Trade and pollution: Evidence from India

TL;DR

This paper investigates whether openness to trade, as induced by India’s 1991 liberalization, worsened local water pollution. Using a district-level panel (1987–1997) that links tariff exposure to river-water quality across 117 districts, the authors employ a shift-share identification with district and year fixed effects, finding that larger tariff reductions raise water pollution, with the median district experiencing about a 0.11 standard deviation increase in the pollution index for a 4 percentage point tariff drop. The analysis leverages eight water-quality metrics and an inverse-covariance weighted index to summarize pollution, and robustness checks—including alternative index constructions, harmonized boundaries, and various standard-error specifications—support the main finding. The results imply that trade liberalization in a developing country can generate environmental costs through industrialization and structural transformation, underscoring trade-offs policymakers must weigh when promoting openness against environmental quality and health outcomes.

Abstract

What happens to pollution when developing countries open their borders to trade? Theoretical predictions are ambiguous, and empirical evidence remains limited. We study the effects of the 1991 Indian trade liberalization reform on water pollution. The reform abruptly and unexpectedly lowered import tariffs, increasing exposure to trade. Larger tariff reductions are associated with relative increases in water pollution. The estimated effects imply a 0.11 standard deviation increase in water pollution for the median district exposed to the tariff reform.

Figures (28)

• Figure 1: Distribution and trend of district-level tariff exposure. Lines show the listed percentile of the data. Shaded area indicates the reform period. Data for 1993 are imputed following topalova2007topalova2010factor.
• Figure 2: a) Trends in water quality, averaged across districts with above and below median changes in tariff exposure (in absolute magnitude). b) Point estimates from a regression of the pollution index on pre-reform average tariffs, interacted with event year indicator variables, omitting 1991, and controlling for district and year fixed effects. Shaded area indicates the reform period.
• Figure 3: Results from the main regression specification, where the estimates show the standardized effect of a contemporaneous change in the tariff measure. More extreme values suggest larger effects of trade on pollution. For DO, the scale is reversed, since DO negatively correlates with pollution levels. District and year fixed effects included in all regressions. Standard errors are clustered by district. Dark grey confidence intervals at the 90% level; lighter grey confidence intervals at the 95% level.
• Figure 4: Results from the falsification exercise, where each water quality metric and the water pollution index are regressed on a four-year lead of the tariff measure. More extreme values suggest larger effects of trade on pollution. For DO, the scale is reversed, since DO negatively correlates with pollution levels. District and year fixed effects are included in all regressions. Standard errors are clustered at the district level. Sample includes data for years between 1987 and 1993. Dark grey confidence intervals at the 90% level; lighter grey confidence intervals at the 95% level.
• Figure 5: Robustness Main specification: inverse covariance-weighted index, contemporaneous effects, district and year fixed effects, standard errors clustered by district. Robustness tests vary analysis choices, as follows: Index, no-missing sample: only district-year observations with no missing data for any metrics. PCA, imputed sample: PCA index, with missing variables imputed. PCA, no-missing sample: PCA index, only district-year observations with no missing data for any metrics. Winsorized: extreme 2 or 4 percentiles of pollution data winsorized. Station data, district FE: main specification using monitor station-level data. Station data, station FE: monitoring-station fixed effects. State-year FE: global trends modeled with state-year fixed effects. State-year cluster: standard errors clustered by state-year. Two-way cluster: standard errors clustered two-way by district and year. Conley: Conley standard errors in addition to clustering by district, over 500 and 1000 km respectively. Lagged: effects of tariffs lagged one year. Non-adjusted: tariff data without adjusting value in 1993. Harmonized districts: unit of analysis is a harmonized district liu2023climate, with area-weighted tariff and pollution variables. Excl. districts w. ban: drops 3 districts affected by a local ban on leather tanneries. Population weighted: using district population in 1991 as regression weights. Dark grey confidence intervals at the 90% level; lighter grey confidence intervals at the 95% level. N = 1,146 observations (2,481 at the station level) in 117 districts, except in non-missing sample (N = 740 in 99 districts), with lagged tariffs (N = 1,029 in 117 districts), harmonized districts (N = 950 in 96 districts), and excluding districts with the tannery ban (N = 1,117 in 114 districts).