Friday, June 06, 2008

Springtime: Flowers bloom, liberals try to draw graphs 

Talking with some public finance specialists this week, I commented on this new study from Minnesota 2020, "Doesn't it seem like we get one of these every year or so?" Yes, they answered. You should just look them up.

The newspapers nevertheless dutifully report that the study shows "less investment public spending of tax dollars, less return."

The organization of the report is to draw two trends and then claim a correlation. The first trend is the decline in government expenditures as a share of economic activity in the state. Twenty pages of that stuff, either as a share of population or share of income. The second trend is a set of measures of economic activity, poverty, education, etc. (even "percentage of deficient bridges" -- that's a new metric!) that are meant to show that Minnesota is slipping relative to other states.

Now "relative to other states" is important. Most of the economic literature since Barro and Sala-i-Martin shows that when you have states that are rich and others that are poor, the poorer states tend to grow faster. This confirms a basic tenet of growth theory that dates back to the work of Robert Solow in 1957. In America, one of the issues that confounds the research on taxes and state growth is that the places that tend to have low state tax rates are often in the South, which has a worse position in economic performance when you look at long time series of performance going back to Reconstruction. This confounding factor creates a problem for correlations like the MN 2020 study. You would need a more careful regression analysis that teased out these complications. (There are other issues besides convergence, but this one might be the most important based on my experience with international cross-sectional regression analysis.) In short, (no) correlation does not mean (no) causation when you've got confounding factors like this. MN2020 don't need no steenkin' ceteris paribus.

But I need to show you more. If we don't believe that they've made a proper argument for no effect, what do we have to show that there is a negative effect from state tax rates to state economic growth. I thought I'd look to see what economists have said about the relationship. This is not a full list but a sample, more on which at the end. I restricted myself to papers I found on a site anyone could access, and to papers that had been through a peer-reviewed process. Thus we'll stay away from any of the publications of conservative or libertarian think tanks (which have many such citations.)

Anyway, here's a sample of three articles and relevant parts of their abstracts; go to the site and try your own search engine to find others if you want:

Tomljanovich, The Role of State Fiscal Policy in State Economic Growth. Contemporary Economic Policy, July 2004.
Using pooled annual U.S. state-level data from 1972 to 1998, a fixed-effects model is employed to examine the effects of changing tax rates on both state per capita output levels and growth rates. The results indicate that higher tax rates negatively influence short-run state economic growth, which lowers state output levels. However, long-run growth is unaffected by changes in state tax rates, even after adjusting for the effects of initial per capita output levels, state expenditures, and aid from the federal government. Nor do changes in state public spending rates and federal aid permanently alter state growth rates, implying that state fiscal policies have only transitory effects on state growth.
Becsi, Do state and local taxes affect relative state growth? Federal Reserve Bank of Atlanta Economic Review, March 1996.
This article presents an overview of relative state growth and relative state and local taxation from 1960 to 1992. After a brief discussion of the theoretical issues, the article surveys simple--but revealing--correlations across states and across time that characterize states' experiences. The correlations indicate convergence but also imply that shocks, including tax policy, matter for long-term growth. ; The author argues that the evidence on the growth effects of taxes has been mixed because empirical models imperfectly separate the growth effects of other government policies that occur simultaneously with tax policies. He demonstrates a simple way to get a more nearly accurate specification. His analysis shows that state and local taxes appear to have temporary growth effects that are stronger over shorter intervals and a permanent growth effect that does not disappear. ; In terms of policy implications, if long-term growth rates seem too low relative to other states, lowering aggregate state and local marginal tax rates is likely to have a positive effect on long-term growth rates.
Mofidi and Stone, Do State and Local Taxes Affect Economic Growth? Review of Economics and Statistics, July 1990.
Estimates for net investment and employment in manufacturing for 1962-82 support this conjecture, indicating that state and local taxes have a negative effect when the revenues are devoted to transfer-payment programs...
Let me add one paper that doesn't meet yet the criterion of peer-review, so should be treated as lesser evidence than the above, but it has a rather sharp conclusion about specification that strikes me as useful.

Reed, The Robust Relationship Between Taxes and State Economic Growth, Working Papers, University of Canterbury (NZ).
I estimate the relationship between taxes and economic growth using data from 1970-1999 and the forty-eight continental U.S. states. I find that taxes used to fund general expenditures are associated with significant, negative effects on economic growth. Further, this finding is robust across (i) alternative variable specifications, (ii) alternative estimation procedures, (iii) alternative ways of dividing the data into five-year periods, and (iv) allowing for individual-specific time and state effects. I also provide an explanation for why previous research has had difficulty identifying this robust relationship.
I've been meaning to put one of these together for awhile, and if someone wants to work with me on a meta-analysis of the literature, I've got access to enough databases to put together a long list of articles. You will, for awhile at least, need to accept that I've not cherry-picked these studies, or you will have to take up my offer. I'm not convinced that there's enough variability in the professional economics results to make a meta-analysis interesting, but I'd be happy to give that a try. Write if interested.