Research

Floods destroy livestock, equipment and other productive assets across the developing world, yet flood relief often targets only the land-poor. This paper asks who bears the loss, through which village markets, and who adapts. I use a three-round panel of 5,473 households in Bangladesh from 2011 to 2019, matched to satellite flood maps, and compare neighboring Unions of the same district that floods cover unequally. The largest inherited landholders lose livestock and working capital, which is still lower three to four years later. Their farms shrink, and their wage and self-employment income falls by 7.6 percent. The rent falls by a fifth, but no group rents in more land. The labor market rations work instead of lowering the wage. The land-poor lose nearly a quarter of their day labor but turn to self-employment, so their combined income falls by at most 4 percent and their consumption by at most 0.05 percent. Labor can be redeployed and capital cannot, so who bears the loss from a flood depends on who owns the capital it hits, not who is poor. In a model where farms without credit choose when to replace lost animals, a grant to the farms that lose a herd lets them rebuild within a year and is worth about 2 percent of consumption. Fairly priced index insurance that pays the full loss does nearly as well at no expected fiscal cost. The same budget given to the land-poor does not move the land or the labor market.

This paper studies whether climate-induced shifts in land productivity can trigger structural adaptation in agriculture by overcoming long-standing frictions in land markets. In coastal Bangladesh, rising sea levels have pushed dry-season soil salinity beyond the threshold at which conventional crops survive. Using census data on over one million farm households and a fuzzy spatial regression discontinuity at the hydrological boundary between the tidally active and mature regions of the Bengal delta, I find that salinity above 12 dS/m moves 32 to 50 decimals of land per household from crops to aquaculture and raises average farm size by 52 percent. The transition runs through the rental market rather than land sales. Reallocation is faster in villages with greater religious diversity and slower where parcels are more fragmented.

Damage functions in climate economics typically assume that rising income and cooling technology steadily reduce the mortality cost of extreme heat. We test this with a birth-cohort decomposition for 276,031 Bangladeshi births from nine DHS rounds merged with ERA5 daily-maximum temperature. At 30–40°C, only the 1980 cohort shows a significant response; each later cohort's response is close to zero and significantly below it, and a heterogeneity-robust difference-in-differences estimator links the flattening to the staggered rollout of rural electrification, implying roughly 718,000 neonatal deaths averted. Above 40°C the effect is large for the 2010 cohort, about 43 additional deaths per 1,000 births per 10 additional days, and no test detects a decline. The tail effect implies 657 to 1,920 excess neonatal deaths per year by mid-century.

Household surveys diagnose chronic disease from a single measurement of each respondent. Using 34,879 blood-pressure readings and 33,179 fasting glucose results from three Bangladesh DHS rounds, merged with ERA5 temperature at the date and hour of measurement, I test whether they also measure the weather. A one-degree rise in the daily maximum lowers systolic blood pressure on the day it is read by 0.56 mmHg and raises next-morning fasting glucose by 1.03 mg/dL; the response is to humid heat, and a fall in haemoglobin points to dilution rather than dehydration. Measured diabetes rises 0.63 and measured hypertension falls 0.84 percentage points per degree, while reported past diagnoses do not respond, and warmer 2022 fieldwork accounts for about a third of the 7.8-point rise in measured diabetes among adults aged 35 and over since 2017–18.

Son preference raises child mortality through sex selection, discriminatory investment and early weaning; we document a fourth channel, birth spacing. Bangladeshi mothers of two daughters conceive a third child sooner, compressing the interval below the WHO-recommended 24 months. Instrumenting a sub-24-month third conception with the sex composition of the first two births, we estimate a 10.2 percentage point higher risk of third-child neonatal death in Matlab and Chakaria surveillance data and 11.4 points in nine Bangladesh DHS rounds. The child harmed is the one the parents wanted, male with even probability. Nationally, this implies about 35,000 excess neonatal deaths over four decades.

We match three rounds of a Bangladeshi household panel to satellite-observed inundation to measure flood damage to pond aquaculture. Any inundation lowers the probability of operating a pond by 2.9 percentage points, a 16 percent relative decline. Severe floods, which inundate more than half the surrounding area, reduce yield by 48 percent, revenue by 47 percent and profit by about 18,500 taka. No coping margin compensates, and household income falls by more than the pond loss alone. Such floods cost the sector about 26 million dollars a year.