Forecasting the federal budget with time-series models


  • Hamid Baghestani,

    Corresponding author
    1. The Economics Institute, CO, U.S.A.
    • The Economics Institute, Boulder, CO 80309, U.S.A
    Search for more papers by this author
    • Received his PhD in 1982 from the University of Colorado. Currently, he is teaching at the Economics Institute in Boulder, Colorado. He has authored articles some of which have appeared in Applied Economics, The Journal of Business, The Journal of Macroeconomics. and the Southern Economic Journal.

  • Robert McNown

    Corresponding author
    1. University of Colorado, U.S.A.
    • Department of Economics, Campus Box 256, University of Colorado, Boulder, CO 80309, U.S.A
    Search for more papers by this author
    • 1971 PhD in Economics from the University of California, San Diego. He is now Professor of Economics at the University of Colorado, Boulder. His research on topics in time-series analysis and forecasting have been published in The American Economic Review, Applied Economics, Demography, and The Journal of Time Series Analysis.


The stochastic properties of conventionally denned federal expenditures and revenues are examined, and cointegration is found. Alternative time-series models-univariate ARIMA models, vector autoregressions in levels and differences, and an error correction model-are specified and estimated using quarterly data from 1955:1 through 1979:4. Updated forecasts for up to three years beyond the sample period are evaluated against actual expenditures, revenues and the deficit. The vector autoregression in levels shows evidence of nonstationarity, which leads to strong biases in the forecasts. The remaining models produce forecasts that are satisfactory by the mean squared error criterion, and the magnitudes of biases at the longer horizons are significantly smaller than those of the official forecasts.