This paper verifies the existence of the inverse productivity relationship in the ICRISAT data and examines the role that land quality and imperfect markets play in generating it. The empirical approach is straightforward, and allows for more direct tests of the land quality or imperfect labor market hypotheses than previous studies. The ICRISAT information on plot-level soil quality and measures of labor and land market failures are used to see if either land quality or market failure can fully account for the inverse relationship. Differences in land quality measured by soil type, irrigation, and the value of farm land largely explain the IP relationship in the random effects profit regression, but not in labor demand regressions. Controlling for labor and land market failures and differences in soil quality eliminates the IP relationship for male labor, but not female labor in the random effects estimates. The inverse relationship is much stronger in fixed-effects than random-effects estimates, suggesting that the farm size variable may be subject to measurement error. Instrumental variables estimation is used to correct possible measurement error in farm size, wiping out the IP relationship in the fixed effects estimates of both profits and female labor demand.