We offer a model of incumbent party seat change in House elections which provides true forecasts-it "calls" the seat change well in advance of the election itself. The model predicts seat change as a function of both quarterly real GNP per capita growth and presidential popularity measured six months prior to the election, plus the time of the election (midterm or not). This model is more attractive than the earlier, strictly midterm, models propounded by Tufte and Hibbs. It is clearly superior in forecasting the 1982 results. In general, the model's forecasts for postwar House races are very close to the actual result, off only five or six seats at the median, regardless of whether the forecast is made ex post or ex ante. As a final exercise, we apply the model to 1984, in order to forecast the prospective Republican House seat change.