SD25 - Revenue Forecasting in the Executive Branch: Process and Models

  • Published: 1991
  • Author: Joint Legislative Audit and Review Commission
  • Enabling Authority: Appropriation Act - Item 13 (Regular Session, 1990)

Executive Summary:
Virginia experienced an unusually large revenue shortfall for FY90, with additional and substantial shortfalls expected for FY91 and FY92. The shortfalls have raised questions concerning executive revenue forecasting - Virginia's forecast accuracy, the forecasting process itself, the forecast models used and their administration, and the effects of tax policy changes and judgmental inputs. This interim report, the first in a series on the executive budget process, addresses these questions.

Forecasting is not an exact science there will always be uncertainty associated with forecasting. Differences between revenue forecasts and actual collections are to be expected. Historically, Virginia's forecast accuracy has been similar to that of other states, the federal government, and national economic forecasting firms. However, revenue forecasts for FY91 and FY92 have been reduced to an unusually large degree: the current shortfall now approaches $1.7 billion for the biennium.

The magnitude of the recent forecast reductions does not appear to be the result of an unsound forecasting process. While improvements could be made to increase accountability for the forecast, the process fully or partially meets the majority of criteria for an optimal forecasting process. Nor were adjustments to the forecast necessitated by problems with the statistical forecasting models. These models, and their administration, appear to be generally sound.

The majority of the shortfall in the General Fund appears to be due to an economic downturn. JLARC staff confirmed this with a technical replication of the statistical models used in the forecasting process. The remainder of the shortfall in the General Fund could be the result of tax policy changes and judgmental inputs, as well as normally-occurring forecast error. The lack of conclusive research on the fiscal impact of tax policy changes and the lack of documentation of judgmental inputs to the forecast make it difficult to independently substantiate the impact of these factors. Forecast error, on the other hand, is an inevitable part of forecasting.

If forecast error is inevitable, how can the legislature better anticipate changing revenue conditions that affect budgetary decisions? One step the General Assembly could consider is an increased role in the forecasting process. Relative to other state legislatures, the Virginia General Assembly plays a minimal role in the forecast process. Increased participation, however, would mean a commitment of more legislative time and money. The feasibility of such a commitment will need to be evaluated against the degree to which the desired outcomes - a more accurate revenue forecast and increased accountability - can be achieved.

In addition, the General Assembly may wish to protect the State's budget in case of unusually large forecast error by establishing a Revenue Stabilization Fund or "rainy day" fund. A proposal for such a fund is described in the JLARC special report, "Proposal for a Revenue Stabilization Fund in Virginia."

Historically, Virginia's Forecast Accuracy Is Similar to That of Other States

The General Assembly has been told that Virginia's forecast accuracy over the past 16 years is two percent; in other words, forecasts differ from actual revenue collections by an average of two percent. However, the two percent figure is based typically on forecasts made only five months from the end of the fiscal year, a standard for forecast accuracy used by the executive branch for many years.

However, the forecasts the legislature actually uses for appropriations are one-and-
a-half to two-and-a-half years from the end of a fiscal year. The accuracy of these longer-term forecasts in predicting collections averages from three to four percent. On average over the past 16 years, Virginia's forecast accuracy of three to four percent is similar to that of other states, national economic forecasting firms, and the federal government.

Recommendation. The one-and-a-half to two-and-a-half year forecasts are those used by the General Assembly to make its budgetary decisions. Consequently, the Department of Taxation should monitor and report differences between forecasts and actual collections for all forecasts which the General Assembly may use in making appropriations.

Forecasts for FY90, 91, and 92 Were Unusually Far Off

The FY90 forecast used by the 1989 General Assembly was 10.3 percent higher than actual collections, much higher than the three to four percent average over the last 16 years. The forecasts used by the General Assembly in its 1990 session to budget for the 1990-92 biennium are now expected to be even further off. For FY91, the difference between the original forecast and the most recent revision is 13.3 percent, while for FY92 there is a 16.8 percent difference. For the 1990-92 biennium, the shortfall has been estimated at $1.7 billion.

Forecast Models and Their Administration Are Basically Sound

The unusually large downturns in the FY91 and FY92 revenue forecasts do not appear to be attributable to either unsound models or inadequate administration of the models. The model assumptions appear to be clearly understood by Department of Taxation staff, and the economic variables (personal income, gross national product, and unemployment rates) used in the models appear to be sufficient. In addition, the mathematics underlying the models appear sound. While accounting for regional conditions seems problematic, it is unclear whether the added complexity of formally modeling regions of the State separately would improve the accuracy of the revenue forecast models.

Most of the Downward Revisions to the Revenue Forecast Can Be Attributed to Declining Economic Indicators

Sixty percent of the forecast reduction in the major tax sources of the General Fund for FY91 and FY92 (approximately $926 million) can be attributed to declines in economic indicators. In general, as economic indicators used in the models have declined, the model forecasts have declined. JLARC staff used the Department of Taxation's models of major tax sources and the corresponding economic data used in each model to reach this conclusion.

Some of the Shortfall Cannot Be Independently Substantiated

Over 40 percent of the decline in the major tax sources of the General Fund cannot be attributed to changes in economic indicators driving down the model results. This amount (approximately $622 million) may be due to the effects of tax policy changes and judgmental inputs. JLARC staff were unable to substantiate independently the estimated impacts of tax policy changes because sufficient data are not yet available to verify the estimates. Also, while the forecast models are well documented, judgmental adjustments used to amend forecasts are not.

Recommendation. The Department of Taxation should take steps to enhance its research and monitoring capability to predict better the impact of tax policy changes. Specifically, the Department should:

• Conduct more research to verify the estimated fiscal impacts of specific federal and State tax reforms.

• Collect itemized statistics of income.

• Develop additional sampling strategies for analyzing income tax returns.

• Separate out the components of non-withholding payments on monthly revenue reports.

The Department should also document changes to its forecasts based on judgment.

Recommendation. The Department of Taxation should ensure that sufficient documentation of each revenue forecast exists so that the Department of Taxation staff, or any other group reviewing the forecasts, can replicate each forecast. In the event judgmental adjustments are made to a forecast, they should be identified. A record should be made of the size of the adjustments and of the forecast errors with and without the adjustments.

Legislative Involvement in the Forecast Could Be Increased

Greater legislative involvement in the forecasting process could theoretically improve forecast accuracy and would improve accountability for the forecast. There are a number of ways to increase legislative input, depending on the level of involvement desired by the General Assembly. No one option for increased legislative involvement is clearly preferable to others. Any option increasing legislative participation would cost Virginia's citizen legislature time and money, however. Therefore, the various options for increasing legislative involvement in the revenue forecasting process should be studied further. In the meantime, JLARC staff can continue to monitor the forecast process and models until the General Assembly determines what, if any, expanded role it wants in the forecasting process.

The General Assembly may also wish to formalize a collection-monitoring process and designate a legislative staff agency to execute it. Currently, the General Assembly relies primarily on the executive branch to analyze and interpret revenue collection data. Formalizing such a collection-monitoring function in the legislative branch could provide the General Assembly an alternative early warning system to identify potential shortfall years. For instance, an analysis of individual income tax collections by JLARC staff indicates that FY90 experienced slow collections beginning in the first quarter, and continuing throughout the year. Through its first five months, FY91 is following a similar pattern, indicating the downward revisions in the forecast made thus far are justified.

Recommendation. The JLARC Subcommittee on the Executive Budget Process may wish to consider options for increasing the level of legislative involvement in the revenue forecasting process and report to the full Commission by December 1991. Moreover, the General Assembly may also wish to formalize the collection monitoring function in the legislative branch in order to allow an independent analysis of collection trends.

Improvements in the Forecasting Process Could Increase Accountability

In addition to increased legislative involvement, other improvements in the forecasting process would enhance accountability. First, the executive branch should submit revenue-related reports on a regular basis, as required by the General Assembly. These reports have not always been submitted as required. In addition, the General Assembly may wish to clearly specify that an interim forecast, when used by a Governor to reduce appropriations or withhold allotments, should follow the same process as specified in the Code of Virginia for December 15 forecast estimates.

Recommendation. The Secretary of Finance should monitor the submission of the "Monthly Report to the Governor" and the "Monthly Revenue Report" to ensure full compliance with Section 4-8.01(f) of the 1990 Appropriation Act.

Recommendation. The General Assembly may wish to specify in Part 4 of the Appropriation Act or by revision of statute that any interim forecast upon which the Governor reduces appropriations or withholds allotments, and which is not submitted during a legislative session, should follow the forecast process as set out in Section 2.1-393 of the Code of Virginia.