In light of the recently released Goldman Sachs study on comparative state performance during the recession (h/t Michael Leachman; no link, but thanks to study author Zach Pandl for sending me more info on the study) and the widely-quoted analysis at Political Math, both of which use raw employment numbers as their core, I've thought about what a better measure would be.
The problem with raw job numbers, and we see it in the case of Texas, is it makes a state look better than it is if it has a rapidly increasing population that job growth is unable to keep up with. Since December 2007, the state has added 247,000 jobs, but its labor force grew by 739,000 through June 2011, which is why the unemployment rate has gone up by 4 percentage points.
Pandl's study calculates June 2011 employment as a percentage of December 2007 employment, for all states. From the graph he sent me, it appears that North Dakota is tops at about 109%, Alaska next about 104%, DC third at about 103%, and Texas fourth at about 102%. I asked Pandl, what if we use a different dependent variable for your regression analysis, unemployment rate in July 2011 (newly available in the past week) as a percentage of the unemployment rate in December 2007. If we want to know how much worse a state's unemployment got, this is a pretty intuitive measure. I hope Pandl considers this calculation, because it turns out the results are substantially different. The table below ranks the states from best to worst (lowest percentage to highest percentage) on this measure.
State | July 2011 P | Dec 2007 P | Ratio 1 | |
North Dakota | 3.3 | 3.3 | 100 | |
Alaska | 7.7 | 6.5 | 118 | |
Oklahoma | 5.5 | 4.5 | 122 | |
Nebraska | 4.1 | 3.2 | 128 | |
Arkansas | 8.2 | 5.9 | 139 | |
Vermont | 5.7 | 4 | 143 | |
Michigan | 10.9 | 7.6 | 143 | |
New Hampshire | 5.2 | 3.6 | 144 | |
Minnesota | 7.2 | 4.9 | 147 | |
Kansas | 6.5 | 4.4 | 148 | |
Iowa | 6 | 4 | 150 | |
Ohio | 9 | 6 | 150 | |
Maine | 7.7 | 5.1 | 151 | |
Mississippi | 10.4 | 6.8 | 153 | |
Wisconsin | 7.8 | 5 | 156 | |
South Dakota | 4.7 | 3 | 157 | |
Missouri | 8.7 | 5.5 | 158 | |
New York | 8 | 4.9 | 163 | |
South Carolina | 10.9 | 6.6 | 165 | |
West Virginia | 8.1 | 4.9 | 165 | |
Pennsylvania | 7.8 | 4.7 | 166 | |
Kentucky | 9.5 | 5.7 | 167 | |
Massachusetts | 7.6 | 4.5 | 169 | |
Oregon | 9.5 | 5.6 | 170 | |
Illinois | 9.5 | 5.5 | 173 | |
Virginia | 6.1 | 3.5 | 174 | |
DC | 10.8 | 6.1 | 177 | |
Louisiana | 7.6 | 4.2 | 181 | |
New Mexico | 6.7 | 3.7 | 181 | |
Connecticut | 9.1 | 5 | 182 | |
Indiana | 8.5 | 4.6 | 185 | |
Tennessee | 9.8 | 5.3 | 185 | |
Texas | 8.4 | 4.5 | 187 | |
Wyoming | 5.8 | 3.1 | 187 | |
Colorado | 8.5 | 4.5 | 189 | |
Maryland | 7.2 | 3.8 | 189 | |
Hawaii | 6.1 | 3.2 | 191 | |
Washington | 9.3 | 4.8 | 194 | |
Rhode Island | 10.8 | 5.5 | 196 | |
California | 12 | 6.1 | 197 | |
Arizona | 9.4 | 4.7 | 200 | |
North Carolina | 10.1 | 5 | 202 | |
Georgia | 10.1 | 4.8 | 210 | |
New Jersey | 9.5 | 4.5 | 211 | |
Delaware | 8.1 | 3.8 | 213 | |
Montana | 7.7 | 3.6 | 214 | |
Nevada | 12.9 | 5.8 | 222 | |
Florida | 10.7 | 4.7 | 228 | |
Utah | 7.5 | 3.2 | 234 | |
Alabama | 10 | 4 | 250 | |
Idaho | 9.4 | 3 | 313 |
Sources: http://www.bls.gov/news.release/archives/laus_01182008.pdf, Table 3 (seasonally adjusted); http://www.bls.gov/news.release/pdf/laus.pdf, Table 3 (seasonally adjusted). Note that both figures are the preliminary ones from the initial press release for the month in question; Texas and Massachusetts have both been adjusted downward by 0.1 point for December 2007, for example, but I didn't want to download 51 spreadsheets to make this table. “Ratio 1” is simply the July 2011 rate divided by the December 2007 rate, expressed as a percentage.
As with Pandl's study, North Dakota and Alaska do very well, taking the top two spots. His other two strong performers, DC and Texas, fall slightly below the median (Virginia), however. Four of his five weak performers (California, Florida, Arizona, and Nevada) all do pretty poorly here, but so do states like Idaho (313% of December 2007 unemployment rate), Alabama (250%), Utah (234%), etc. His fifth poor performer, Michigan, actually does quite well (143%) on this metric, though it should definitely be discounted since the state has been losing population overall.
At this point, I don't have an explanation of why some states do better than others, since I don't have Pandl's dataset to run regressions myself. Energy works for the top three states, but not states like Louisiana, Texas, and Wyoming. I don't currently have the data on exposure to subprime mortgages or presence of high-end services (both of which Pandl found to be statistically significant) or for the many variables he did not find significant. But people with access to a lot of state-level data might want to take a look at this, or perhaps Pandl will do so.
The bottom line is that this view of employment performance undermines glib references to a Texas Miracle and challenges us to find an explanation of what really differentiates the states' employment performance. We should recall, too, that employment is not the only dimension of distress in our current economic situation (foreclosures immediately spring to mind), but it is an important one in its own right and contributes to the other major problems as well. Understanding why some states did better than others may give us some clues on what states should do differently, but we may find instead that a non-replicable factor like the energy industry remains statistically significant after controlling for other variables.
No comments:
Post a Comment