More Cotton, Same Workers: How Bt Technology Raised Output without Shifting Labor in India
Bt cotton significantly increased cotton acreage and output in India, but these gains were uneven and constrained by land scarcity, especially in labor-abundant districts. Despite major productivity improvements, labor market frictions prevented workers from moving out of agriculture, leaving India’s employment structure largely unchanged
Produced within the World Bank’s Development Research Group and drawing on evidence from the Food and Agriculture Organization (FAO), the International Crops Research Institute for the Semi-Arid Tropics (ICRISAT), India’s National Sample Survey Office, and the Reserve Bank of India, the study revisits the long-running debate over Bt cotton and structural transformation in India. When genetically modified Bt cotton was approved in the early 2000s, it was widely seen as a breakthrough that could raise farm productivity and release labor from agriculture into manufacturing and services. Two decades on, the evidence shows a more complex outcome: Bt cotton reshaped what farmers grow and how much they produce, but it did not fundamentally alter where India’s workers are employed.
Bt Cotton and India’s Agrarian Turning Point
Bt cotton arrived amid an agronomic crisis. The spread of American cotton hybrids had left Indian farmers increasingly vulnerable to pests, particularly the American bollworm, triggering heavy insecticide use, rising costs, and mounting debt. Bt cotton, engineered to kill bollworms from within the plant, dramatically reduced pesticide requirements and labor-intensive crop management. Approved first in central and southern states in 2002 and later in northern India in 2005, adoption was initially slow but then accelerated sharply. By the early 2010s, Bt cotton covered nearly all cotton acreage, transforming India into one of the world’s largest cotton producers.
Productivity Gains without Labor Exit
Using district-level data from 1990 to 2011 and exploiting variation in agro-climatic suitability for cotton, the study estimates the causal impact of Bt cotton through a difference-in-differences framework. The results on production are unequivocal. Districts that stood to gain more from Bt cotton expanded the share of land devoted to cotton and recorded large increases in output, particularly after 2004–05 when adoption intensified and seed prices fell following state interventions. By 2011, cotton acreage in highly exposed districts had risen by more than one-third relative to pre-adoption levels, while output increased by around 30 percent.
Yet employment patterns tell a different story. Despite these large productivity gains, the share of the working-age population employed in agriculture remained essentially unchanged. Across multiple datasets, specifications, and robustness checks, there is no evidence of a significant reallocation of labor out of farming. The technology raised output, but it did not free workers in any meaningful way.
Why Land and Labor Matter
To explain this disconnect, the paper emphasizes relative factor endowments. Bt cotton is best understood as a labor-augmenting technology: it allows farmers to produce more with the same labor input. But cotton cultivation in India relies on land and labor as complementary inputs, not substitutes. Where land is scarce relative to labor, productivity gains quickly run into diminishing returns.
Empirically, this mechanism is clearly visible. Districts with high labor-to-land ratios experienced significantly smaller increases in cotton acreage and output than land-abundant districts facing similar technological improvements. In labor-rich areas, Bt cotton effectively made labor even more abundant relative to land, intensifying land constraints and limiting expansion. Importantly, these sharp differences in production responses did not translate into differences in employment shares, reinforcing the idea that land scarcity and diminishing returns shape outcomes long before labor reallocation occurs.
Frictions that Freeze Structural Change
Labor market frictions provide the final piece of the puzzle. Using pre-existing migration intensity of low-skilled workers as a proxy for labor mobility, the study shows that districts with fewer migration barriers did see some increase in agricultural employment following Bt cotton’s introduction. Workers flowed in from non-agricultural activities and from non-employment. However, these effects were modest and uneven, and they disappeared in districts where migration costs were high.
Paradoxically, these same frictions help explain why cotton expansion was sometimes larger in less mobile districts. When labor cannot move easily, labor-to-land ratios rise more slowly, delaying the onset of diminishing returns to land and allowing cotton acreage and output to expand further. The paper also rules out alternative explanations based on crop variety or quality: Bt cotton led to nationwide convergence toward American upland cotton, and farm-gate prices show no systematic differences across districts.
Rethinking Technology and Transformation
The central lesson is sobering but important. Technological change alone does not guarantee structural transformation. When inputs are complementary and one factor, such as land, is scarce, labor-augmenting innovations may deliver large output gains without reshaping employment. In India’s case, Bt cotton transformed cotton production but left the broader employment structure intact. More broadly, the study cautions against assuming that agricultural productivity growth will automatically release labor. Endowments, institutions, and frictions ultimately determine how economies adjust, and whether technological promise translates into structural change
- FIRST PUBLISHED IN:
- Devdiscourse

