Menu
Tax Notes logo

Economic Analysis: Will 'Shock' Macroeconomics Smooth the Path to Tax Reform?

Posted on July 31, 2017

Previous researchers, and I myself, have been mistaken before and probably will be again.

— Christopher A. Sims, 2011 Nobel Prize Lecture

Like it or not, our next round of major tax legislation will not be crafted around long-standing fine points of principled tax policy that are so familiar to our readers. Instead, it will depend critically on assumed growth effects of tax reform — or, more likely, tax cuts, which cannot be estimated with much certainty even by our most advanced econometricians.

Hemmed in by both unprecedented long-term fiscal constraints that cry out for caution and by irresponsibly generous vows of major tax relief, the path of least resistance for Republicans will be to cherry-pick favorable macroeconomic estimates. This will allow them to assume that tax cuts and other policies they favor will not only substantially boost employment — a huge political win in and of itself — but will also simultaneously increase taxable income from policy-induced growth.

The first draft of this strategy was outlined on May 23 with the release of President Trump’s bare-bones budget. It assumed that an unspecified matrix of Republican policies would increase average annual economic growth from the 1.9 percent projected in the Congressional Budget Office baseline to 3 percent. More recently, on July 18 the House Budget Committee released its draft budget resolution, which assumed that deficit reduction, deregulation, healthcare reform, and tax cuts would increase the annual economic growth rate to 2.6 percent. This assumed additional growth would painlessly yield House Republicans the ability to claim an extra $1.8 trillion of funds over the next decade for tax cuts and deficit reduction.

Given that these bold predictions will be playing a critical role in the looming tax battle, it is a good time to review what economists know about the effects of taxes on growth. The branch of economics devoted to taxes, called public finance, has always been a lively area of study and debate. But strangely enough, academic macroeconomists who are best equipped to assess the effects of taxes on the overall economy largely neglected fiscal policy for decades. From the 1970s through the 1990s, despite its preeminence in undergraduate textbooks, most leading macroeconomists gave short shrift to the idea that tax and government spending policy could help modulate business cycles. Instead, embracing the teachings of Milton Friedman and rejecting those of John Maynard Keynes, they collectively developed a near-obsessive focus on monetary policy. Further, in what can clearly be called a blunder of outrageous proportions by the profession, practically none of the attention lavished on finance looked beyond monetary affairs to the inner workings of credit markets.

On their own, macroeconomists began learning that monetary policy was not as important as they once thought. As Christopher A. Sims stated in 2011: “Erratic shifts in monetary policy are not the main source of cyclical variation in the economy” (“Statistical Modeling of Monetary Policy and Its Effects,” Nobel Prize Lecture, Dec. 8, 2011). And, of course, whatever reluctance academics had to study the macroeconomic effects of fiscal policy was blown away by the Great Recession of 2007-2009 and the subsequent policies of stimulus (in the United States) and austerity (in Europe).

‘Explosion in Academic Work’

The revival of academic interest in fiscal policy brought with it new and sophisticated statistical techniques to examine tax policy that had originally been developed to study monetary policy. Non-economists who follow tax issues might be surprised to learn that these new models have received little attention on Capitol Hill. One exception was the testimony of Kevin A. Hassett to the House Ways and Means Committee in February 2016. He described to committee members a “recent explosion in academic work that relies on a significant methodological innovation.” This introduction to lawmakers of this new brand of research is particularly notable because the highly respected Hassett is now Trump’s nominee to chair the Council of Economic Advisers.

Although the models most often used by the Joint Committee on Taxation, CBO, Tax Foundation, and Urban-Brookings Tax Policy Center are continually updated to include new data and advances in theory, they basically use highly structured frameworks that require a multitude of assumptions about private and public sector behavior. These assumptions (for example, about the response of labor supply to changes in taxes) are informed by the latest research, but the staffs that feed these assumptions into their models understand the uncertainty of the underlying estimates. However, the models themselves treat these estimates as gospel. This can easily create a false sense of confidence about the reliability of their results (despite the unending reminders from government economists about the uncertainty of their forecasts). In this broad sense, the models used in policy shops are similar to those first used in the 1960s and still in use by commercial forecasters today despite academics’ widespread disdain for them. “Because of their very structure, [conventional large-scale econometric models] largely postulate rather than document an effect of fiscal policy on activity,” wrote Olivier Blanchard and Roberto Perotti in a much-cited 2002 study (“An Empirical Characterization of the Dynamic Effects of Changes in Government Spending and Taxes on Output,” 117 Q.J. Econ. 1329 (2002)).

The new statistical methods at the cutting edge of macroeconomic research rely much less on theoretically derived preconceptions of behavior. As much as possible, they let the data tell their own story. But the data talk in riddles. One of the most challenging problems is disentangling the two-way causation that plagues the relationship between most economic variables. Does the stock of money increase because the economy is growing? Or does growth in the money supply cause the economy to grow? Did a certain tax cut really stimulate the economy? Or did Congress pass a tax cut because deficit-reducing growth allowed lawmakers to pass tax cuts? When this two-way causality exists, an estimate intended to measure the effect of policy changes can be significantly biased.

As a result of this endemic problem, macroeconomists over the last few decades have engaged in hundreds (or perhaps thousands) of empirical studies trying to identify the components of policy that are “exogenous shocks,” that is, policy variables in which the causality is one-way. If the causality is one-way, there is a greater chance that the measured statistical relationship will accurately tell us by how much and when a policy lever (like a tax cut) will affect economic growth.

There are two basic methods of identifying exogenous shocks. The first is a math-intensive approach called vector autoregression (VAR). Using VAR methods, economists identify policy shocks as the components of changes in policy variables (like taxes, government spending, and money supply) that cannot be predicted by any past data. They are surprises in a statistical sense. And as such, it is assumed policymakers did not formulate them in response to observable economic conditions (and thus there is no bias-creating causality from the current economic conditions to the policy variable). An influential paper by Blanchard and Perotti (cited above) applied this approach and found that taxes had a multiplier effect of 1, which in this context means the effect of a tax cut equal to 1 percent of GDP had its peak effect equal to a 1 percent increase in GDP eight quarters after taking effect.

The second approach is almost entirely math free and is based on historical narrative. In reference to U.S. tax policy, the groundbreaking work was done by Christina D. and Paul H. Romer as reported in two papers finalized in 2009 and 2010. (Christina Romer was chair of President Obama’s Council of Economic Advisers from January 2009 through September 2010.) Romer and Romer painstakingly reviewed the historical record of tax policy changes back to 1950 and then set about dividing them into two categories: those changes that were responses to current economic conditions (such as a recession) and those that were not responses to current economic conditions (such as legislators’ belief in the probity of tax relief or the need for long-term deficit reduction). Because the first group (endogenous tax changes) would produce downward-biased estimates, Romer and Romer excluded these and only included the effect of exogenous changes (shocks) in their estimates of taxes on growth. Deviating sharply from the accepted wisdom that evolved from the Blanchard and Perotti research, Romer and Romer estimated a multiplier effect of 3 that peaked 10 quarters after enactment of exogenous tax cuts (“A Narrative Analysis of Postwar Tax Changes,” unpublished paper, June 2009; and “The Macroeconomic Effects of Tax Changes: Estimates Based on a New Measure of Fiscal Shocks,” 100 Am. Econ. Rev. 763 (2010)).

Implications for Policy?

In his testimony before the Ways and Means Committee, Hassett wrote of this new research: “The evidence it has generated, in fact, is so striking that it is essential that U.S. policymakers begin to incorporate it into their thinking.” So what are its quantitative implications? Let’s first look at one of the simpler examples directly from Romer and Romer’s studies, the Omnibus Reconciliation Act of 1990. Signed into law on November 5, 1990, this legislation had a one-time shock of $35.4 billion in the first quarter of 1991 (0.4 percent of GDP) and of zero for the following 12 quarters, according to Romer and Romer. (Unfortunately, using these methods, the measurement of the effects of shocks beyond three years is highly unreliable.) The JCT at the time estimated the net revenue gain (excluding extension of temporary provisions) for each of the fiscal years from 1991 through 1995 to be $18 billion, $30 billion, $27 billion, $31 billion, and $30 billion. As the revenue estimate suggests, the legislated tax increases were intended to be permanent.

Figure 1 shows Romer and Romer’s estimate of the percentage change in GDP after an exogenous tax cut equal to 1 percent of GDP. The peak impact of the tax increase is approximately 3 percent at 10 quarters and declines slightly thereafter. Over the three-year period, the average reduction in GDP is 1.7 percent. Scaling these figures to interpret the effect of the 1990 tax increase, the tax shock of 0.4 percent of GDP in the first quarter of 1991 reduced GDP by 1.2 percent below what it would have been in the third quarter of 1993. On average over the three-year period ending at the end of the first quarter of 1994, the 1990 act’s tax increase reduced GDP by 0.68 percent.

figure

What implications might the estimates like those of Romer and Romer have for upcoming tax legislation? As noted above, under the House Budget Committee’s draft budget resolution, the economy is assumed to grow on average at a rate of 2.6 percent instead of the 1.9 percent estimated in the CBO baseline. This means that under the House plan, GDP in 2027 will be 9.7 percent larger than under the CBO baseline. If tax cuts accounted for half of this new growth (with the rest attributable to healthcare reform, deregulation, and lower deficits), and using reasonable assumptions about a phase-in of the new policy, it would require a tax cut with an annual cost of approximately $50 billion in 2018 increasing to approximately $370 billion annually by 2025. In terms of the 10-year cost usually cited in government circles, the total tax cut would amount to more than $2.5 trillion over 10 years to get the economy halfway to the assumed 9.7 percent cumulative increase over baseline projections in 2027. If the Romer and Romer estimates are considered too high, the necessary tax cuts would have to be proportionately larger.

Of course, since 2010, macroeconomists have been busy updating and revising these earlier studies. At first it appeared as though the magnitude of the Romer and Romer estimates would not stand up to scrutiny. For example, a 2012 paper by Carlo Favero and Francesco Giavazzi reported that when variables in addition to Romer and Romer’s tax shocks were included in estimating equations, the multiplier effects of exogenous tax changes were reduced from 3 to 1 (“Measuring Tax Multipliers: The Narrative Method in Fiscal VARs,” 4 Am. Econ. J.: Econ. Pol’y 69 (2012)).

But soon after that, Karel Mertens and Morten O. Ravn developed a more sophisticated hybrid approach combining VAR methods and Romer and Romer’s data and pinpointed the shortcomings of the estimates by Blanchard and Perotti. Mertens and Ravn estimated a multiplier that peaked at 3.2 after five quarters (“A Reconciliation of SVAR and Narrative Estimates of Tax Multipliers,” 68 J. Monetary Econ. s1 (2014)). In a 128-page review of recent research, Valerie A. Ramey concluded that large multipliers similar to those estimated by Romer and Romer and by Mertens and Ravn were consistent with her own work with the underlying data (“Macroeconomic Shocks and Their Propagation,” National Bureau of Economic Research Working Paper 21978, Feb. 2016).

Outstanding Issues

Instead of building toward a new consensus, advances in econometric modeling are provoking economists to reexamine the most fundamental questions about the effects of tax policy on the overall economy. A sampling of recent research outlined below gives a sense of the varied activity macroeconomists are pursuing and the variety of results they are finding.

How do changes in marginal effective tax rates differ from changes in average tax rates?

All of the above studies examined the effect of average tax rates on the economy. But economists usually believe that the effects of marginal effective tax rates are more important than those of average rates. In a 2011 paper, Robert J. Barro and Charles J. Redlick found that marginal tax rate shocks have a larger effect on GDP than those of average tax rates (“The Macroeconomic Effects of Government Purchases and Taxes,” 126 Q. J. Econ. 51 (2011). In a 2015 paper using different data and statistical methods than Barro and Redlick, Mertens estimated similar effects on GDP of exogenous changes in marginal tax rates and, in concert with Hassett’s testimony, concluded that “federal tax policy seems to operate through incentive effects rather than disposable income and demand stimulus” (“Marginal Tax Rates and Income: New Time Series Evidence,” National Bureau of Economic Research Working Paper 19171, Sept. 2015).

Do different types of taxes (that is, individual income, corporate, transfer payments) have different effects on the overall economy?

Everybody understands that different types of tax changes should have different effects on the economy, but macroeconomists are only beginning to explore them. In a 2013 paper, Mertens and Ravn separated individual income tax shocks from corporate tax shocks and found that individual income tax cuts were more effective in creating jobs in the short run than cuts to corporate profit taxes. They also found that changes in corporate tax rates did not appreciably change corporate tax revenue (“The Dynamic Effects of Personal and Corporate Income Tax Changes in the United States,” 103 Am. Econ. Rev. 1212 (2013).

In what may seem to some as an unnatural deviation from everyday parlance, academic economists do not consider Social Security benefits and other transfer payments to be government spending, but negative taxes. (So, a warning to conservatives — some studies that show growth effects from tax cuts could also be interpreted as showing growth effects from increases in mandatory spending.) In a 2016 paper, Romer and Romer found that a permanent increase in Social Security benefits raises aggregate consumer spending roughly 1-for-1 after checks are issued but that the effects decline sharply after one year. This contrasts with their findings about income tax shocks that are larger and more persistent (“Transfer Payments and the Macroeconomy: The Effects of Social Security Benefit Increases, 1952-1991,” 8 Am. Econ. J.: Macroeconomics 1 (2011).

What effects of tax cuts are anticipated before they take effect?

Economic theory and common sense tell us that if a tax cut is anticipated, it will reduce GDP before the effective date because taxpayers will delay some taxable activity until lower taxes take effect. This is one area where so far there seems to be little disagreement. The data strongly indicate that anticipated but not yet implemented tax cuts reduce output and investment. (See, for example, Mertens and Ravn, “Empirical Evidence on the Aggregate Effects of Anticipated and Unanticipated U.S. Tax Policy Shocks,”4 Am. Econ. J.: Econ. Pol’y 145 (2012).) This research implies that the conventional wisdom that there will be some sort of tax cut in late 2017 or early 2018 is temporarily slowing economic growth.

Do the effects of tax cuts vary with a government’s level of indebtedness?

If a government cuts taxes when its debt levels are high, taxpayers have good reason to expect the tax cut will not be permanent. Among many others, Ethan Ilzetzki, Enrique G. Mendoza, and Carlos A. Végh find that high-debt countries generally have lower tax multipliers than countries with more stable finances (“How Big (Small?) Are Fiscal Multipliers?” 60 J. Monetary Econ. 239 (2013)). This research implies that if current projections of soaring U.S. debt levels are not reduced, the positive effects of tax cuts may well be less than those estimated from historical experience. As former CBO Director Douglas Holtz-Eakin told the House Budget Committee on June 7, “Businesses, entrepreneurs, and investors perceive the future deficits as an implicit promise of higher taxes, higher interest rates, or both.”

Supply or Demand?

Finally, it is essential to understand whether the widely estimated positive effects of tax cuts are supply- or demand-side effects. If they are from the demand side, and if the current economy is near full employment (as most but not all economists believe), it is far less likely that any tax cuts enacted soon will mimic the positive estimated effects that the studies are detecting. Most of the research cited above is agnostic or supports the idea that tax shocks are demand-side stimulus. William G. Gale and Andrew A. Samwick point out that the rapid response of GDP to tax shocks is similar to the response of GDP to shocks in government spending, which generally is not considered to have significant supply-side effects (“Effects of Income Tax Changes on Economic Growth,” in Alan J. Auerbach and Kent Smetters, eds., The Economics of Tax Policy, Oxford, 2017, p. 13). On the other hand, Hassett makes the case in his congressional testimony that the Romers’ and the high-end VAR results are consistent with convincing new evidence that the responsiveness of labor supply to tax cuts is larger and more rapid than commonly assumed. This means tax cuts could be effective in promoting growth even in the current environment where the economy is peaking.

The central policy goals of the tax legislation that will likely begin to take shape in earnest this fall are job creation and fiscal responsibility. Congress’s ability to achieve both of these goals depends critically on how much the economy responds to tax changes. So right now would seem to be the perfect time for an airing out of the debate on the quantitative effects of taxes on growth. Unfortunately, we are likely to see more magic-wand assumptions about growth rates that are at the outer limits of what most economists consider possible (even if Congress adopted economically ideal policies). Despite all the considerable advances in empirical research, our understanding of tax policy effects on the macroeconomy is still a work in progress. Given the humility of economists who have spent their professional careers studying this topic, it is amazing how so many unschooled legislators and media commentators confidently assert that tax cuts will project the economy to new heights.

Copy RID