A shadow-price profit frontier model is developed to examine production efficiency of Chinese rural households in farming operations. The model incorporates price distortions resulting from imperfect market conditions and socioeconomic and institutional constraints, but retains the advantages of stochastic frontier properties. The shadow prices are derived through a generalized profit function estimation. The shadow-price profit frontier is then estimated and an efficiency index based on the estimated profit frontier is computed and decomposed to household characteristics. Empirical results using data from China's Rural Household Survey for 1991 reject the neoclassical profit maximization hypothesis based on market prices in favor of the general model with price distortions. Farmers' resource endowment and education influence their response to the market restrictions, thus alter their performance in terms of efficiency. The estimated efficiency index ranges from 6% to 93% with a sample average of 62%. Households' educational level, family size and per capita net income are positively related to production efficiency. Households living in mountain areas or with family members employed by the government or state industries are relatively inefficient. Reducing market intervention, allowing right of use of farm land to be transferred among households, encouraging migration of excess farm labor, and promoting farmers' education will improve rural households' efficiency in agricultural production.