Agricultural Productivity and Industrial Growth. Evidence from Brazil

This project studies the effect of productivity growth in agriculture through the adoption of a new technology, genetically modified soybean seeds, on industrial development in Brazil.

Grants Round: 

Productivity growth in agriculture can accelerate industrial growth by: releasing labour from the agricultural sector; raising incomes (which increases demand for manufactured goods); and increasing savings (which provides finance for entrepreneurs in the industrial sector). Some of these positive effects only occur in closed economies, however. In open economies comparative advantage in agriculture may actually retard industrial growth. Estimating the direction and size of these effects is therefore an important issue for designing policy.

This project aims to provide empirical evidence on the strength of each of these channels, using firm-level data to study the effects of the adoption of a new agricultural technology, genetically modified soybean seeds, on industrial development in Brazil. During the decade after the technology was introduced in 1996, the output of soy doubled in Brazil, becoming the most important crop in the country. To identify the causal effects of the change in agricultural productivity brought by the new technology, we use two sources of exogenous variation in the profitability of technology adoption. First, as the technology was invented and commercially introduced in the U.S. only in 1996, it clearly did not exist before this date. Thus, we use the period before 1996 as our pre-treatment period, and the period afterwards as our treatment.

Second, the new technology has the potential to increase yields of soy in some areas, but not in others, depending on geographical and weather characteristics. We use data on theoretical soil yields at a very disaggregated geographical level. These yields are calculated by incorporating local soil and weather characteristics into a model predicting yields for each crop given certain climate and soil conditions. Since these predicted yields rely only on weather and soil information, but not on actual yields, they are exogenous. We can then exploit the predicted differential impact of the new technology on yields across municipalities in Brazil as our source of cross-sectional variation in agricultural productivity. This design allows us to investigate whether municipalities with higher predicted increases in agricultural productivity also experienced higher increases in agricultural yields and income. More interestingly, it allows us to understand whether exogenous shocks to local agricultural productivity lead to increases in the size and efficiency of the local industrial sector. We will use firm level datasets to construct on measures of industrial development such as the size of the industrial sector, aggregate and firm-level productivity, and firm-level investment in new technologies and product quality.