This paper analyses and forecasts annual time series of aggregate real income per head in the US. The approach integrates elements from recent univariate time series analyses with multi-equation macromodels in which policy feedback rules have been endogenized. The main conclusions are as follows: First, aggregate real per capita income is subject to significant trend reversion. This conclusion comes through more clearly by examining the data at an annual rather than the more usual quarterly frequency and by incorporating multivariate economic contents in the income process. Secondly, there is significant evidence for the Lucas (1976) or Haavelmo (1944) critique: in the US there appears to have been a shift in the structural macropolicy reaction function causing a corresponding shift in the reduced form income forecasting equation. This is associated with increased concern in the late 1980s over the size of US budget deficits. Thirdly, with the above proviso, useful real income forecasts can be made as far as three years ahead. Finally, the paper provides empirical evidence for the effectiveness of monetary policy on real output or income. The change in the short-term interest rate is highly significant in forecasting income growth up to three years after the change.