Unaided observation suggests that political leaders manipu1ate forecasts. Lance Brouthers study tests this supposition in a systematic way by looking at the errors in election years and comparing them to those in non-election years. Forecasts of revenues and expenditures (and deficits) were examined for a 35 year period from 1950 through 1984 in the United States. The forecasts studied were those presented to congress by the Office of Management and Budget (OMB) in January of each year. As expected, the errors in off-year congressional election years were much higher, especially with respect to receipts. The error in the federal deficit was more than twice as high (36.4% vs. 15.7%). Interestingly, the errors were no higher for presidential election years.
Differences were found also in the forecast errors depending upon who controls the White House: When Republicans were in office, the OMB forecasts tended to be more accurate than when Democratic presidents where in office. Democratic administrations overestimated the deficit and Republican ones underestimated it.
The data presented by Brouthers yield some additional observations. First, as shown in the table that I constructed from these data, forecasting does not seem to be getting better over time.
It is also interesting that, contrary to common belief, forecasting was more accurate in the so-called turbulent times of the 1970s. If one takes the 1970s after the oil crisis (using 1973 to 1979), the mean absolute error for OMB forecasts was 4.2%, a result that compares favorably with the average error of 6.2% for the other years during the 1950-1984 period. This result is in agreement with prior empirical research [Daub (1981), Daub and Peterson (1981) and Makridakis (1981)].