The Art of budget forecasting

We have not yet reached the end of the first month of the new fiscal year but already Treasurer Scott Meacham has publicly predicted that state General Revenue collections will fall far enough short of the forecast to trigger an official revenue shortfall. We have already shared our recommendations for how the state should respond to a shortfall if one does materialize.  This seems like a good time to provide a closer look at the state’s track record on forecasting and provide some suggestions for what we may be able to do differently.

Each year the Legislature makes appropriations based on an official estimate of revenues for the upcoming year that is certified by the Board of Equalization. For the General Revenue Fund (GRF), the Lottery Fund, and some smaller funds, the Legislature may only appropriate up to 95 percent of the certified estimate; for other funds (including the 1017 Education Reform Fund), the Legislature can appropriate up to 100 percent of the estimate. If revenues come in below 95 percent of the certified estimate, this triggers a revenue shortfall, which requires the State Budget Director to reduce appropriations in a proportionate amount.

The revenue estimates for the major taxes that constitute the bulk of the GRF are developed by the Oklahoma Tax Commission based on economic models developed in collaboration with the Oklahoma State University Center for Applied Economic Research.  OSU and the Tax Commission use a range of public and proprietary forecasts about economic growth, consumer confidence, oil and gas prices, and other variables to generate their estimates.

The results of this process reveal that forecasting is as much art as science. As shown in the chart, in five of the past eight years, actual GR collections have come in more than seven percent above or below the estimate. In only one year was the variance between estimates and actuals less than four percent. The pattern seems to be that in bad times, revenues do far worse than expected (2002-03, 2009-?) while in good time, revenues do far better than expected (2005-06).

Within the overall picture, the accuracy of revenue forecasts varies considerably across the different major taxes.  The record of forecasting gross production tax collections is quite dismal, which may come as no surprise given the volatility of oil and gas prices. In six years out of eight, collections have come in more than 15 percent above or below the estimate, and the average annual variation has been 26.1 percent. At the other extreme, sales tax forecasts tend to be most accurate: in seven of the last eight years, actual sales tax collections have come within five percent of the estimate, and the average variation has been 3.3 percent.   Income tax forecasts occupy a middle ground, with collections coming within five percent of the estimate in three of eight years, with an average variation of 9.2 percent.

We are left asking two questions. First, can we do a better job of forecasting revenues?  There is no obvious or immediate answer, but it would seem worthwhile to study Oklahoma’s forecasting methods and outcomes in comparison to those of other states – particularly those with a similar reliance on gross production revenues –  to see if others are having any greater success.  If some states have better records of accuracy, we should think about borrowing their methods. While we are studying official and binding short-term forecasts, we should also be working to improve our capacity to develop professional five-year budget forecasts to guide longer-term decisions about revenues, expenditures and service levels, as we recently suggested.

Second, assuming that forecasting will always remain imperfect, what should we do about it?  We cannot entirely eliminate uncertainty and imprecision from the budget process, but we could perhaps make the General Revenue Fund less dependent on the most volatile revenue source, the gross production tax. Oklahoma would benefit from a more predictable and consistent flow of funds for its public services. Forecasting is an important step, but strategic use of revenues may be necessary as well.