The Capital Gains Wars
by Jane G. Gravelle
Essays on Economics in Government is a series of occasional essays in which all SGE members are welcome to contribute original essays which draw upon and recall special experiences during their time in government.
An economist laboring in the policy-making arena may find his or her technical tools of the trade rarely on display. As we sometimes say, as we are laboring to explain fundamental economic ideas (such as why rebates of value added taxes do not confer a trade advantage), a little economics goes a long way. There are no separating hyper-planes or Kuhn-Tucker conditions in our work.
But technical issues became the center of political interest during a period that I like to call the “Capital Gains Wars.” It was at the end of the 1980s, after the legislation of the early and mid- 1980s that had cut rates and indexed the individual income tax. Before that time, periodic tax cuts were common to return the “fiscal dividend’ that arose from bracket creep. And, tax cuts are always popular. However, operating under restrictive budget rules in the wake of the 1981 and 1986 tax revisions, it was hard to envision any type of tax cut. Except, of course, one that raised revenue.
Capital gains tax cuts were claimed to do so. Interest in “supply side” economics and behavioral changes that could raise revenue from tax cuts, which played an important role in the 1981 tax cut, remained. But these notions did not enjoy much legitimacy from the economics profession; nor had the 1981 tax cuts caused the revenues to pour in. However, a new and burgeoning literature on realizations responses to changes in capital gains tax rates suggested that cutting capital gains taxes might do the trick. Capital gains tax rates, which had enjoyed a great deal of favorable treatment through the entire history of the tax code, had actually been raised as part of the simplification and base broadening in the 1986 individual tax reform (and slight tax cut). Complaints had already started.
The administration proposed a 30% capital gains exclusion, which would move the top rate back towards about 20%. Both Treasury and Joint Committee on Tax (JCT) revenue estimators prepared estimates, but while Treasury was saying revenue would be raised, JCT was saying it would be lost. Mind you, both of them included significant realizations responses – and their elasticities were actually close given the range of empirical estimates. The static revenue loss was estimated by the JCT to be about $100 billion over five years, but the estimated loss was only $11.4 billion with their realization response. Using a lower baseline, Treasury estimated a static loss of about $75 billion, but a gain of $12.5 billion with their realization response.
Treasury officials argued that their elasticities were nevertheless quite conservative given the range of estimates in the literature and included these arguments (along with a table of elasticities) in testimony before the Senate Finance Committee.
There were also some side skirmishes in the battle, one involving distributional effects. If one prepared distributional effects after taking into account large projected realizations responses, these distributional tables could show the capital gains tax cuts as, not a benefit, but a burden – and Treasury was also doing these tables. JCT was doing standard distribution tables that assumed no behavioral response.
Another side skirmish was the argument that cutting these taxes would increase savings, economic growth and, therefore, revenues. If capital gains tax cuts increased savings, so the argument went, then they understated the revenue feedback. Thus, feedback effects that did not take into account these growth effects were understated.
I was asked to examine the issue of revenue effects. The growth argument was relatively easy for me to deal with because I addressed these issues all the time. Trying to sort out the empirical evidence on the realizations response was tougher. It required going out to the tool shed and getting my old econometrics tools (which had rusted a bit) out. Plus, econometrics was not my favorite subject, an attitude that stems, I think, from sitting through econometrics class while I had morning sickness. I also gathered the fifteen studies – the twelve cited by Treasury and three additional ones I found – to review.
Some of my paper involved trying to get as many of the elasticities as possible to be comparable and relevant to the proposed legislation, as many were estimated in a semi-log or net-of-tax form that caused the tax elasticity to vary with the tax rate. The value at the sample mean was not always the proper value.
But the important issue was that the magnitude of the elasticities depended on whether they came from cross section or time series. Measured at the proper tax rate, all of the time series estimates were below one – the magic number that, roughly speaking – causes a tax cut to have no effect on revenue. They ranged from 0.27 to 0.89, with a median and mean of around 0.65. The cross section (and panel data that included strong cross section elements) elasticities were higher on average and more variable – estimates ranged from 0.55 to 3.8 with a mean of about 2 and a median of about 1.5.
According to my comparison, Treasury estimates were high relative to the time series estimates although low relative to cross section, while the JCT assumption was about typical of time series estimates.
Both approaches had problems, some in common and some specific to the approach. Cross section estimates have the particular problem that they may largely reflect transitory effects; there may also be individual specific differences in preferences that are correlated with tax rates. Time series estimates suffer from aggregation bias, small sample size, and limits on the ability to control for other macroeconomic factors. Moreover, they cannot capture the adjustment path. My basic conclusion was that the problems with cross section estimates were much more serious than those with time series and that there were some reasons to find even times-series estimates overstated.
That paper is how I ended up participating in a hearing called to sort out the debate on the econometric studies. And so it was that representatives of the Treasury and the JCT convened before the Senate Finance Committee to argue the merits of cross section vs. time series regressions, in a memorable hearing. It should have been a great day for economists to debate these technical issues in a policy arena that really mattered. The only problem is that the officials who did the testifying were lawyers who presumably knew nothing about statistical inference (as was likely the case of the Senators). It was a most entertaining hearing, as I waited to present my own testimony. And I never forgot this one time when the technical tools of the trade gained such prominence, only to be wielded by proxy.
As a postscript: It subsequently dawned on me that there was a natural bound to the realizations response – since realizations cannot exceed accruals – that might be instructive and I developed historical data and used a simulation technique to try to find the upper limit of the elasticity consistent with the various functional forms used in regressions. My analysis suggested an elasticity of no more than 0.5 and very clearly ruled out the high elasticities in some of the cross section studies. That paper encouraged two Treasury researchers to undertake a new cross section study that used state variation in tax rates to identify permanent effects – a paper that found a very large transitory elasticity but a negligible permanent elasticity. At the same time the surge in realizations that occurred in 1986 just before tax rates were to go up demonstrated the enormous power of transitory effects. All of this suggested that the JCT as well as Treasury permanent elasticities were too high.
The administration’s proposal was not enacted, although the capital gains tax rate was capped and did not increase with the general rate increases in 1993. Ultimately, however, while the debate about realization responses and new research may have helped shelter the JCT from criticisms that their response was too low, there were no changes in their assumed elasticities when the cut was eventually enacted in 1997. At that time the rate was set at 20%, roughly to the Bush administration’s original proposal. It was cut again in 2003 to 15%. The reductions were done through a rate reduction rather than exclusions and these rate reductions were not a preference for the alternative minimum tax (AMT) – an ironic development since the original purpose of the minimum tax was clearly to tax the capital gains preference. So capital gains ended up being more lightly taxed than before 1986 and more lightly taxed than at any time in history (and presumably scored as costing less revenue than suggested by the newer empirical studies). The AMT is no longer aimed at capital gains, but is increasingly affecting middle class people with children, whom it was never intended to affect (because exemptions were not indexed for inflation). And the econometrics debate has never resurfaced.
Jane Gravelle is a Senior Specialist in Economic Policy, Government and Finance Division at the Congressional Research Service. The views in this article do not reflect the views of the Congressional Research Service