Abstract Erosion is a pressing problem throughout rural China, but the problem is most severe within the Loess Plateau of north central China. The Loess Plateau is losing an average of one centimeter of soil each year due to erosion and related environmental problems. The eroded loess clogs irrigation systems, restricts navigation on all rivers within the region, fouls urban and industrial water systems, and for agriculture represents a loss of a vital resource that this very poor area cannot afford. While land stabilization in the region has been given a very high priority by regional and national planners for economic and environmental reasons, the selection of the most appropriate measures for land stabilization remains controversial. This research, based on a survey of 330 households in Zhongyang County, Shanxi Province, PRC, estimates the increases in gross production and yields which can reasonably be expected from improvements to cultivated slopeland through land stabilization using two different types of terrace construction. The first of these involves the construction of terraces built almost entirely with human labor. The second replaces most of this human labor with earth-moving equipment used to construct broader and considerably more expensive terraces. We estimate and compare differences in yields and production costs for common crops grown on these two types of terraced land as well as two types of unstabilized land. Specifically, the types of land evaluated in the study include: 1) terraces constructed by human labor alone (rengong titian); 2) terraces constructed using earth-moving equipment (jixie titian); 3) flat unterraced land (yuandij not yet severely affected by erosion; and 4) by way of a control, un-improved slopeland (podi]. The crops evaluated are corn, millet, soybeans, sunflower, white potatoes, and sor-ghum-the main subsistence and economic crops in the region. These data are used to generate production functions for each of the six crops to determine the contribution of terrace construction to gross production. Multiple analysis of variance (MANOVA) is used to determine if mean yields for these six crops are significantly different on different land types. Results of these analyses indicate that yields for all crops grown on mechanically constructed terraces are significantly higher than yields for crops grown on terraces constructed by hand, while yields of crops grown on both types of terraces remain lower than for crops grown on flat land which has not yet faced serious erosion. From the perspective of farm households living in this region, the proper management of existing flat land and the increased construction of mechanically constructed terraces are the best strategies for improving output and farm income while also helping to slow erosion. Government agencies beyond the region, however, continue to favor terraces constructed by hand owing to lower costs. The implications for local, regional, and national policy directives are discussed.