Input from Academia

Dr. George Fulton ⁴ , Research Scientist, University of Michigan, provided a unique perspective to the Dynamic Revenue Estimating Seminar. Dr. Fulton has been involved in forecasting Michigan economic and fiscal activity for over ten years as a member of the Research Seminar in Quantitative Economics (RSQE). He has also worked for a number of years with the Michigan Departments of Commerce, Treasury, and Transportation, and the Jobs Commission, evaluating economic development strategies and conducting policy analysis. In addition, Dr. Fulton is a nationally-recognized expert in using the input/output model of the Michigan economy developed by Regional Economic Models, Inc. (REMI) and has used the REMI model to analyze the economic impacts of auto plant closings in Michigan, Michigan Economic Growth Authority (MEGA) tax incentives, an,d numerous other economic development plans.

At the seminar, Dr. Fulton discussed the current role of the University of Michigan in the state revenue forecasting and policy analysis process; technical details of the REMI model of the Michigan economy, and examples of how the REMI model has been used thus far; the context of the models used for dynamic revenue estimating; and important issues in making the process operational.

The Role of The University of Michigan

Since 1973, the University of Michigan Research Seminar in Quantitative Economics (RSQE) has been under contract with the state to provide and maintain an econometric model of the state -economy. The RSQE model produces quarterly forecasts of U.S. and Michigan economic activity and major state tax revenues. The forecast horizon is two to three years. The RSQE model is available to the House and Senate Fiscal Agencies and the Department of Treasury via modem. All three agencies have the capability to modify the assumptions of the model and use it extensively in the consensus revenue estimating process.

The RSQE hosts a conference on the economic outlook for the U.S. and Michigan each year. In addition, Dr. Fulton and other members of RSQE present forecasts to the Consensus Revenue Estimating Conference twice each year, the Governor's Economic Round Table once each year, and the State Officers' Compensation Commission on a bi-yearly basis.

Regional Economic Models-Inc. (REMI)

Dr. Fulton also discussed a variety of technical and conceptual issues related to the REMI model. Conceptually, REMI is similar to a standard input/output ⁵ model because it incorporates buying and selling transactions among industries at a detailed level. The REMI model goes beyond standard input/output models, however, because it also traces the implications of economic actions over time. The model incorporates:

• interindustry transactions and endogenous ⁶ final demand feedback effects,

• substitution among factors of production in response to changes in relative factor costs,

• migration that occurs in response to changes in labor market conditions, and

• changes in state and local shares of the export market due to changes in regional production costs and profitability.

The REMI model is composed of output and industry demand equations for 49 private, non-farm industries; three government sectors; the farm sector; and consumption equations by industry based on real disposable income received by consumers. Real disposable income equations describe:

• wage and salary income,
• other labor income,
• property income,
• personal contributions to social insurance, and
• transfer payments.

The model also includes three types of investment equations: residential investment, nonresidential investment, and equipment investment. Government spending equations predict government spending for six components: federal civilian, military, and state and local expenditures for education, health and welfare, public safety, and miscellaneous ⁷.

Examples of how the state REMI model has already been used include: economic impact of the K.I. Sawyer military base closing, the economic/revenue impacts of announced General Motors plant closings, impacts of higher oil prices on Michigan, and the impacts of the Mitsubishi plant location decision, among others.

Issues in Making the Process Operational

Dr. Fulton raised a number of important points about making the proposed process (of generating dynamic estimates of tax policy changes) operational. Incorporating more sophisticated modeling into the process may necessitate changing the procedure for making requests for bill analyses to allow a longer lead time to perform an analysis.

Any dynamic model will require maintenance and testing on an ongoing basis. The technical work will require a great deal of sophistication and a thorough understanding of the issues. Important questions for consideration include:

• Who should do the technical work?

• Should the work be done in-house, or should some arrangement be made with academia or the private sector?

• If the work is done in-house, is the permanency of staff an issue?

• If the work were done to some extent in an academic setting, would that environment be compatible with the requirements of the process?

Dr. Fulton also stressed the importance of the "credibility of message and the messenger" to the whole process.

Private Sector Experience

Michael Vlaisavljevich ⁸, Managing Director of the Barents Group, KPMG Peat Marwick, gave a presentation titled, Dynamic Revenue Estimating: Meeting The Challenge. Mr. Vlaisavljevich noted that both economic theory and research indicate that taxes influence behavior through a variety of complex interactions that typically include successive rounds of impacts. Although the long-term magnitude and direction of these "feedback effects" are debatable, he suggested that these effects should be incorporated, to the extent technically feasible, into fiscal impact and revenue estimates.

The Dynamic Revenue Estimating Process:
Mr. Vlaisavljevich used the diagram shown in Figure 1 to explain the dynamic revenue estimating process. The process involves the following major steps:
1. Baseline Economic Forecast:
   The first step in the process is to prepare a forecast of economic activity. This "baseline" forecast assumes that no changes in tax policy occur for the forecast period.

2. Direct (Static) Revenue estimates of the Proposed Tax Change:
   Using a micro-simulation model, or some other method, the direct or static revenue impact of a tax change is estimated. The static estimate is predicated on the baseline economic forecast.

3. Dynamic Economic Impact Model:
   The static revenue impact is then fed into the dynamic economic impact model to estimate the effects the tax policy change will have on economic activity. This model estimates the change in economic activity by estimating the impact the tax policy change will have on key factors such as taxpayer behavior, employment, and business investment.

4. Revised Revenue Estimates:
   The revised economic forecast that incorporates the changes in economic activity predicted by the dynamic model is then used to estimate the dynamic revenue impact of the proposed tax change. This new estimate is called a dynamic estimate because it reflects both the direct (static) effect and the secondary (dynamic) effect of the tax policy change.

Managing the Process:
Mr. Vlaisavljevich also discussed how the process of incorporating dynamic analysis into the current methodology might best be managed. Key to the process is the credibility of the methodology and the assumptions incorporated into the model(s) that would be required to perform dynamic analysis. He also suggested that this type of analysis should be reserved for major tax policy changes.

Important issues with respect to managing the process were discussed as well. State policy makers are often accustomed to receiving analyses almost instantaneously. Since instant analyses that incorporate dynamic effects are not feasible due to model set-up and review requirements, the legislative process may be slowed down if dynamic analyses are incorporated. As with any new technology, it is important to guard against unrealistic client (i.e., policy-maker) expectations and maintain credibility. Mr. Vlaisavljevich suggested that there may be a need to obtain outside assessments of the models/technology to verify that models are state-of-the-art, and are being used by competent professionals. It will also 'be important to ensure that model applications and objectives are credible.

The California Experience

Dr. Bruce Smith ⁹, California Department of Finance, presented an overview of Computable General Equilibrium (CGE) models and discussed the California experience. In August 1994, the State of California adopted legislation that requires the California Department of Finance to incorporate the effects of the dynamic responses of economic agents when evaluating the fiscal effects of changes in tax policy. In September 1995, Dr. Smith was hired to build, maintain, and operate the California CGE model.

Computable General Equilibrium (CGE) Models:
In his first presentation, Computable General Equilibrium Models, Dr. Smith discussed what CGE models are, what the good and bad aspects of CGE models are, and the process California used to build a CGE model of the California economy. Computable General Equilibrium models are the most recent contribution to the field of applied economic modeling. While the basic economic theory underlying CGE models has been part of the professional literature for some time, implementation of the theory required the advent of sophisticated computer software and technology.

In its most basic form, a CGE model is a description of the relationship between and among producers, households, government, and the rest of the world. In order to capture the "dynamic effects" of tax-policy changes on an economy, the model must do a number of things simultaneously. It must track the income of individuals and firms (because this is the basis of income taxation) and it must track the sale of the goods and services that serve as the basis of excise taxes. In order to be dynamic, it must also track the effects of taxation on the economy's use of labor and capital, as well as other economic reactions to state fiscal policy.

California's Dynamic Model:
No model can account for every interaction in the economy; hence, the model must aggregate sectors of the economy. The California model divides the economy into 75 sectors: 28 industrial sectors, two factor sectors (capital and labor), seven household sectors (by household income), one investment sector, 36 government sectors (seven federal, 21 state, and eight local), and one sector that represents the rest of the world.

The biggest problem with modeling various sectors of the economy is a lack of accurate data. For example, in order to model the California economy, data for the industrial sectors had to be estimated using national industrial data collected by the Bureau of Economic Analysis in 1987. The data were "scaled-down" to the state level and "adjusted" to approximate economic conditions in California in 1995-96.

It is also important to note that although California has a working Computable General Equilibrium model and has been testing the model by simulating certain tax policy changes, the California Department of Finance has not yet presented any estimates of the dynamic effects of actual legislation to the California Legislature. The reaction of policy makers and the way these estimates stand up to further scrutiny remain to be seen.

The Minnesota Experience

Dr. Robert Cline ¹⁰, Director of Tax Policy Research, Minnesota Department of Revenue, gave a presentation titled, Minnesota's Experience with Dynamic Revenue Estimating. Whereas the California Legislature mandated that a dynamic model be developed and used to estimate the dynamic effects of all substantial tax policy proposals, the Minnesota Legislature asked the . Office of Tax Policy Research to analyze the dynamic effects of only one major proposal: exempting capital purchases from the state sales tax. Minnesota's experience is important in that the results of the analysis, the model, and the process were all deemed credible by both state policy makers and industry representatives.

The process started in 1993 when the Capital Equipment Advisory Council (consisting of 14 members appointed by the Governor, the Senate, and the House) was created and charged with evaluating the proposal. The Tax Research Division was assigned the responsibility of doing the actual dynamic estimates. The Council also established a technical group consisting of private sector tax specialists, economists, and state agency staff. This group met frequently for a period of five months to discuss methodology, data, and model results. Hence, although specialists within state government developed the model and generated the estimates, industry representatives and specialists in the private sector had the opportunity to provide input and monitor the process. In Dr. Cline's view, the openness of the process mitigated potential controversy over the final estimates produced by the model.

Minnesota's Dynamic Model:
The dynamic model used in the Minnesota analysis, see figure 2, incorporated an economic forecasting model developed by Regional Economic Models, Inc. (REMI). The REMI model is an input/output model composed of aggregate demand equations for consumption, investment, government spending, and exports from Minnesota that determine total spending in the Minnesota economy. A set of equations determines the shares of Minnesota and U.S. purchases supplied by Minnesota firms, and models how state/U.S. shares respond to relative business costs -- including the cost of capital. The REMI model was used in conjunction with a cadre of micro-simulation models developed by the Tax Research Division.

Model Assumptions:
Dr. Cline's discussion demonstrated that, in analyses of this type, a number of important (and sometimes ad hoc) assumptions are unavoidable because no model is capable of capturing all the interactions that occur over time when tax policy changes. For example: What should be assumed about the behavior of other states?

Tax incidence analysis ¹¹ of the relative effect of a state tax rate demonstrates that taxes on capital are germane. We also know that states actively compete and respond to changes in competitor states. The Minnesota and California simulations assumed no changes in the tax rates of other states. However (at least in the long run, and perhaps in the not-so-long run), other states are likely to respond in kind. Hence, the competitive advantage, which causes at least some of the feedback effects, will be eroded over time.

Scott Jordan ¹², Deputy Director, Office of Tax Policy Analysis, Massachusetts, discussed Massachusetts' experience with dynamic estimates. Massachusetts developed a dynamic model in 1992. Model development was overseen by an advisory panel consisting of legislative leaders, academic and business economists, and a private consulting firm. The model was built by Price Waterhouse. The advisory panel provided input and oversight on the project.

The model produced much lower dynamic feedback effects than were expected by some state officials. Because of this, the initial results were heavily criticized, mistrusted, and largely ignored by the Legislature. The model is still being used by staff to assess the impact of tax policy changes, but on a more limited basis than originally envisioned. To date, three analyses have been completed and presented to the Legislature. The Office of Tax Policy Analysis has recently hired an economist with CGE experience, and is in the process of developing a new CGE model of the state economy.

⁵ This type of model incorporates the fact that the goods/services produced by one industry (output) may be an input to the production process in another industry. Therefore, policies that initially affect one industry may have economy-wide effects.

⁶ The way policies that initially affect one industry also impact the rest of the economy.

⁷ For an extensive description of the REMI model see; Treyz George I., Dan S. Rickman and Gang Shao; The REMI Economic-Demographic Forecasting and Simulation Model; International Regional Science Review, Vol 14, No. 3, pp. 221-253, 1992.

⁸ For the past 9 years, Mr. Vlaisavljevich has directed consulting projects in over 20 states including business tax competitiveness model projects for North Carolina and Kentucky; tax policy simulation models projects for Rhode Island, Arizona, Kansas, Iowa, Kentucky, Minnesota, and Pennsylvania; and a multi-year budgeting project for North Carolina. He has also managed projects to produce a forecasting model for Guam; and economic impact analyses for the Department of Defense, Intel, and Argonne National Laboratory. Prior to becoming a consultant, Mr. Vlaisavljevich was Tax Policy Director for the Wisconsin Department of Revenue for eight years and Senior Fiscal Analyst with the Wisconsin Legislative Fiscal Bureau for five years.

⁹ Dr. Smith received a Ph.D. in Economics in 1993, from the University of Texas at Austin. His dissertation topic was Optimal Regional Economic Policies from Computable General Equilibrium Models. He taught briefly at San Antonio College. On September 1, 1995, Dr. Smith began working for the California Department of Finance, designing and building a CGE model of the California Economy to be used for tax policy analysis.

¹⁰ Dr. Cline received a Ph.D. in Economics from the University of Michigan. He has served as the Director of Tax Policy Research, Minnesota Department of Revenue, since 1989, and is currently managing the Pennsylvania Tax Blueprint Project for Price Waterhouse in Washington, D.C. Dr. Cline was Director of the Office of Revenue and Tax Analysis, Michigan Department of Management and Budget, in 1984-86; has held research positions at the U.S. Advisory Commission on Intergovernmental Relations and the Urban Institute; and has taught at the University of Michigan, Georgia State University, the University of Minnesota, Hope College, and Manchester College.

¹¹ The analysis of which individuals/firms bear the ultimate burden of taxes.

¹² Mr. Jordan earned a bachelor's degree in economics in 1989 from the University of Massachusetts at Amherst. He served as a staff research analyst for three years, was named manager of the Tax Policy Analysis Unit in 1993, and Deputy Director of the Office of Tax Policy Analysis in 1995.