In a recent Kansas City Star commentary, Saint Louis University economics professors Lisa Gladson and Jack Strauss criticized my Show-Me Institute study of the earnings tax, They claim that there is no statistically significant relationship between earnings taxes and the growth of metropolitan statistical areas (MSAs). They also imply that I made such a claim in my study. In fact, I made no such claim and these authors are mischaracterizing my study. The purpose of this note is to set the record straight on my research. I will here demonstrate that my findings are perfectly consistent with theirs.

Let me briefly review my empirical evidence. I looked at more than 100 MSAs in the United States. The dependent variable I studied was the ratio of aggregate income reported by the U.S. Census for each primary city in that MSA to the aggregate income for the entire MSA. Basically, I measured the size of the city economy relative to the suburban economy. I then estimated a regression in which the city-to-MSA income ratio was the dependent variable and the earnings tax rate was the independent variable. The result was a significant negative correlation; in other words, cities with higher earnings tax rates tend to have smaller economies relative to their suburbs than cities with lower earnings tax rates. This result held both when using the 1990 Census as the data source and the 2000 Census as the data source.

In the article, I argued that this empirical finding was entirely consistent with sophisticated models of business and household locations within local economies. My analysis was grounded in a widely cited paper by Haughwout and Inman (2001). In it, they analyzed a model economy in which people living in a city took the tax structure as given and made consequent location decisions. When dealing with aggregate economic data, even at the city level, economists do not have a laboratory to run their experiments. Facts are extracted from the data — these are the economists’ observations — and economic theory is employed to account for these facts. Other models exist, but the Haughwout and Inman model can account for the many empirical regularities observed in city growth around the country.

Professor Strauss took an alternative approach and asked a different question. He estimated a regression in which the dependent variable is the growth rate of income in an MSA. The growth rate was computed for the period 1969 through 2007. His regression uses at least two independent variables. One is a dummy variable set equal to one for an MSA in which the primary city has an earnings tax, and set equal to zero otherwise. The other independent variable is the income level for the MSA in 1969, the so-called initial income level. The basis for Prof Strauss’ inquiry was the notion of economic convergence. If the coefficient on the initial income level is negative and statistically significant, the evidence indicates that cities with higher initial income levels tend to grow slower than cities with lower initial income levels. The convergence hypothesis follows, because the evidence suggests that cities starting off poor tend to catch up to the initially richer cities. The convergence hypothesis has been applied to cross-country datasets, and is useful for explaining why Japan and Korea — and, more recently, China — grow so fast. The return to capital is higher in these poor countries, and provides an opportunity for them to catch up to those already-rich countries. The convergence hypothesis cannot explain why Sub-Saharan Africa remains so poor. For our purposes, it is not obvious why convergence should apply to MSAs, but I will blog about this lesson more fully in the future.

In Prof. Strauss’ results, the coefficient on the earnings tax dummy is not significantly different from zero after one includes the MSA’s initial income level as an explanatory variable. He concludes that the earnings tax does not explain economic growth. He interprets these findings as indicating that cities across the United States are catching up and not affected by earnings taxes. I would proffer a slightly different interpretation. Metro areas that started off with lower incomes in 1969 are, on average, catching up to cities that started off with higher incomes. His unit of measurement is the MSA, not the city. This is kind of interesting. At first pass, convergence can characterize MSAs across the United States. The more difficult part is why rural areas are not catching up. If convergence can be attributed to low capital in the low-income metro areas, then it seems that rural areas — ones with low capital accumulation — would catch up to high-income urban areas. For me, the nagging problem about the convergence hypothesis is that Rocheport should start looking like Columbia in terms of capital. But Rocheport’s economy does not look like the Columbia economy. Agglomeration, or increasing returns, probably has something to do with this. As I said, I will save this digression for a future blog.

I do not dispute Strauss’s findings; however, he did mischaracterize my research. He asked a different question than I did. I contend that my question is more relevant for the question of earnings taxes. People can avoid the earnings tax by eschewing the political jurisdiction in which the tax is implemented. Insofar as the suburban area is not a perfect substitute for the city area, economic efficiency is lost and the earnings tax is distorting people’s behavior.

Consider the following situation for illustrative purposes. Suppose there are two cities. In City A, the metro area’s income increased at a 1-percent annual rate between 1969 and 2007. However, all the businesses moved out of a city and into the suburbs in 2000. In City B, the metro area’s income increased at a 1-percent annual rate between 1969 and 2007. In both City A and City B, Prof. Strauss’ measure of income growth would be 1 percent. If I further told you that City A had a 1-percent earnings tax and City B had no earnings tax, then according to Prof. Strauss’ unit of measure — the growth rate of income in the MSA — would indicate no statistical relationship between the earnings tax dummy and the growth rate of MSA income.

In contrast, I estimate the regression for city income to MSA income in 2000 and find a negative relationship between the earnings tax rate and the city-to-MSA income. The purpose of this illustration is to point out that his results do not contradict mine. His findings do not render my interpretation of the evidence as faulty. He asked a different question and got a different answer. Thus, a reasonable person could walk away believing both results are accurately depicted. Because we have different units of measurement for the city economy, we are clearly asking (and answering) different questions.

My results were mischaracterized in the Kansas City Star’s recent op-ed by Prof. Strauss. My goal is to properly characterize both sets of results. I am sure that more “teachable” moments will be forthcoming. Let’s proceed with skepticism before we accept any economic interpretation of the results.