The U.S. Census Bureau’s Small Area Income and Poverty Estimates (SAIPE) program provides annual estimates of poverty within school districts, counties and states for different age groups. The SAIPE program’s estimates are model-based, and use single year American Community Survey (ACS) data and administrative records. There is potential for improving estimates of the current year’s parameters by using previous years’ data in the small area model. Two methods for utilizing multiple years of data are compared: the first method uses a bivariate normal distribution for the model errors from different time periods, and the second method assumes an AR1 structure on model parameters. Gains in efficiency using multiple years of survey data for estimation of a small area parameter are investigated using each method. In addition, estimates of the increase or decrease over time of a small area parameter are constructed, as well as credible intervals for the change over time. An example using state-level SAIPE data is presented.