OBJECTIVES: Conditional expectation methods have been widely used for their simplicity, and often without adequate justification. We tested the performance of treatment effect estimators of two conditional expectation methods: the linear outcome-equation model and a propensity score method, under different levels of unobserved selectivity. METHODS: We used the Monte Carlo method to generate data with varied degree of unobserved selectivity. The parameters in the simulation were calibrated using an asthma patient population selected from Medi-Cal eligibles between January 1995 and December 2000. The simulated outcome variable can be considered as log-cost, or any continuous outcome variable in healthcare. For each level of unobserved selectivity, 500 data sets were generated each with 1500 observations. Both the treatment effect in the general population (ATE) and the treatment effect in the treated population (TT) were estimated. Two conditional expectation methods, a linear outcome-equation model and a propensity score method, were applied to estimate treatment effects. The sensitivity of treatment effect estimators was measured using root mean square relative error (RMSRE) and relative bias. RESULTS: Our result showed that both conditional expectation methods were sensitive to unobserved selectivity. The relative bias of ATE estimate from the linear outcome-equation model (propensity score method) was greater than 50% when the combined unobserved selectivity (treatment group and the control group) was greater than 15% (8%). The result for RMSRE showed similar level of sensitivity. The results also showed that a consistent estimate of treatment effect on the treated (TT) could be achieved under conditions less strict than those required to achieve a consistent estimate of average treatment effect (ATE). CONCLUSIONS: Our study shows that the estimates from either conditional expectation methods are acceptable only when there is no, or small, unobserved selectivity, a condition not satisfied in many retrospective healthcare studies, especially those based on administrative claims databases.