Seasonal heteroskedasticity refers to seasonal changes in variability in a time series occurring over calendar years. When present in economic indicators, it can affect seasonal adjustments and trend estimates used for understanding historical patterns in the data, analysis of current trends, and policy making. In this paper, we investigate how seasonal heteroskedasticity affects signal extraction results in the forms of trend estimates and seasonal adjustments from some standard seasonal time series models enhanced to include a seasonally heteroskedastic irregular component. We apply these models to a regional time series of U.S. housing starts that shows higher levels of variability in the winter months. Comparing signal extraction results from the original and enhanced forms of the models shows the importance of accounting for seasonal heteroskedasticity in these time series to both signal extraction estimates and their error variances.