Examines the extent to which one-time surveys can obtain precise unbiased estimates of expenditures for outpatient medical care. Survey data are evaluated for item nonresponse, bias, and measurement error. Item nonresponse appears to be high for persons whose health care is financed through the public sector, but is otherwise infrequent. A comparison of nonsurvey estimates of mean total and out-of-pocket physician and dental expenditure with estimates from two surveys using indirect methods (including RAND's Health Insurance Study) revealed that out-of-pocket physician expenditure is overestimated but that other measures are unbiased. Estimates using direct, self-administered methods appear to be biased upward. Commonly used record check methods for evaluating survey data will make random error appear as bias. A procedure developed in the study for estimating response error variance shows substantial error in the measurement of dental expenditure: 44 percent of the total variance in survey data and 39 percent of the variance in records.