Conventional small area estimation methods combine generalized linear model synthetic estimates made using covariates with direct survey estimates. Since "borrowing strength" from covariates to make quality synthetic estimates is a key motivation in small area modeling while almost all such information is collected through surveys, we recognize the need for building models that combine survey data and incorporate uncertainties in both surveys. In this study, we use the American Community Survey (ACS) to improve the disability estimates from the Survey of Income and Program Participation (SIPP). In particular, we discuss the estimation results from a bivariate Fay-Herriot model and a measurement error model as well as a comparison of estimated mean square errors.