A household-level switching regression model is implemented to examine potential selectivity bias for rural households under high and low levels of investments in soil conservation in El Salvador and Honduras. In the presence of selectivity bias, separate stochastic production frontiers are estimated for low and high adopters. The main results indicate that households with higher levels of investments in soil conservation show higher average TE than those with a lower level of investments. Constrains in the rural land and credit markets are likely explanations for these differences. The results also indicate that for farms with lower levels of investments in soil conservation access to credit is a significant factor explaining the sources of inefficiency. Conversely, households with higher levels of investments have the highest partial output elasticity for land, the highest levels of TE and the smallest farms. These results are consistent with the presence of a failure in the land market which would limit access to land to the more efficient producers.