Matt Yglesias doesn’t think that Figure B in Tom Hungerford’s new paper on corporate tax rates and economic growth “proves” a non-relationship between the two. He’s right, of course, but then goes on to argue that including this chart specifically, and including charts in blogs generally, is a borderline shady practice. This I don’t really get.
So, I shall make the case for the use of charts with a chart, below.
But, first, a quick thought on another point in Matt’s post. He notes that we at EPI aren’t keen on seeing corporate tax rates cut reformed anytime soon because we like revenue and we “feel that the corporate income tax raises it in a distributionally progressive way.” We do like revenue, but our thoughts on the incidence of corporate tax rates are not “feelings”—they’re pretty evidence-based. The CBO, for example, used to assume that 100% of the corporate income tax fell on owners of capital (a very well-off bunch relative to the rest of the population). Now they assume that 3/4ths does. So our view that most of the incidence of the corporate income tax falls on capital is not some wild-eyed heterodox view.
We’d probably quibble even with the change to put some of the incidence on wages (though I’m not a huge expert in this one), and we know of the growing effort by some economists to claim that the incidence of the corporate income tax is actually borne by workers, but we’re convinced by the counter-evidence (see here (PDF) and here (PDF) for some of this evidence). A side-note—it’s always fun to see people who claim in one venue that corporate tax rates should be cut reformed because the incidence is actually borne by workers claim in another that the corporate income tax is bad because it “double-taxes” capital income. Really, you can’t have both claims.
Ok, on to the role of charts in blogging and policy papers. Yes, Figure B in Tom’s piece doesn’t end the argument. Luckily there’s Figures C and D and even a bonus appendix table, complete with regression coefficients.
But I’d argue Figure B is still worth including. The way that I approach an empirical-y question is to first look at broad trends in data to see if any obvious relationship seems to jump out. I obviously don’t stop there. Sometimes there’s no clear dual-axis relationship between two variables, but one emerges once other controls are added. Of course, some of the best advice I’ve heard on empirical economics (I’m pretty sure I heard it from a talk by Richard Freeman, but could be wrong) is that if you don’t see a relationship between two variables in broad, visual data, you should be a bit skeptical if a strong one emerges in later probing.
And, of course, sometimes a relationship that does appear in the broad visual data (or, in the ocular econometrics, in the pejorative phrase sometimes used) disappears pretty quickly when you start probing more. Either way, so long as the additional testing is done, taking readers with you on your investigatory walk of looking from broad to more narrow evidence doesn’t seem too shady to me.
And to be clear, it’s true that year-to-year changes in GDP are not going to be influenced much by anything beyond monetary and fiscal policy changes—but by plotting these year-to-year changes over a bunch of years, it could in theory become clear that average GDP growth was faster in one period than another. And if this decline in average growth was associated with, say, an increase in corporate tax rates, well at least the opposing case for corporate tax cuts reform would survive Figure B.
So, dual-axis charts that plot income growth against something else can sometimes be useful or intriguing. Below is the earlier-promised chart.
This is the average annual change in incomes for the bottom 90% of households plotted against the top marginal tax rate for the last 60ish years. Looks pretty clear to me that growth of the bottom 90% incomes is actually noticeably higher in the pre-1973-ish period (the red-line tends to hover higher in the earlier period) and that its decline to a lower average growth rate (outside the late 1990s boom) for the generation that followed looks plausibly coincident with the decline in the top marginal tax bracket. This, of course, “proves” nothing on its own. But I think it’s intriguing. And in fact a very similar relationship survives some more rigorous testing (check out Table 1 in this pdf). It might make a good Figure A (or B) in a new paper.