Does industrial water pollution harm agricultural production?

While toxic foam regularly appears on lakes and rivers in major Indian cities, water pollution has not received as much attention as air pollution. Examining the impact of industrial water pollution on agriculture, this article demonstrates that there is a large, sudden rise in pollutant concentrations in nearby rivers downstream of industrial sites. Despite this, there is no significant impact on crop yields.

Pollution levels in low- and middle-income countries (LMICs) are often far worse than in high-income countries. Despite this, most of what we know about the costs of pollution comes from developed countries (Keiser and Shapiro 2019, Currie and Walker 2019), with little basis to extrapolate these findings to LMICs. In India, water pollution has not received as much attention as air pollution, even though it is a major issue. Toxic foam regularly appears on lakes and rivers in cities like New Delhi and Bengaluru (Möller-Gulland 2018), and fish die-offs are now a common occurrence (Vyas 2022). While public pressure has led to stricter air pollution regulations, similar efforts have not significantly improved water quality (Greenstone and Hanna 2014).

Even in high-income countries, it has been tricky to put a clear price tag on the social costs of water pollution. Although people care deeply about water quality according to surveys, studies often fail to find big economic impacts of water pollution. This might mean the costs are genuinely low – or it could just be that water pollution is hard to study. Limited data, the challenges of modelling how water pollution moves, and the sheer number of pollutants have all made this a difficult question to study, potentially leading to underestimates of its true effects (Keiser and Shapiro 2019).

Our study

We look at how industrial water pollution affects agricultural output in India (Hagerty and Tiwari 2024).¹Agriculture is an obvious sector to focus on because it uses more water than all other sectors combined – about four times as much, as per estimates by the Food and Agriculture Organization (FAO) (2018) – and irrigation water is almost never treated. Agriculture can also be found almost everywhere, so it is likely to be affected by the major sources of pollution. We zoom in on 48 industrial sites labeled “severely polluted” by India’s Central Pollution Control Board in 2009. These sites are among the world’s most polluted industrial areas (Mohan 2021), making them an important place to study the impact of water pollution.

How does water pollution affect crops?

Industries dump all sorts of waste into water, including heavy metals, effluents with high salinity or abnormally low or high pH, and toxic organic compounds. These can seriously harm crops by stunting growth or affecting nutrient uptake (Bajpai 2013, Sudarshan et al. 2023, Scott et al. 2004). Experiments using polluted water for irrigation have found major impacts on rice, such as poor growth and bad taste (World Bank and State Environmental Protection Administration, 2007). But such experiments do not replicate real-world situations, where pollution from the source can follow complicated pathways before it reaches farms. Our study addresses this problem by carefully modelling how pollution reaches crops.

Importantly, not all industrial waste is harmful. Some effluents contain useful chemicals like nitrates, phosphates, and potassium – the same ingredients found in fertilisers. If these chemicals show up in small amounts, they can actually help crops grow (Hawkins and Risse 2017, Bedane and Asfaw 2023, Zhang and Lu 2024). Because of this mix of harmful and helpful effects, the overall impact of industrial pollution on crops is an empirical question that we address in our research.

Research design

Our research design exploits a key feature of water pollution: unlike air pollution, it almost always flows in only one direction from its source. When industries release wastewater into rivers, pollution levels spike downstream while remaining relatively unaffected upstream; yet these areas are likely similar otherwise. This creates a natural comparison between downstream and upstream areas, allowing us to assess the economic impact of water pollution.

We address three key methodological challenges:

i) Sparse monitoring data: Instead of isolating the effect of specific pollutants, we estimate the overall impact of highly polluting industrial sites, bypassing the limitations of noisy, infrequent, and biased water quality monitoring data.

ii) Complex pollution transport: We use precise hydrological models to better map upstream and downstream areas relative to the industrial sites.

iii) Lack of high-resolution crop yields: To measure crop yields, we employ machine learning to predict yields from six different satellite-derived vegetation indices including NDVI (normalised difference vegetation index), EVI (enhanced vegetation index) and others. These indices, developed by earth scientists, are reliable predictors of crop yields across diverse settings (Running et al. 2004, Burke and Lobell 2017, Lobell et al. 2022). Using village-level microdata, we train multiple models and select the one that performs best. This model has nearly four times the predictive power of earlier approaches in the literature.

Findings

The first big takeaway is that industrial sites considered “severely polluted” release massive amounts of pollutants, contributing heavily to water pollution. Figure 1 shows that surface water pollution jumps three to six times in rivers downstream of these sites compared to upstream areas. This scale of pollution from these sites has not been publicly quantified before.

Figure 1. Surface water pollution measurements at monitoring stations

Notes: (i) Graphs plot mean values of each parameter within quantile bins of distance from the industrial site. Positive distance indicates a monitoring station is downstream of the site; negative is upstream. The fitted polynomial lines indicate how pollution might change with distance (ii) The estimated effect of pollution at the site is reported within each panel along with the robust p-value.

Our second key finding is surprising: crop yields downstream of polluted sites are not much lower than those upstream. Figure 2 shows a small 3% drop in yields, but the 95% confidence interval includes zero², and we can rule out yield declines greater than 7%. This suggests that even the localised impact of pollution on crop yields is small. These effects are likely even smaller further downstream, where pollution levels dissipate.

However, larger yield reductions are observed in areas where pollution exposure is more likely, such as villages that are served by canals, those near rivers, or those that have shallow groundwater so that pollution can more easily percolate. For instance, yields in canal-irrigated villages drop by a statistically significant 10%, as shown in Panel (b) of Figure 2.

Figure 2. Average crop yields in the village, as predicted from satellite data

Notes: (i) Graphs plot mean values of each parameter within quantile bins of distance from the industrial site. Positive distance indicates a monitoring station is downstream of the site; negative is upstream. The fitted polynomial lines indicate how crop yields might change with distance (ii) The estimated effect of pollution at the site is reported within each panel along with the robust p-value. (iii) Panels (b), (c), (d) restrict the village sample.

Why are the effects small?

Three reasons help explain why pollution may not have a big overall impact on agricultural output:

i) Limited crop exposure: Not all crops are exposed to polluted water. Groundwater pollution, for instance, does not increase much across the full sample.

ii) Pollution dilution: As pollution travels, it gets diluted through sedimentation, filtration, and diffusion, which reduces its concentration by the time it reaches farms.

iii) Beneficial components: Industrial effluents sometimes contain nutrients that act like fertilisers. We find suggestive evidence that sites that release more nutrients have smaller effects on crop yields.

We also checked if farmers might be offsetting pollution damage by changing agricultural practices, like using more fertilisers or altering irrigation patterns. There is little evidence of change in inputs, and there is no noticeable impact on household consumption or poverty rates.

Conclusion

We studied how water pollution from severely polluted industrial sites in India affects agricultural output. These sites cause significant jumps in river pollution, but their impact on crop yields is surprisingly small. This is likely the case because (i) most crops are not directly exposed to pollution, (ii) pollution is diluted before it reaches farms, and (iii) some pollutants include nutrients that benefit crops. This finding is also in line with the literature from the developed world, which mostly does not find large non-health benefits of cleaning up water pollution (Keiser and Shapiro 2019).

However, our findings do not imply that industrial water pollution is harmless. While its impact on agriculture might not be large, there could be other serious costs, such as to human health and to ecosystems. Understanding these broader effects is an important area for future research.

Notes:

1. Various data sources have been used for the study, such as Census (2011), Socio Economic and Caste Census (2012), SHRUG (Socioeconomic High-Resolution Rural-Urban Geographic Platform) database, Ministry of Agriculture’s survey on cost of cultivation (2015-2018), and so on.

2. A 95% confidence interval (CI) means that, if you were to repeat the experiment over and over with new samples, 95% of the time the calculated CI would contain the true effect. CIs that do not include zero imply statistical significance at the 5% level.