Patterico's Pontifications

5/31/2009

Dealergate and Clinton Donors

Filed under: Obama — DRJ @ 11:40 am

[Guest post by DRJ]

Via Michelle Malkin, Marla Singer at Zero Hedge found “… a significant and highly positive correlation between dealer survival and Clinton donors.”

She wonders:

“Did Maureen White pass a list of Clinton donors to her husband for preferential treatment in the Chrysler dealer beheddings game? We don’t know. “Maybe” is about the best answer we can give with this data. But “Maybe” should never have been an option anyone even had to wonder about.”

Singer cautions these are preliminary findings but also states she does not expect them to substantially change in the final version.

The story, I’m sure, will continuue.

EDIT: More on minority dealer closings here.

— DRJ

55 Responses to “Dealergate and Clinton Donors”

  1. Story? What story?

    Diffus (6b5636)

  2. 80% of the dealers closed made zero political donations.

    Just a few days ago, the same folks were claiming that Obama’s apparatchiks were closing the dealers that did donate to Hillary out of revenge.

    Stay away from this story if you want to maintain your credibility.

    It’s Jamil Hussein all over again.

    poon (093c46)

  3. I thought the Jamil Hussein story was very interesting. Still do.

    DRJ (2901e6)

  4. Blogger runs regression analysis and finds nothing statistically significant. Film at 11.

    Internet Tough Guy (22501b)

  5. I take it you didn’t notice this Zero Hedge chart, ITG.

    DRJ (2901e6)

  6. DRJ, I saw the chart. I also read this following the chart:

    Granted, that P-Value (0.125) isn’t enough to reject the null hypothesis at 95% confidence intervals (our null hypothesis being that the effect is due to random chance), but a 12.5% chance of a Type I error in rejecting a null hypothesis (false rejection of a true hypothesis) is at least eyebrow raising. Most statistians would not call this a “find” as 95% confidence intervals are the gold standard for this sort of work.

    It reminded me a bit of the pirates vs. global warming chart. I can run regressions all day in SPSS that show all sorts of “interesting” things given a sufficiently low CI.

    Internet Tough Guy (22501b)

  7. IIRC, you guys almost got Jamail Hussein killed.

    Very interesting for him, I guess.

    As for the chart, 3000+ Chrysler dealerships, a vast majority of them owned by people who made zero political donations over the past 18 years.

    Only by tossing out these apolitical owners does the sparse data look sort of(not really) interesting.

    It’s like only looking at the soldiers who served at Abu Ghraib and then judging every soldier in the U.S. Army by their behavior.

    poon (093c46)

  8. ITG,

    So you treat evidence of some correlation as equal to evidence of no correlation?

    DRJ (2901e6)

  9. DRJ, I treat statistics as what they are. I’m a fan of Zero Hedge, but that post is weaksauce. The poster even points out that there is no statistical significance to the analysis.

    Internet Tough Guy (22501b)

  10. Keep ridin’ this pony. It’ll take you where you need to go.

    Larry Reilly (45e7a4)

  11. ITG,

    The post says the preliminary report shows a “significant and highly positive correlation” between dealer survival and Clinton donors. How do you translate that into “no statistical significance”?

    DRJ (2901e6)

  12. #9 – Please reread what Zero Hedge said. “not enough to reject at the 95% level of confidence”. Not “no correlation”. Big difference.

    The data show a correlation at the 87.5% LOC. Not “gold standard” proof, but it seems more than just random chance.

    If you want the 95% LOC, you want a carefully designed study, with independent variables (orthogonal variables) and repeated experiments to confirm statistical relevance. Also, data analysis can be degraded by outside influences (i.e., uncontrolled factors) that may be uncodable.

    Having done tons of study on engineering data after the fact (i.e., not controlled experiments), if you get anything over 80% LOC, you tend to believe there is something there.

    Of course, better methods for analyzing non-orthogonal data sets would be better, but few people understand the methods, and fewer know how to interpret the data. If you don’t believe me, look up “partial least squares” methods.

    So just because you don’t get precise results from a “clean” data set does not mean that something strange was not happening.

    So if you are so convinced that there is nothing to see, what’s wrong with FOIAing the methodology used to decide who remains open and who gets early retirement?

    Dr. K (0fdffa)

  13. DRJ, just because the poster said there’s a “significant and highly positive correlation” does not mean there is a statistically significant correlation. Statistics is a well defined field where words mean things. The OP you cite is using a linear regression instead of an analysis of variance, so I’m going to call foul on that alone, based on the type of data she is trying to analyze. Further, the OP tries to wash away the low CI by saying a 95% CI is “totally arbitrary.” Maybe on blogs, but not by anyone that uses stats for actual work.

    This analysis wouldn’t get a passing grade in a classroom and would likely get you fired in a professional setting. If the OP is a quant, I’d personally stay far, far away from any investment advice.

    Internet Tough Guy (22501b)

  14. So if you are so convinced that there is nothing to see, what’s wrong with FOIAing the methodology used to decide who remains open and who gets early retirement?

    The money quote from the linked “weaksauce” post:

    Did Maureen White pass a list of Clinton donors to her husband for preferential treatment in the Chrysler dealer beheddings game? We don’t know. “Maybe” is about the best answer we can give with this data. But “Maybe” should never have been an option anyone even had to wonder about.

    Obama promised transparency.

    Where is it?

    Paul (creator of "Staunch Brayer") (290fb9)

  15. Dr. K, I’ll join any calls for FOIAs on information surrounding federal intervention or influence in Chrysler’s decisions. Since you are familiar with statistics you should notice that the OP is using multiple categorical variables in a linear regression analysis. See any problems yet? Then we could get into chi-squared, fisher’s exact, etc.

    Internet Tough Guy (22501b)

  16. #13 – The OP you cite is using a linear regression instead of an analysis of variance, so I’m going to call foul on that alone, based on the type of data she is trying to analyze

    ITG – I call BULLSHIT. Regression analysis IS analysis of variance.

    Where do you think the measures of significance come from? They are ratios of the sums of squares that are determined from an analysis of variance.

    But what the hell do I know? I have only been using statistics in my profession (and I’m not a professional statistician) for over 20 years.

    Dr. K (0fdffa)

  17. ITG,

    We evaluate events by questioning what happened, and the progression here seems consistent with the way scientific inquiries typically develop. For instance, initial questions were based on anecdotal observations about the terminated dealers and that led bloggers to submit various hypotheses.

    From there, the inquiry shifted to surveys of all terminated dealers and comparisons to their political donations. Others looked at dealers that weren’t terminated, while Zero Hedge began a more sophisticated regression analysis. Isn’t testing of hypotheses the next step in the scientific method?

    I’m sure there are further steps that will be taken but these seem like worthwhile first steps given the short amount of time that has elapsed.

    DRJ (2901e6)

  18. Hillary donors: 3/54 dealerships shut down = 5.55%
    Obama donors: 1/69 dealerships shut down = 1.44%

    My super duper statistical analysis fails to find the Hillary bias.

    poon (093c46)

  19. #15 – It surely is possible to treat the data that way correctly.

    What do you think anova of 2-factor experiment is? It’s a linear regression of a response between 2 categorical variables (High/Low, D/R, etc.).

    Now before you all get too bent out of shape, if you are trying to fit a straight line to a model that is better described as a parabola (still a linear model, BTW – see below), guess what?

    I mean we are talking:

    y = A + Bx versus
    y = A + Bx + Cx^2

    Both of these are linear in the coefficients (which is why it’s called a linear regression).

    You get lower correlation and the linear coefficient (B) is not as significant as the quadratic coefficient (C) if the model is truly a quadratic.

    By using 2 categories, the best you can model is the y = A + Bx – this is still a linear regression. And it could mean that the level of significance is lower than the “gold standard”, but there is still a relationship.

    Dr. K (0fdffa)

  20. #18 – poon (if that is truly your real name), compare the Hillary supporters versus the Republican supporters as a total of the closures.

    Now tell me that it was truly random.

    Dr. K (0fdffa)

  21. Poon is not to be engaged in at any level – it just leaves droppings at various times of the day here.

    Dmac (1ddf7e)

  22. Dr. K,

    Of the Chrysler dealers that did make a political donation, they donated at a 10:1 ratio Republicans:Democrats.

    You could make a far better case that not making any kind of political donation was the leading cause of dealership closure.

    Revenge of the political class as a whole?!?!

    Nah, not sexy enough.

    poon (093c46)

  23. DRJ, I welcome all investigation into potential government wrong doings, and in no way do I want to disparage statistics. Stats is a field that I used to be involved in and it’s fascinating (to me). I just want to make sure that people don’t read too much into the OP’s analysis. By all means, more people should learn how to use SPSS and R.

    Dr. K, it’s possible to treat the data that way correctly, but the OP didn’t do that. If you gave me information on the American League or National League preferences of dealership owners, I might be able to show a “significant and highly positive correlation” between that and closings. It doesn’t mean there’s an MLB conspiracy involved.

    Internet Tough Guy (22501b)

  24. Some of you couldn’t figure your way out of a wet paper bag. It is not about the overall % of Republican/Democrat dealers closed. It is whether, in any given local area where the dealerships are close enough to pick up customers or otherwise benefit from a closed dealership, there was a preference to a donor or supporter of the Democrats.
    If an objective review of the closings finds that it was done fairly, then that is how it should have been–and that should be announced to all.
    But, if those with political “juice” were allowed to stay in business over others, then that is a different matter entirely.
    Only a close look at each local area will resolve this. Stats will come from that, not the other way around.

    ventu (7bc469)

  25. “where the dealerships are close enough to pick up customers or otherwise benefit from a closed dealership”

    This was goalpost move #3

    #1 Dealerships that made donations to Republicans were diproprtionately closed..oops, nope.

    #2 Dealerships that gave to Obama’s Democratic rivals were closed…oops, nope

    #4 Dealerships that gave to Obama’s Democratic rivals were kept open…

    Turns out the dealerships that were shut down are being sold to better businesspeople, so “shutting down” a dealership actually hurts its nearest competitors.

    Next goalpost move in 3…2…1

    poon (093c46)

  26. “Turns out the dealerships that were shut down are being sold to better businesspeople, so “shutting down” a dealership actually hurts its nearest competitors”

    poon – Do you have any proof for that assertion? Please show your work.

    daleyrocks (5d22c0)

  27. Daley, poon has yet to show anything to back up any assertion it made on this thread.

    Paul (creator of "Staunch Brayer") (290fb9)

  28. Lithia, which operates 90 dealerships in 13 states, had two of its 29 Chrysler dealerships on the cut list — one in Colorado Springs, Colo., and one in Omaha, Neb. DeBoer said Lithia could have lost all 29 of its Chrysler stores. Instead it expects to see a net gain because it plans to buy five Jeep or Chrysler franchises on the cut list and add them to existing stores.

    http://www.registerguard.com/csp/cms/sites/web/business/13699048-41/story.csp

    poon (093c46)

  29. Paul – Or any other thread that I recall. poon seems very fond of asspulls. Perhaps it should consider a more appropriate screen name.

    daleyrocks (5d22c0)

  30. poon – Lithia is a dealer who supports Dems, correct?

    daleyrocks (5d22c0)

  31. Nate Silver (who deals with stats on a daily basis) does a good take down of the OP’s analysis.

    Internet Tough Guy (22501b)

  32. poon – Lithia is a dealer who supports Dems, correct?

    Amazing how poon left that little detail out.

    Paul (creator of "Staunch Brayer") (290fb9)

  33. Yeah, Paul,

    A dealership that supports Democrats…had two of its dealerships closed!

    Gee Wilikers!

    Get Rush, Newt and Karl on the batphone!

    *Resists urge to point and laugh as the faux scandal unravels*

    poon (093c46)

  34. ITG – Silver’s most recent post on the subject is as misleading as his first effort, merely quoting Marla’s first sentence about a 12.5% P factor, but ignoring the rest of her paragraph acknowledging why most statisticians would not consider that statistically significant at the 95% level. Dishonest on Silver’s part due to selective quoting – Absolutely.

    daleyrocks (5d22c0)

  35. poon,

    As this Bloomberg link notes, Lithia is expected to gain 5 dealerships and lose 2 as a result of the Chrysler terminations:

    “It’s a net plus for us,” Sid DeBoer, chairman and chief executive officer, said in an interview today. “The good news for Lithia is that we have a manufacturing partner that will be stronger.”

    That’s good news for Lithia since it lost $5.1M in the last quarter of 2007 and had continued losses in 2008 despite selling several stores.

    DRJ (2901e6)

  36. daleyrocks, the OP has a CI of 87.5% and admits that “Most statistians would not call this a “find” as 95% confidence intervals are the gold standard for this sort of work.” This is in the same paragraph she says “Why would there be a significant and highly positive correlation between dealer survival and Clinton donors?” Those two statements seem to be at odds, especially since “significant” has a specific meaning in the world of statistics.

    Did the OP find any correlation between dealer closings and Huckabee donors? Answer that and you’ll know why this analysis is bunk.

    Internet Tough Guy (22501b)

  37. DRJ,

    Same story as mine. Lithia might buy 5 dealerships from the list.

    Might.

    Lithia might also just lose two dealerships.

    You’d think the biggest car dealer donor to Hillary Clinton’s Presidential campaign would have done a little better than losing two dealerships, eh?

    According to goalpost #4, that is.

    poon (093c46)

  38. Yeah, and I wonder why when confronted with facts, ITG decides it isn’t worth his time to engage my answers to his hueristic “argument”.

    I am specifically referring to the issue that linear regression IS analysis of variance.

    Or the fact that this was NOT a controlled experiment, so the 95% CI is arbitrary. Especially when one realizes that 95% was arrived at over a century ago when some people at a party decided that 95% was just too darn strong for coincidence.

    Rgularly, serious data analysts use CIs of less than 95%, especially when there are no replicates (i.e., repeated experiments) or when there is some question of uncontrolled variables that skew traditional analysis.

    But, ITG uses statistics as a drunkard uses lampposts – for support rather than illumination.

    Dr. K (0fdffa)

  39. If the drive-by’s are this invested in proclaiming that there is nothing to se, then it prolly bears looking at. Or Barcky could release the criteria he and Gattner used to decide which contracts to terminate, and Barcky could quote to us the part of the Constitution gives him the authority to do this in the first place. Absurd, I know.

    JD (870a39)

  40. I get a giggle about how Internet Tough Guy presents his own fantasies in such a way as to imply they are factual, e.g.: Did the OP find any correlation between dealer closings and Huckabee donors? Answer that and you’ll know why this analysis is bunk.

    Uh, this is rhetorical bunk. If other correlations occur with similar confidence, then show them. Pretending they exist is just dishonest.

    SPQR (26be8b)

  41. ITG @36 – I believe Marla chose not to use the words “statistically significant” together for a reason, because used together, those words do have meaning to knowledgeable observers. Nice try.

    daleyrocks (5d22c0)

  42. poon – Please feel free to present your proof that the owners of the Lithia Group are superior business people to those whose dealerships are being closed at any time. Please show your work. Circular reasoning does not constitute proof.

    daleyrocks (5d22c0)

  43. Dr. K, I’m unsure about what you want me to address. The analysis is a mess. Maybe there’s something to the Clenis speculation, but there’s nothing statistically significant. The OP’s use of the word “significant” while discussing statistical correlation is a bit disconcerting, but feel free to dig into the “significant and highly positive correlation” more. Maybe you’ll find Vince Foster’s real killer while you’re there.

    SPQR, I don’t have the data set from the OP, so I can’t run the analysis. I would guess that I could come up with all sort of interesting correlations if I had the data set. Finding nonsensical correlations with a CI > 95% used to be a game in college.

    Internet Tough Guy (22501b)

  44. Daley,

    It’s the Phony Outrage of the Week Brigade that’s pushing this latest non-story.

    It should have died at its first wrong iteration (Dealerships were being shut down because they donated to Republicans).

    In this fifth or so(wrong and absurd on the surface) iteration, I’d say the burden of proof is on the folks manufacturing it.

    Thee damage is done.

    Rush repeated it on the air.

    Everyone involved in it is a little less credible.

    And the Republicans have wasted another week or so.

    17 months to the next election.

    What say we keep the ball rolling and move the goalposts again and say the dealerships were shut down based on…I don’t know…how about the religion of their owners?

    poon (093c46)

  45. And that’s what you are doing, Internet Tough Guy, presenting your guesses as evidence.

    They aren’t.

    SPQR (26be8b)

  46. SPQR, the OP asked a question, saw no correlation existed, eventually found something that was interesting, but not statistically significant, and created a story to fit the result. Go ahead and run with that one.

    Internet Tough Guy (22501b)

  47. #43 – You should acknowledge the following points about the analysis:

    1) Regression analysis is appropriate for 2 level coded variables because it is in fact an analysis of variance.

    2) This was not a designed experiment with adequate controls to ensure outside influences. As such, confidence intervals can be expanded. While a 95% CI is the “gold standard” it is not necessarily applicable in this case.

    3) If you look closely at the first table presented, the p=0.875 for the “Clinton” regression is the one that has the highest potential significance. The next nearest is the “Obama” regression that has p=0.509; far too low to even be considered. Nobody can agrue that any other data even shows the hint of a trend.

    4) There are other, more arcane analytical analyses that remove the cocorrelation between supposedly independent variables. These were not done, but could prove interesting.

    My guess is that because this is not a before-the-fact designed experiment, but an analysis of data collected, there are uncontrolled factors that may be skewing the results.

    For example, if a multiple linear regression using other data such as volume of sales in units volume of sales in dollars, population in a 10-mile radius of the dealership (or in the county), distance to nearest dealer, and so forth, we could see some interesting patterns.

    However, all we have is the anomoly of some dealers being favored over others and that Chrysler management did not make the decisions on who stayed and who went.

    Since there are questions, the task force should show their methodology. There are two reasons for this:

    1) Transparency was promised, so be transparent.

    2) The methodology could be tested by running the numbers and see how close to “ideal” the closing list adheres to the supposed methodology. Of course, reasons for deviations should also be publicized.

    Dr. K (0fdffa)

  48. Say, does poon’s writing style seem familiar?

    It sure does to me. Let’s see:

    Comments constructed of of one sentence paragraphs, statements, assertions and projections without any visible means of support, mocking tone throughout…

    Hmnn…I just can’t shake the feeling poon once clamimed to be a “Staunch Republican”…under a different screen name.

    I can’t shake the feeling that poon is yet another banned commenter that just can’t quit us…a certain commenter that has returned several times under different screen names…a certain one that inspired my own screen name.

    JD, daley, AD, nk, Scott Jacobs, SPQR: what say you?

    Paul (creator of "Staunch Brayer") (290fb9)

  49. That analysis needs to be refined. The safe dealers need to be tossed out.

    But what is presented means very little right now. More to come.

    HeavenSent (1e97ff)

  50. true believers, enjoy some disconfirmation..

    http://www.macombdaily.com/articles/2009/05/30/news/srv0000005465630.txt

    Conservative bloggers insist that dealer owners who contributed to Republican candidates were disproportionately targeted for closure, with the White House Auto Task Force calling the shots. But a review of Federal Election Commission records reveals a different story.

    In Macomb County, three dealerships were closed in the Chrysler bankruptcy process and seven were saved. The FEC records show no noticeable difference in the political leanings of the owners of the 10 businesses.

    Anthony Viviano, president of the Metro Detroit Dodge dealer’s association and owner of Sterling Heights Dodge, which will remain in operation, said politics was never a factor in the process.

    “That’s just a bunch of baloney,” said Viviano, former president of the Detroit Area Dealers Association.

    According to the FEC, Viviano donated $4,150 to federal candidates over the past 10 years, with most of the money flowing to former Republican senator Spence Abraham and GOP Rep. Candice Miller. The only deviation was a $500 contribution to Democrat Carl Marlinga’s 2002 bid for Congress.

    As with every other Chrysler dealer in Macomb, Viviano did not donate to President Obama or any other 2008 presidential candidates.

    The FEC records show a consistent pattern. The surviving dealers donated to Republican Michael Bouchard’s 2006 Senate campaign, the National Republican Congressional Committee, the Republican National Committee, the Michigan Republican Party, Abraham and Miller.

    One exception was Carl Galeana, who has a bipartisan track record. Over the past 10 years, Galeana, vice president of Galeana Automotive Group, has contributed a combined $16,700 to Abraham, Bouchard, Miller, Marlinga, Democratic Reps. Gary Peters and Mark Schauer and the Democratic Senatorial Campaign Committee.

    Galeana’s mixed record of political assistance has, presumably coincidentally, matched the mixed results he has experienced during the Chrysler and General Motors march toward bank

    ruptcy protection. Galeana’s Van Dyke Dodge in Warren survived, but the future of his other dealerships — Saturn of Warren and Saturn of Lakeside in Macomb Township — remains in limbo.

    Two Macomb dealers — Mike Riehl of Roseville Chrysler Jeep and Robert Brent of Orchard Chrysler Dodge Jeep in Washington Township — made no federal contributions over the past decade. Both stayed off the casualty list.

    A Wednesday editorial by The Washington Examiner indicated that a number of Chrysler dealerships across the nation slated for closure are owned by people who have donated to Republican candidates or Democrats who opposed President Obama in the 2008 primaries, such as Hillary Clinton. Bloggers quickly latched onto the report and talk-radio commentators followed.

    But it appears the flaw in the premise is that most dealer owners prefer GOP candidates, not just those owners who are targeted by Chrysler’s 25 percent reduction in dealerships.

    According to the Washington-based Center for Responsive Politics, the National Auto Dealers Association membership gives 69 percent of its contributions to Republicans. Some calculations indicate that the tilt toward the GOP is even more pronounced among all auto dealers.

    White House press secretary Robert Gibbs said Thursday that the list of dealerships marked for extinction was crafted by Chrysler executives, not the Auto Task Force.

    Viviano, a 43-year veteran of the auto business, has a keen sense of the anger expressed by many dealers who will soon shut their doors. Meadowbrook Dodge in Rochester Hills, in which he holds a majority stake, in partnership with his two sons, will be shutting down.

    Though the business was highly profitable and ranks in the top 20 nationally for sales volume among Chrysler dealers, Viviano told his sons, Sal and Nino, that the property did not fit the automaker’s plan to favor dealerships that have room for expansion and upgrades.

    “There was a reason for this. They are sitting on just 4.6 acres. There’s nowhere to expand,” he said. “We (at Sterling Heights Dodge) are sitting on 20 acres and we have two huge showrooms. Our capacity is four times the size of (nearby competitors).

    ad hominem responses in 3, 2, 1..

    glasnost (f9f106)

  51. true believers, enjoy some disconfirmation..

    …which was blown up by the very first commenter at your link.

    http://www.investors.com/NewsAndAnalysis/Article.aspx?id=478099

    Paul (creator of "Staunch Brayer") (290fb9)

  52. Am I getting this right, a liberal poster at this site is complaining about moving goal posts? Must resent someone going after their monopoly.

    Have Blue (854a6e)

  53. “Anthony Viviano, president of the Metro Detroit Dodge dealer’s association and owner of Sterling Heights Dodge, which will remain in operation, said politics was never a factor in the process.”

    glasnost – Since Viviano is such an authority on “the process”, why doesn’t he tell everybody how it was really done?

    The article doesn’t say what you think it says. Thanks.

    daleyrocks (5d22c0)

  54. John Lott looked at the numbers, and found no there there.

    Xrlq (62cad4)

  55. […] Patterico’s Pontifications » Dealergate and Clinton Donors. […]

    Dealergate and Clinton Donors | JoeWebb.com (66cbdf)


Powered by WordPress.

Page loaded in: 0.4198 secs.