Chrysler Shakedown? Updates and Statistical Analysis

My last blog posting considered the possibility that the Obama administration may have targeted a disproportionate number of Republican-donating Chrysler dealerships for closure. Not having the time to do the original research myself, here are some links to several items of interest related to this story:

Directorblue

Washington Examiner

Gateway Pundit

Red State

Libertarian Republican

theblogprof

These alone (and there are more) have a wealth of data—and they're linking more all the time, it seems—that seems to demonstrate some statistically interesting conclusions.

Since most Americans (alas!) are woefully weak when it comes to statistics, I am going to try to focus on the key issues that must be statistically demonstrated to (a) exonerate the Obama administration, or (b) blow open the greatest Washington scandal since Watergate.

Here are some of the assertions that are beginning to be made:

1. A disproportionate share of the dealers on the closure list gave political contributions to Obama opponents and Republicans.

2. A disproportionate share of the dealers on the closure list are in "red state," GOP-leaning districts across the country.

3. The administration claims that the closed dealers were less profitable and/or losing money may in fact not be true.

Here is how they should be statistically analyzed:

1. Two key sets of data must be obtained: The percentages of dealers giving to each of the various political groups (Republican, Obama for president, Hillary Clinton, etc.; and those who did not give to any political groups) in 2008, and possibly previous years; and the percentages of each of those groups of dealers which were put on the closure list.

For example, it has been speculated that car dealers in general are Republican leaning. Let us say, then, that 60% of dealers give to the GOP [Note: Data used as examples is fictitious and only for the purpose of illustration]. If 75% of the closed dealers are GOP donors, it would then be relatively easy, once the data is in hand, to conclude if this is a statistically significant difference, and if so, to what level of confidence (90% confidence? 95%? 99%? 99.999%? I'm not joking about that last one, either). Confidence intervals, by the way, are quite interesting as a statistical topic.

The same can be done with the Obama donors, or with any other political group. The data could potentially indicate that there are no statistically signficant differences, in which case this whole story will, of necessity, quickly fade away. But just imagine a press release like this:
"We [insert statistician(s) name(s) here] conclude, with a 99.9% level of confidence, that there is a statistically significant difference between the political giving of dealers overall and those who were closed." The impeachment proceedings should begin within days.

2. This will be the easiest of the three, and the process mirrors #1 above. The two key sets of data this time are how many dealers are in "red" areas, and how many of the closed dealers are in "red" areas. [Note: Defining "red" geographically could be done at the state level, or at the congressional district or county levels. Here in Michigan, a blue state overall, the west side of the state tends to be red, while the east side is deep blue. No doubt the analysis will be done at state, district and county/local levels.]

Tests for statistically significant differences are then performed, as with #1.

3. This one will be the most challenging of the three to assert from a statistical viewpoint, as there will likely be various viewpoints on what constitutes "profitable," "growing," etc., combined with the fact that many dealerships are quite private about their financials. [Question: If the car companies are government owned, can a FOIA request unlock this data?] Furthermore, whatever statistics Chrysler may have used to comprise the list of dealer closures would need to be incorporated into the analysis.

Assuming the data becomes available in sufficient detail, the essence of the analysis will be a comparison—but there is more than one way to do it.

• Comparison of all dealers to closed dealers. I would be surprised—nay, shocked—if there was not a statistically significant difference here, but the magnitude of the difference will be informative.
  
• Comparison of closed dealers to dealers in the lowest 25% of profitability overall. It would also be interesting—and fairly easy once the data is known—to determine the degree of overlap in these groups. Which brings us to...
  
• Comparison of closed dealers to the lowest 25% profitability dealers who are allowed to remain open. This one could be fascinating, because if the closed dealers are not worse off from a profitability perspective than the lowest quarter of those who remained open, the stink will rise.

Whether or not the Obama administration more closely resembles the blindfolded Lady of Justice or the cast of the Godfather remains to be seen.

Godspeed, accuracy, and best wishes to all those data crunchers out there.