How to Forecast the Fiscal Impacts of Tax Incentive Programs

In this column, originally published by The Pew Charitable Trusts in September 2024, Jim Landers, professional practice associate professor, Enarson Fellow, director of graduate/professional studies at the John Glenn College of Public Affairs at The Ohio State University, discusses the complexities and challenges of estimating the fiscal impacts of tax incentive programs.

Evaluation perspectives

Forecasting the Fiscal Impacts of Tax Incentive Programs: Unraveling the Revenue Puzzle

Jim Landers
Professional Practice Associate Professor, Enarson Fellow, Director of Graduate/Professional Studies
John Glenn College of Public Affairs
The Ohio State University

State legislative fiscal offices fulfill a variety of responsibilities, the most important of which may be preparing estimates of the potential revenue or spending impacts of proposed legislation. I recently spoke to a meeting of state legislative fiscal office directors at the NCSL Legislative Summit about the complexities and challenges of estimating the fiscal impacts of tax incentive programs. This is a different direction from our typical discussions about incentive evaluations, but it's an important topic that we will explore in this edition of the newsletter.

Why are fiscal estimates of tax incentive programs important?

Estimating the cost of tax incentives has important implications for legislative decision-making, state budgets, revenue forecasts, and even incentive evaluations. These estimates inform legislative decisions on proposed incentive programs, or on expansions of existing incentive programs. Likewise, these fiscal impact estimates may gain importance for decision-making on state budgets if the expected revenue loss is sufficiently large to have discernible impacts on revenue growth.

Revenue loss estimates for extremely costly incentive programs may have significant implications for revenue forecasts. Such incentives may lead to larger forecast errors and probably require forecast adjustments during the initial years of program implementation when their impact is not discernible in the data series used for forecasting. The revenue loss estimates also establish a baseline for estimating changes in incentive utilization and revenue loss because of future legislative changes.

Retrospective reports of revenue loss from incentives, coupled with estimates of future revenue loss, could also be important informational components of incentive evaluations. This data informs policymakers about the scale of an incentive program that they are evaluating and may revise or eliminate. The incentive amounts that taxpayers claim—alongside income and tax liabilities reported on their tax returns—can be used to examine the income distribution of the incentive, as well as the relative impact of the incentive on the recipient’s tax liability and income. The latter could be used as an indicator of whether the incentive is plausibly affecting recipients’ spending, employment, or investment decisions.

Incentive claims (revenue loss) data, and data on incentives awarded by the agency administering the incentive, also provide insight about the contingent revenue loss from incentive awards being carried over to succeeding years. These contingent liabilities could have important implications for budget decisions and revenue estimates, especially for large-scale incentive programs.

The existing terrain

While countless tax incentive programs are in place across the U.S., new programs continue to spring up, requiring careful analysis of their potential budgetary and revenue impacts. The Council for Community and Economic Research (C2ER) reports that states offer almost 1,000 tax incentive programs. More than two-thirds of these incentives are tax credits that offer the recipients a dollar-for-dollar reduction in their tax bills.

While states authorize many tax incentives that are relatively minor in scope and impact, many are extensively used and lead to significant forgone revenue. For instance, Ohio’s 2024-25 tax expenditure report estimates that the revenue loss from credits against the state’s Commercial Activity Tax for research and development, job creation, and job retention alone could be as high as $200 million annually. Likewise, Indiana’s 2022 tax expenditure report estimates that the state’s Economic Development for a Growing Economy (EDGE) job creation and retention tax credit and research and development tax credit could result in $120 million to $130 million in revenue loss from the state’s individual and corporate income taxes.

C2ER also identifies several new incentive programs authorized by states since 2023. These programs include incentives for venture capital, agricultural businesses, high-tech companies, and semiconductor manufacturing. In addition to specific incentives, states have recently been enacting individual and corporate income tax cuts to spur investment and employment. The Tax Foundation reports that 25 states reduced individual income tax rates and 13 states reduced corporate income tax rates between 2021 and 2023.

The cumulative impact of these incentive programs and tax rate cuts means reduced revenue available for state budgets as well as tax burden shifts, necessitating careful review and potential recalibration of these policies.

Costs and benefits

Estimating the fiscal impact of proposed incentive programs could include program costs as well as benefits. However, the cost side the direct budgetary impact will probably be the focus. Estimates of program costs should include both direct and indirect costs. The direct costs are obvious (revenue loss from tax incentives and direct spending on grants and other subsidy programs) but can be difficult to estimate. This is due to lack of pertinent data and uncertainties about take-up rates by firms and the scale of qualified investment or employment qualifying for incentive dollars. While capping annual incentive awards is foremost a means of controlling cost, it also simplifies the estimation process.

The indirect cost should include spending related to program administration or higher demand for public services resulting from incentivized business activity. Large company relocations receiving incentives may lead to significant (potentially localized) spending increases on roads, infrastructure, public safety, and schools. These could be directly related to the company’s facilities or to the in-migration of people employed because of increased economic activity. These indirect costs, while significant, could potentially be overlooked when costing out the impact of these programs and are challenging to predict for many of the same reasons outlined above.

Precision versus scale

How we interpret fiscal impact estimates for proposed incentives is also important. Reasonably speaking, while we may report these estimates in very exact terms, they are not a precise reflection of the potential revenue impact of the tax incentive. Rather, the estimates should be viewed as an indicator of the scale of the potential revenue impact. They shouldn’t be used as an exact indicator of whether a tax incentive could lead to $5 million or $5.1 million in forgone revenue, but whether the revenue loss could be $5 million or $50 million. Besides underscoring the care that analysts should exercise when using these types of estimates, it also highlights how caps on incentive awards can successfully control the cost of these programs.

Methodology

Analysts should select a methodology that leads to valid and defensible revenue impact estimates. Assumptions, data, methodology, and other information used to generate the estimate should be reasonable, as there may be a host of individuals and groups who will have different opinions about the fiscal impact of the incentive program.

Static estimation is the most basic approach to estimating the revenue impact of an incentive. It assumes that the behavior of the incentive recipient will not change. Static estimates essentially ignore microeconomic theory that people and businesses respond and change their decisions and behavior when they receive an incentive such as a tax credit or a grant. Consequently, static estimates should be avoided as they lack reasonableness and could be difficult to defend.

However, analysts often must provide static estimates when sufficient data or research is unavailable to support adjustment for potential behavioral changes by incentive recipients.

In contrast, microdynamic estimation assumes that the recipients of tax incentives change their behavior. That is, we assume that incentive recipients respond to the incentive by increasing their economic activity (e.g., spending, employment, investment) beyond levels that would have occurred in the absence of the incentive. The additional economic activity generates additional revenue that offsets some of the direct revenue loss of the incentive program. While microdynamic estimates add complexity (and maybe more uncertainty) to the estimation process, they tend to be accepted as more plausible than wholly static estimates. Still, microdynamic estimates require analysts to make reasonable assumptions about the response of individuals or businesses to tax incentives and, if available, use empirical estimates of these response rates to guide their assumptions.

Macrodynamic estimation takes things a step further—a big step further. Macrodynamic estimation accounts for the incentive’s static and microdynamic revenue impacts plus the effects of the incentive on the overall economy and full tax system. The macrodynamic effects would occur because of spending by new employees of the businesses receiving tax incentives and increased spending by these businesses throughout the supply chain, which then generates additional spending by employees of those suppliers.

Still, additional information gained by dynamic estimation may come at steep cost. The economic data and modeling required to generate macrodynamic estimates (e.g., computable general equilibrium models or regional economic models like REMI or IMPLAN) are expensive, complex, and time-intensive to deploy. They also require many more estimates of or assumptions about household behavior, interindustry relationships, and the response of businesses to the spending, employment, and investment changes of other businesses receiving tax incentives. Mikesell (2012) reports that different macrodynamic models can generate estimates that vary significantly and that may be highly sensitive to changes in analysts’ underlying economic assumptions. Moreover, Mikesell reports that state macrodynamic estimates suggest that the secondary effects of tax changes or incentives are relatively modest.

Give your estimates a smell test

No matter the analytical approach selected, analysts should assess the robustness of their methodology and plausibility of their baseline fiscal impact results. This can be done by varying economic assumptions; estimating variables, sample data, estimator values; or generating estimates using an alternate methodology.

When employing an econometric model to estimate, for instance, the revenue impact of a policy change, analysts can perform various sensitivity tests to examine the variation in model outputs or fiscal impact estimates. These include adding explanatory variables to the econometric model, interacting other explanatory variables, changing the composition of the estimating sample, inputting extreme values for the explanatory variables in the econometric model to generate impact estimates, or inputting different values of the estimators produced by the econometric model (e.g., using the confidence intervals of the estimators instead of the mean) to generate impact estimates. Frequently, however, analysts aren’t estimating elegant econometric models but are instead cobbling together impact estimates using secondary data sources, estimates from academic studies, and their own educated assumptions about behavioral responses to tax incentives or tax rate or tax base changes.

For instance, I can recall estimating the revenue impact of a proposed reduction in Indiana’s insurance premium tax. The tax is imposed on the total premiums on insurance policies sold by a company in the state. I generated a static estimate of revenue loss from the proposed rate cut using several years of company-level tax return data. The reason I went with the static estimate was due to the structure and imposition of the tax, the insurance market in the state, and trends in insurance premiums and insurance company employment in Indiana over 15 years, including two previous reductions in the tax rate. From that analysis I assumed that there was a low likelihood of insurance companies relocating to Indiana because of the rate cut. Thus, I assumed no significant increase in insurance premiums written in Indiana (the tax base) or employment gains generating additional income tax revenue for the state. Thus, fiscal impact modeling using the tax return data was informed by significant data analysis of the insurance industry in Indiana.

Conclusion

Estimating the cost of tax incentives is challenging work, but the results have important implications since they inform budget decision-making, decisions on tax policy, and management of the financial resources of state government. Consequently, it is imperative that budget analysts, fiscal researchers, and incentive evaluators continually reflect on the validity of their data and methods and the plausibility of their fiscal and economic impact estimates.