Retracted CRS Report on Taxes and Growth Flawed, But Still Cited
October 23, 2012
A recent Congressional Research Service report to Congress purported to show no link between the top income tax rates and economic growth. The report, written by CRS staffer Thomas L. Hungerford, was titled “Taxes and the Economy: An Economic Analysis of the Top Tax Rates Since 1945“, CRS 7-5700, R42729, September 14, 2012. The report’s methods were so seriously flawed that the study could not possibly pick up any relationship between taxes and growth, making its results, or lack of them, meaningless. The report has since been withdrawn from the CRS web site.
This type of CRS report to Congress is not available to the public on the Library of Congress web site. These reports to Congress can only be accessed on line at the CRS by Members of Congress and their staffs. Often, however, the reports are released by the Members of Congress, their staff, or CRS employees, and are picked up by other web sites. The Hungerford paper is still being cited on various websites – including the New York Times.
There are two glaring errors in the study’s approach. One is that it looks at the wrong time frame, the one year change in the rate of growth or productivity versus the change in the top tax rates. That is easier than looking at multi-year averages with lags, but it is wrong. It takes one to five years for a reduction in the tax on capital income to generate all the associated additions to the stock of equipment, and between one and ten years for structures. So productivity and income rise significantly, but gradually, after a tax cut that makes investment more attractive. Looking only at the first year effect throws out about 95 percent of the outcome. It is like looking under the streetlamp for car keys that were dropped in the parking lot many yards away, because that is where the light is.
The other major error is that it does not hold other factors constant. Shifts in monetary policy, changes in other taxes, regulatory actions, even weather-related effects on the economy can have significant impacts on the measured growth rate in a given year. If the tax change in the top rates is only one of many events during the year, its effects may be hidden in the statistical noise, especially if the other events are not even in the model. Indeed, one of the few partially correct observations in the study is that the top rates affect so few taxpayers that it might be wrong to expect to find a large impact on total GDP. That is all the more reason to make sure that other variables are accounted for, to reduce the noise.
For example, in 1981, the first Reagan tax reductions (including reductions in the top tax rates and the rate on capital gains) were phased in over three years. Some months before the tax bill was enacted, the Federal Reserve moved to tighten money to break the back of the ongoing double digit inflation. The resulting recession completely overwhelmed the effect of the tax rate reductions until 1983. The study ignores the change in Fed policy.
In the Tax Reform Act of 1986, the top income tax rates were reduced, along with the corporate tax rate. But other taxes on capital income were increased. The net effect was a rise in the tax burden on capital, and a slower rate of economic growth after the reform.
In the other direction, the increase in the top tax rates during the first Clinton Administration was accompanied by a continuing reduction in the rate of inflation, courtesy of the Federal Reserve. The lower inflation rate reduced taxes on investment by reducing the taxation of fictitious inflation-related capital gains, and by boosting the real value of the capital consumption (depreciation) allowances. These other factors reduced the amount an investment needed to earn, pre-tax, to be considered worth doing, and so raised the growth rate.
In fact, the GDP effect from changing the top rates is measurably large, when isolated, especially when measured in relation to the amount of revenue supposedly collected. One would expect the effects to be biggest where the rates are highest, because the after-tax change — 1 minus the tax rate — is the greatest, and the productivity and earnings per hour of the affected taxpayers are the highest.
Missing variables are the bugaboo of regression analysis. Even as undergraduates, we are warned about the possibility of spurious correlation (or non-correlation) when other factors are ignored. One classic example is the high correlation that one finds when regressing the incomes of school teachers in Massachusetts against spending on liquor in that state. They rose together due to inflation, not due to any disproportionate marginal propensity to drink on the part of the teachers.
We’ll leave that issue for the next CRS report.