Why Government Budget Estimates Are Often Wrong

This column first appeared on TaxVox.

Government budget and tax analysts who estimate future federal revenues and spending are among the most talented people I know. They probably are a lot more accurate at what they do than typical academics or business consultants. However, their estimates frequently understate the true long-term costs of tax cuts or spending.

When estimators miss on the low side, it is often because they are trying to project costs of those government programs and tax subsidies that are both permanent or “mandated,” absent new legislation, and essentially open-ended.

Unlike programs where the government appropriates a fixed amount of money each year, the costs of mandated programs don’t need to be appropriated to be spent. Meanwhile, open-ended programs leave important determinants of cost, such as demand or price or definitions about who or what qualifies for the program, to decisions made by beneficiaries or service providers. When the two elements are combined, Congress effectively cedes long-term control of the costs to private individuals acting in their own, not necessarily the public’s, interest.

Estimation is simpler for mandatory programs or permanent tax benefits that are not open-ended. For example, estimators know that the cost of the child tax credit will equal the amount of the credit times the number of eligible children. Similarly, Social Security retirement benefits might grow rapidly but can be estimated with fair accuracy because they are set by formulas based on lifetime earnings that provide only limited discretion to recipients.

That is not so for Medicare, where beneficiaries and providers have often been allowed to appropriate resources to themselves. Consumers demand more access to healthcare treatment even as providers—who mostly are compensated based on volume—happily increase the supply of those treatments. Because there is little effective market discipline, health care provided by government creates a perfect fiscal storm. For example, suppose a drug company markets a new drug under government patent protection; it then sets a price; consumers demand the drug to address the ailment it treats; and—often—Medicare pays.

A similar phenomenon occurs with the open-ended tax subsidy for capital income, which is taxed at lower rates than ordinary income. Since ordinary income is taxed at a top rate of 37 percent while long-term capital gains are taxed at a top rate of 23.8 percent, taxpayers have an enormous incentive to recharacterize their income to benefit from the lower rate. The classic recent example: Hedge funds that have converted a share of their managers’ labor compensation income into lower-taxed long term capital gains income (carried interest).

Over the years, Congress and the IRS have played a game of whack-a-shelter with respect to preferential tax rates for capital income. Smart lawyers find a new way to turn ordinary income into lower-taxed capital gains, government (through either legislation or regulation) shuts it down, and then taxpayers and their advisors find another approach. That process makes it impossible to estimate government revenues for the long-run since estimators are supposed to assume the permanence of what is an inherently unstable law.

Forecasting capital gains revenue is even more difficult because investors can choose when to realize gains and, thus, pay the tax. As a result, gains can be earned over decades but are not taxed (and this generate no revenue) unless “realized” through an asset sale.

In a classic article, Nobel prize winning economist Joseph Stiglitz explained how individuals could take advantage of these arbitrage opportunities to reduce the taxes they pay. Given their voluntary nature, a large share of gains is never taxed because they are held until death—when their assumed cost in the hands of the heir is “stepped-up” to the market price at the time the person making the bequest passes away.

The newest example of an open-ended tax shelter is the Tax Cuts and Jobs Act’s 20 percent individual income tax deduction for income from pass-through businesses such as partnerships and sole proprietorships. It blows a hole in the government fisc so large than a Mack truck could be driven through it—as long as the operator is a sole proprietor. Congress attempted to limit the benefit to some types of earners and some types of businesses, but tax lawyers are busily finding ways to convert excluded businesses into qualified ones, and wage earners into independent business owners.

The structure of these types of laws makes estimating difficult enough. But two other factors make forecasting even more challenging.

The first is that the compounding of cost growth may take place years in the future and congressional scorekeeping conventions generally limit projections to the first 10 years, the so-called “budget window.”  Underestimating a growth rate by a couple percent per year, for instance, compounds to a very large number over time.

The second is that estimators may be reluctant to project very large costs in the absence of empirical evidence. For example, the 20 percent tax deduction for pass-through income is new, and there is little information upon which to predict the magnitude of gaming that will occur. The revenue estimator doubtlessly will assume some gaming, but may not be imaginative and daring enough to forecast without much data a large multiplier for what lawyers or providers, in absence of further whack-a-shelter legislation, will invent for their clients.

Much is wrong with a system that allows enactment of open-ended mandatory spending programs and tax preferences. Until we repair that system, it is worth remembering there is a built-in bias towards underestimating their long-term costs.