Iyer asked in the comments to a recent post on state taxes and prosperity whether the picture would change if we looked at a different period (1996-2006), because 2007/08 was so anomalous.
For many data series, it’s very easy (as demonstrated here) to cherry-pick periods that seem to “prove” a given thesis. Far better to look at a whole bunch of periods of different lengths and see if there’s any pattern.
In this case, there doesn’t seem to be:
| Correlation between states’ income tax burdens as a % of state income (2008) and states’ change in median income, for the period: | ||||||||||||||
| Ending Years | ||||||||||||||
| Starting Years |
1986 | 1988 | 1990 | 1992 | 1994 | 1996 | 1998 | 2000 | 2002 | 2004 | 2006 | 2008 | ||
| 1984 | 29% | 41% | 26% | 26% | 9% | 15% | 12% | 14% | 18% | 17% | 30% | 11% | ||
| 1986 | 28% | -5% | -5% | -19% | -9% | -11% | -8% | -7% | -9% | 9% | -16% | |||
| 1988 | -31% | -28% | -34% | -24% | -23% | -23% | -23% | -25% | -11% | -29% | ||||
| 1990 | -1% | -17% | -7% | -8% | -6% | -4% | -7% | 13% | -12% | |||||
| 1992 | -19% | -5% | -6% | -5% | -2% | -6% | 13% | -12% | ||||||
| 1994 | 9% | 6% | 9% | 11% | 9% | 28% | 2% | |||||||
| 1996 | -4% | -2% | 1% | -2% | 13% | -6% | ||||||||
| 1998 | 3% | 7% | 3% | 19% | -3% | |||||||||
| 2000 | 4% | 1% | 20% | -5% | ||||||||||
| 2002 | -3% | 20% | -8% | |||||||||||
| 2004 | 27% | -7% | ||||||||||||
| 2006 | -30% | |||||||||||||
Think of each number representing a scatter plot with a trend line, like the graph in the previous post. Negative correlations mean higher taxes/slower growth; positive correlations are the opposite.) I’ve used percentage here because I think they’re easier to grasp at a glance. 20%, for instance, means a correlation usually represented by .20.
Averages:
All periods: -.18%
Periods >9 years: -2%
Periods >19 years: 4%
Averaging may not be terribly valid, but it is a quick way to do what you were doing anyway: looking at the chart and trying to eyeball the average.
These are vanishingly small correlations; don’t even need to calculate a p function to know they’re deep within the range of randomness, chance.
Counts:
All periods: 42% of 78 periods show positive correlation.
Periods >9 years: 42% of 36 periods show positive correlation.
Periods >19 years: 66% of 6 periods show positive correlation.
Draw your own conclusions.
Comments
3 responses to “State Taxes and Prosperity, Revisited”
Thanks. I had checked only 84-06 and looks like it was 0.3.
I will play with this data set next weekend, since I see some unusual patterns.
On another note, Krugman and others have repeatedly said that tax cuts do not increase tax revenues, which I believe. But no one has published long term trend data? Would you have access to it?
Thanks
@Iyer “Krugman and others have repeatedly said that tax cuts do not increase tax revenues, which I believe. But no one has published long term trend data? Would you have access to it?”
Yes, there’s plenty out there, though I don’t have cites handy. (Can say: even Mankiw — the great champion of “dynamic scoring” — says unequivocally that the less-taxes-more-revenues hooey isn’t true.) I wouldn’t spend any of my time on it. It’s loony tunes — only supported by carefully cherry-picked, post-hoc-ergo-propter-hoc anecdotes. We’re nowhere near the peak of the (back-of-the-envelope-inscribed) laffer curve.
Thanks. I now have your blog as one of my daily references (in my multitude arguments with those on the right). Lets popularize the websites more. Perhaps you could guest column on Huffpo or something?
good luck.