This article provides a study design and analytic methodology for evaluating and comparing the quality of survey data in the case of face to face and telephone interviewing. Under the proposed design, the mode differences are decomposed into measurement bias and non„response bias components. The measurement bias is estimated from the interview-reinterview data using latent class analysis which simultaneously estimates the true prevalence rates and the classification error rates for the measures of the characteristics of interest. Nonresponse bias for the face to face survey is estimated from a followup survey of the face to face nonrespondents. Nonresponse bias for the telephone survey is estimated using an error-corrected estimator of the true prevalence rate. The methodology is illustrated using data from a special study conducted for the U.S. National Health Interview Survey (NHIS). Although the study population is limited to Texas and California, the analysis provides new insights regarding the nature of the mode effects for these two interview modes while illustrating an innovative design for assessing mode bias.