In this dissertation, I present three essays that consider the environmental consequences of technological change, from an international perspective. The first two chapters use firm-level production data to estimate the response of CO2 emission intensity to changes in competition in foreign markets. The first chapter estimates this response with respect to foreign demand shocks, i.e., a positive shock to exports. The second chapter exploits a specific liberalization episode to estimate the impact with respect to foreign competition shocks, i.e., a negative shock to exports. Both papers are co-authored with Helene Ollivier. The final chapter analyzes the decision to adopt genetically engineered seeds in different countries around the world, and the attendant impacts on supply and land-use. This last chapter is co-authored with David Zilberman and Steven Sexton and was previously published in Environment and Development Economics.

The first chapter investigates the impact of exporting on the CO2 emission intensity of manufacturing firms in India. Recent papers have argued that export market access encourages firms to upgrade technology, which lowers the emission intensity of production; however, data limitations confound previous attempts to separately identify productivity impacts from simultaneous changes in prices and product-mix. We present a model of how these alternative channels could also explain the results documented in the literature. Then, using a highly detailed production dataset of large Indian manufacturing firms that contains information on physical units of inputs and outputs by product, we are able to decompose the overall firm impact into three components -- prices, product-mix, and technology. Export impacts at the firm level are identified from import demand shocks of foreign trading partners. We find that prices systematically bias down estimates of emission intensity in value, that firms adjust emission intensity in quantity through changing output shares across products, but that firms do not lower emission intensity within products over time (technology). The results imply that the productivity benefits from market integration alone are not enough to induce clean technology adoption.

The second chapter investigates the ``third-party'' impact of trade liberalization on the environmental performance of firms in countries that lose market share as a result of the liberalization. If competition matters for exporting (as previous research indicates), and exporting matters for emission intensity, then emission intensity reductions in liberalized markets may be offset by emission intensity increases in countries peripheral to the liberalization. To test for this indirect effect, we exploit quasi-natural variation arising from the elimination of quota constraints on textile and apparel exports to the US between 1994 and 2007. Using a detailed panel of production and emission data at the firm-product level, we find that Indian exporters in Prowess lost on average 14% export sales as a result of liberalized trade between the US and India's competitors. This loss of export sales was accompanied by an increase in CO2 intensity of 9%. The results do not appear to be due to fuel-switching, but there is suggestive evidence that capital investments and switching to higher emission intensity varieties may have played a role. Overall, the results support the importance of international competition for production and pollution decisions of firms around the world.

The final chapter uses aggregate data to estimate supply, price, land-use, and greenhouse gas impacts of genetically engineered (GE) seed adoption due both to increased yield per hectare (intensive margin) and increased planted area (extensive margin). An adoption model with profitability and risk considerations distinguishes between the two margins, where the intensive margin results from direct ``gene" impacts and higher complementary input use, and the extensive margin reflects the growing range of lands that become profitable with the GE technology. We identify yield increases from cross-country time series variation in GE adoption share within the main GE crops- cotton, corn, and soybeans. We find that GE increased yields 34% for cotton, 12% for corn and 3% for soybeans. We then estimate quantity of extensive margin lands from year-to-year changes in traditional and GE planted area. If all production on the extensive margin is attributed to GE technology, the supply effect of GE increases from 5% to 12% for corn, 15% to 20% for cotton, and 2% to 40% for soybeans, generating significant downward pressure on prices. Finally, we compute ``saved" lands and greenhouse gases as the difference between observed hectarage per crop and counterfactual hectarage needed to generate the same output without the yield boost from GE. We find that all together, GE saved 13 million hectares of land from conversion to agriculture in 2010, and averted emissions are equivalent to roughly 1/8 the annual emissions from automobiles in the US.

Increased demand for agricultural produce for food, fiber, feed, and energy generates a tradeoff between high prices and environmentally costly land conversion. Genetically engineered (GE) seeds can potentially increase supply without recruiting new lands to production. We develop a simple adoption model to show how first-generation GE increases yield per hectare. We identify yield increases from cross country time series variation in GE adoption share within the main GE crops- cotton, corn, and soybeans. We find that GE increased yields 34% for cotton, 32% for corn, but only 2% for soybeans. The model also predicts that GE extends the range of lands that can be farmed profitably. If the output on these lands are attributed to GE technology, then overall supply effects are larger than previously understood. Considering this extensive margin effect, the supply effect of GE increases from 10% to 16% for corn, 15% to 20% for cotton, and 2% to 39% for soybeans, generating significant downward pressure on prices. Finally, we compute \saved" lands and greenhouse gasses as the difference between observed hectarage per crop and counterfactual hectarage needed to generate the same output without the yield boost from GE. We find that all together, GE saved 21million Ha of land from conversion to agriculture in 2010, or 0.41 Gt ofCO2emissions (using a constantCO2/land conversion factor). These averted emissions are equivalent to roughly 1/3 the annual emissions from driving in the US.

This paper estimates the effect of coal-red power plants on infant mortality in India. We nd that a one GW increase in coal-red capacity corresponds to a 14% increase in infant mortality rates in districts near versus far from the plant site. This effect is 2-3 times larger than estimates from the developed world. Our effects are larger for: (1) older plants, (2) plants located in areas with higher baseline levels of pollution, and (3) plants burning domestic rather than imported coal. The environmental benefits from policy aimed at the power sector are thus likely to be substantially higher if targeted at older plants located in more polluted areas tailored to burn domestic rather than imported coal.