical and human capital stock (Skoufias 2003; Yamauchi, Yohannes, and Quisumbing 2008a, 2008b). It is now widely accepted that climate change will not only increase the frequency of two types of natural disasters that affect agriculture and rural households- droughts and floods- but also alter rainfall patterns, thereby changing farming practices, household behavior, and welfare. According to the IPCC Fourth Assessment Report, anthropogenic emissions may be responsible for at least a 40-centimeter sea-level rise by the end of the 21st century (IPCC 2007). Such increases in sea levels cause the salinization of groundwater and surface water sources, jeopardizing the supply of drinking water and the capacity to produce crops and displacing populations.
Examining the relationship between severe weather events and wages is particularly relevant for the design of future development and climate change strategies. Households migrate or seek labor in rural agricultural and nonagricultural markets to diversify their portfolio and buffer against risk (Kochar 1999; Rose 2001; Cameron and Worswick 2003; Takasaki, Barham, and Coomes 2010). It therefore is crucial to understand how resilient those markets are to shocks and the extent to which the markets can absorb the excess labor induced by shocks. Policies aimed at improving the reallocation of labor aftershocks (e.g., by protecting or facilitating migrant labor markets) may be a lower-cost alternative to other investment-heavy candidates.
In Bangladesh, annual flooding is considered a normal part of the agricultural cycle. However, severe floods, such as the one that occurred in 1998, can have devastating short-and long -term impacts. Unlike the "normal" floods that occur annually, the 1998 flood lasted until mid-September in many areas, covering more than two-thirds of the country and causing more than 2 million metric tons of rice crop losses (equal to 10.45 percent of target production in 1998-1999) (del Ninno et al. 2001). Using district-level data, a recent study evaluates the impact of riverine floods on agricultural wages in Bangladesh (Banerjee 2007). Banerjee finds that agricultural wages decline by 5 percent in flood-prone areas and by 14 percent in severely exposed areas during "extreme" floods in the short term. We build upon that work by evaluating both the short-term and long-term effects of the most severe flood experienced by households in Bangladesh, using a household panel survey that was specifically collected for this purpose.1 In addition to using a household panel, our data extend beyond the time period used in Banerjee to account for the long-term impacts of the 1998 flood. Our paper also makes an additional contribution by measuring the flood effect on nonagricultural wages and identifying specific mechanisms that dampen damages in the short and long term.
We estimate reduced-form wage regressions using the Bangladesh Flood Impact household panel survey spanning immediately after the September flood to five and a half years postflood. Our identification strategy depends on the inclusion of district and time fixed effects to control for unobserved spatial and time heterogeneity, as well as reported preflood wage information to control for the initial labor market conditions of each village. We find that for every one-foot deviation from the usual flood depth, daily wages on average declined approximately 2 percent. Upon distinguishing between short-and long-term effects on wages, we find that the long-term impacts dominate. In particular, variations in wages a year after the event are not attributable to the deviations in the flood depth that occurred in 1998. However, we do find a statistically significant impact of the magnitude of the 1998 flood on the variation of wages after five and a half years. The persistence of damages five years past the natural disaster is consistent with findings related to drought effects on labor markets in Brazil (Mueller and Osgood 2009) and on growth patterns in Ethiopia (Dercon 2004). Wages declined between 4 and 5 percent for every one-foot increase from the usual flood depth more than five years after the major flood. Further distinguishing between agricultural and nonagricultural labor markets, we find that agricultural markets experienced a wage loss of 4 percent for every one-foot increase, which remains constant over time. In contrast, nonagricultural labor markets experienced a greater decline in wages of around 7 percent per one-foot increase in the flood depth from normal conditions, with greater losses over time. The persistent negative impact of the flood on credit dependence in Bangladesh may be partially responsible for the negative long-term impact on investment in labor and other related markets (del Ninno, Dorosh, and Smith 2003). Our findings also corroborate the limitations of the food assistance programs in Bangladesh, which enhanced food availability in the short term but had no bearing on long-term household purchasing power (Quisumbing 2005a).
We also evaluate the roles of factors that can cushion labor markets from severe flood damages. In particular, we measure how soil drainage capacityand proximity to markets and bazaars may dampen the impact of floods on wages. Nearby markets and bazaars may provide surplus workers access to additional outlets for employment. Our results indicate that labor markets in predominantly clay-soiled areas (with low drainage capacity) were more severely affected than other areas in the short term. We also find that labor markets closer to the weekly market or bazaar were less affected than those further away. This suggests that the lack of auxiliary labor markets for workers exacerbated the impact of floods on wages. Although the analysis is representative of a modest number of villages in Bangladesh, our work suggests that future development and disaster relief policies might consider increasing ss to additional labor markets through investments in infrastructure and transportation.
In what follows, we provide a theoretical framework and a review of the literature describing why wages may be affected by natural disasters in the long term (Section 2). In Section 3, we describe the household panel survey. We present our empirical model, identification strategy, and empirical results in Section 4. Our concluding remarks are discussed in Section 5.
- International Food Policy Research Institute
- Copyright © International Food Policy Research Institute