Quantcast

Do Taxes Affect Economic Growth?

by Matt Mitchell on September 21, 2012

in Economic Freedom, Economic Growth, Tax and Budget

The CRS has a new report by Thomas Hungerford that has attracted some attention. It seems to suggest that taxes do not affect economic growth. To be precise, it seems to suggest that the top marginal tax rates of two taxes in particular—the personal income tax rate and the capital gains tax rate—have little statistically significant effect on economic growth.

A few comments:

First, as William McBride of the Tax Foundation notes in an excellent post, the study only examines two taxes.

The largest tax on investment is the corporate income tax, but the CRS report ignores corporate rates, even though other studies have found corporate taxes to be the most economically damaging.

Second, Will also rightly notes that the study focuses exclusively on the statutory rates of these two taxes, ignoring their actual incidence.

Because Congress has larded up the tax code with piles of credits, exemptions, and deductions, statutory rates often have little relationship to the rates people actually pay (just ask GE). For that, we need an estimate of effective marginal tax rates. As it turns out, many (most?) researchers who study taxes in the U.S. do attempt to get at this. Barro and Redlick’s piece is one example. It employs a tax model which accounts for the “complexity of the federal individual income tax due to the alternative minimum tax, the earned-income tax credit (EITC), phase-outs of exemptions and deductions, and so on.” Using this measure, Barro and Redlick find taxes do have “significantly negative effects on GDP.”

Third, the piece makes no attempt to account for reverse-causality (what economists call endogeneity).

Put simply, tax rates do not change randomly. If they did, that’d be great for researchers because randomization is the gold standard of the scientific method. But because policy makers are not so keen to let economists experiment with the national economy, tax rates don’t change randomly. Instead, governments tend to change rates in response to changing economic conditions; they cut taxes when the economy is weak and they raise taxes when the economy is strong. This makes disentangling cause and effect quite difficult.

Imagine we studied new drug treatments this way. Instead of large scale controlled experiments with randomized treatments and placebos, what if we only had one patient, and we only gave her a treatment when her condition worsened? If, after the treatment, her condition deteriorated further, would we conclude that the drug did her in? A simple statistical test would say so: drug applied, condition worsened. But such a test would ignore the fact that she only got the drug because she was sick to begin with! The point is that it’d be irresponsible to conclude anything from such a small sample and without trying to control for reverse-causality.

That’s why economists go to great lengths to mimic the conditions of a controlled, randomized experiment. In the case of tax studies, the best example of this is the study by Christina and David Romer. They painstakingly combed the archives of presidential speeches and government documents to identify tax changes that came about for reasons other than the condition of the economy. They found that these sorts of plausibly exogenous tax changes had quite significant macroeconomic effects. In their words:

Our results indicate that tax changes have very large effects on output. Our baseline specification implies that an exogenous tax increase of one percent of GDP lowers real GDP by almost three percent.

Fourth, I also take issue with the way the report characterizes some of the existing literature. For example, Hungerford writes that:

There is substantial evidence…to suggest that labor supply responses to wages and tax changes are small for both men and women.

This is only true if the question is whether people who already have jobs change their labor supply in response to tax changes. Increasingly, however, the literature has identified another margin that matters: lifetime decisions about schooling, fertility, and work experience. If taxes affect these things, then they still affect labor supply, even if they don’t seem to affect short-term decisions about how much labor to supply. As Michael Keane’s recent piece illustrates, taxes have a very pronounced effect along this margin, especially among women.

The CRS report is interesting. And its results should be added to the body of literature on taxes. But it is hardly reason to throw out decades of other research which suggests taxes do harm growth:

 

  • Pingback: 10 Monday AM Reads | The Big Picture

  • Pingback: How Much Would the Fiscal Cliff Raise Taxes? « Constitutional Free America

  • http://asymptosis.com Steve Roth

    Is it significant that the literature listed here does not include Nijkamp and Poot’s 2003 meta analysis? It’s quite clear there that the historical correlation between tax levels and growth is small, unclear, and/or nonexistent.

    • Matt Mitchell

      Thanks for the comment, Steve. That meta-analysis is about more than taxes and, at several points in the paper, they acknowledge that their sample of tax studies is very limited (only 10). Nevertheless, they don’t seem to find what you think they find. Quoting them: “Among the 10 tax studies in our sample, there were no studies that found that higher tax rates were associated with higher economic growth. Instead, there appears to be empirical support for the hypothesis that higher taxes lower growth (with an overall 60 percent probability) and inspection of the individual cases shows that this is particularly true for studies that focus on marginal rather than average tax rates [MM: nota bene]. However, the small sample size leads to a wide interval of (0.26, 0.89) for the proportion of studies supporting the null hypothesis and indicates that this conclusion must remain rather tentative. As noted earlier, this is one area where further empirical research and meta-analysis are desirable.”

Previous post:

Next post: