1. To put this in historical perspective, the 1988 floods had an average duration of 34 days while the 1998 floods lasted, on average, 59 days (del Ninno et al., 2001 Del Ninno, C., Dorosh, P., Smith, L. and Roy, D. 2001. The 1998 Floods in Bangladesh: Disaster Impacts, Household Coping Strategies, and Response, Washington, DC: International Food Policy Research Institute. Research Report 122 [Google Scholar]).

2. The survey has been used to account for immediate (del Ninno et al., 2001 Del Ninno, C., Dorosh, P., Smith, L. and Roy, D. 2001. The 1998 Floods in Bangladesh: Disaster Impacts, Household Coping Strategies, and Response, Washington, DC: International Food Policy Research Institute. Research Report 122 [Google Scholar]; del Ninno et al., 2003 Del Ninno, C., Dorosh, P. and Smith, L. 2003. Public policy, markets, and household coping strategies in Bangladesh: avoiding a food security crisis following the 1998 floods. World Development, 31(7): 1221–1238. [Crossref] , [Google Scholar]) and long-term 1998 flood damages with a specific focus on asset losses, consumption declines, reductions in nutritional status, and human capital accumulation (Quisumbing, 2005a Quisumbing, A. 2005a. “A drop in the bucket? The impact of food assistance after the 1998 floods in Bangladesh”. Washington, DC: Manuscript, International Food Policy Research Institute.  [Google Scholar], 2005b Quisumbing, A. 2005b. “Targeting and fairness of emergency food assistance: the 1998 floods in Bangladesh”. Washington, DC: Manuscript, International Food Policy Research Institute.  [Google Scholar]; Yamauchi et al., 2008a Yamauchi, F., Yohannes, Y. and Quisumbing, A. Natural disasters, self-insurance, and human capital investment: evidence from Bangladesh, Ethiopia, and Malawi. World Bank Policy Research Working Paper 4909. Washington, DC: World Bank.  [Google Scholar], 2008b Yamauchi, F., Yohannes, Y. and Quisumbing, A. Risks, ex-ante actions, and public assistance: impact of natural disasters on child schooling in Bangladesh, Ethiopia, and Malawi. World Bank Policy Research Working Paper 4910. Washington, DC: World Bank.  [Google Scholar]).

3. A thana (referred to as upazila by the present government) is an administrative unit that is smaller than a sub-district and larger than a village.

4. The sample is representative of the post-flood population in each thana with the exception of the Saturia thana (del Ninno et al., 2001 Del Ninno, C., Dorosh, P., Smith, L. and Roy, D. 2001. The 1998 Floods in Bangladesh: Disaster Impacts, Household Coping Strategies, and Response, Washington, DC: International Food Policy Research Institute. Research Report 122 [Google Scholar]). Sampling in the Saturia thana is based on a previous IFPRI survey, and therefore is representative of the pre-flood population.

5. Our parameter estimates may be subject to recall bias, since our baseline figures are based on the recollection of workers one year prior to the first survey round. Recent work on displaced workers suggests respondents tend to overestimate past earnings, though such error explains a relatively small fraction of wage variation (Oyer, 2004 Oyer, P. 2004. Recall bias among displaced workers. Economics Letters, 82(3): 397–402. [Crossref] , [Google Scholar]). Bound and Krueger (1991) find minimal bias can be attributed to measurement error of this sort, while other studies, for example, Duncan and Hill (1985 Duncan, G. and Hill, D. 1985. An investigation of the extent and consequences of measurement error in labor economic survey data. Journal of Labor Economics, 3(4): 508–522. [Crossref], [Web of Science ®] , [Google Scholar]) have demonstrated the opposite. The potential for inflationary self-reports of pre-flood wages in the Bangladeshi sample may potentially cause an upward bias in the losses calculated in this study. Unfortunately, data limitations prevent a validation exercise with record data.

6. Because Buttenheim's (2006 Buttenheim, A. Flood exposure and child health in Bangladesh. California Center for Population Research Working Paper, CCPR-022-06.  [Google Scholar]) flood exposure measure is constructed at the household level, endogeneity is a valid concern in her analysis. Vulnerability to floods may be correlated with unobserved characteristics at the household level that also affect child health. To address this issue, Buttenheim designs a natural experiment by comparing the difference in anthropometric measures between a cohort of children in the 1998 round of the survey that were and were not flood-exposed to the difference in anthropometric measures between a comparable cohort of children in the 1999 round of the survey that were and were not flood exposed. She also applies a second approach that allows for measurement of cross-cohort effects controlling for individual fixed effects; however the estimates remain sensitive to the endogeneity of flood exposure.

7. The weeks worked variable is subject to measurement error, as a significant portion of households (26%) report greater than 13 weeks, where the maximum value was 16 weeks.

8. Inclusion of the time invariant flood variable precludes the estimation of a household fixed effect version of Equation (1). Although the random effects model can be more efficient, it also assumes that unobservable household characteristics are independent of the covariates.

9. Thana by year (rather than village by year) fixed effects are the appropriate indicator to account for time variant differences in infrastructure, because decisions regarding allocation of funds for (say) infrastructure development are typically not made at the village level, and non-government organisations or microfinance institutions may have offices at or near the thana centre (rather than in the village).

10. We note the possibility of no effect. First, emergency relief programmes may have cushioned markets. Second, it is possible that the out-migration of labour in response to the flood causes the opposite effect. For example, if workers can freely migrate and in-migration is non-existent, labour supply may decrease yielding a positive effect on wages.