Political Sex Scandals, part I

Posted: July 24, 2011 in analysis

Over the years, many nationally prominent politicians have faced sex scandals.  Some like Anthony Weiner, have resigned.  Others, like David Vitter, have remained in office.  Which stay and which do not seems random.  In this article, I will attempt to build a statistical model that uses the features of a sex scandal to predict whether or not it will cause the principal to leave their job.

Unlike my past analyses, this is not sparked by a specific statement or a quotation; the question was suggested by a family member and I am just interested in the phenomenon.

This is a work in progress, however.  I wasn’t originally ready to publish this, and some of the details may be subject to change, but the breaking news this weekend of a sex scandal involving David Wu made me want to get my model working in time to make a prediction about his fate.

First, a few notes to the more statistically minded.  I have some pretty good statistical training, but it is a bit unusual in that it is model based, not test based.  So, I will be approaching this problem by attempting to construct a model that explains the variance among the data points, and which can be used to predict an outcome given a set of variables.  I will attempt to find the “best” model that does this.  This means a model with terms of some significance, but I am not a slave to p<0.05.  I will attempt to avoid the multiple comparisons problem and overfitting the model, but nothing I find (or don’t find) should be taken to be “statistically proven” in any fashion.

My question is primarily about the nature of the scandal itself.  That is to say, I hope to discover if certain aspects of the scandal are predictive of the outcome.  Is it how kinky the sex is?  Whether you cheated on someone?  What party you belong to?  Or something more complicated and subtle?  I am, for the most part, not looking at external political forces that may be at work behind the scenes.  That is, other than in very simple ways, I am not trying to measure how politically vulnerable a person is to determine if that was what led to their demise (or survival).  Most of these effects will turn up in my model as external variance that it does not explain; if there is a lot, it means that these effects (or the fickleness of fate) play a larger role than the inherent nature of the scandalous acts.

In order to make any calculations at all, we need a data set.  I chose to consider US Congresspeople, US Presidents, and state governors, and credible candidates for the above, who were involved in some sort of sex scandal with sufficient news coverage for me to research it, while in office or running for office within the last 30-35 years.  I generated a list of 36 people who met these criteria.

Some of the variables to measure about the subjects are simple.  I recorded the party of the person, the year of the events, their seniority in their current position, and the outcome of the scandal.  In simple terms, “won election”, “won re-election”, and “term ended” are positive results, and “resigned”, “lost re-election”, “did not run for re-election”, and “dropped out of race” are negative results, but in a more advanced version of the model, each result gets a different score, with “resigned” at the bottom and “won re-election” at the top.

The rest of the variables were more difficult to measure.  I chose some basic descriptors of sex scandals and attempted to judge, on a scale of 1 to 10, where each scandal fell on it.  Which descriptors I used was difficult and somewhat arbitrary to choose, and sometimes deciding where a particular scandal fell was difficult.  The variables I chose (with example scandals to show extreme cases) were:

  • Intensity: The degree or duration of the claimed act.  Mark Sanford gets a 10 here, since he had a long-term affair and found his soul mate.  Eric Massa gets a 3, since all he did was tickle some people inappropriately and tell some dirty jokes.
  • Unfaithfulness: Did the person have a duty to someone that they violated with their acts?  Most married people score about 8 here, but Anthony Weiner, with a new marriage to a pregnant wife, gets a 10.  Several unmarried people get a 1.
  • Kinkiness: Is the act described in itself unusual, odd, quirky, distasteful, or somehow outside mainstream perceptions of normality?  Gay sexual activity rates higher here, not in a personal judgement, but as a representation of mainstream opinion.  Jon Hinson rates an 8 for gay oral sex in the bathroom of the House of Representatives.  Many scandals in which the details of the sexual behavior were never discussed are scored as 1s.
  • Coercion: Were the acts coercive, either in a power dynamic and subtle way, or as outright sexual harassment/assault?  All affairs with staffers, interns, or pages score a little here, since its hard to separate the power dynamic from the sexual relationship.  Gus Savage gets a 10, since his case involves violent rape, and many vanilla affair scandals get a 1.
  • Hypocrisy: Does the scandal reflect a disconnection between the person’s public positions and actions and their private choices?  Eliot Spitzer gets an 8 here, since he was well known as the New York Attorney General for prosecuting prostitution rings.  Many people get a 2 or 3 for extramarital affairs that contrast with their general “family values” image.
  • Plausibility: All the other ratings have been about the claimed events.  This one is how likely they are to have happened substantially as described.  Most of these are rather high, due to full or partial confessions.  This is likely due to selection bias; the unconfirmed and denied cases of 20 years ago are forgotten.  The lowest score here is Nikki Haley, who adamantly denies an affair with Will Folks, who has presented circumstantial evidence only.  This rates a 4.
  • Contrition: Did the person admit responsibility, ask forgiveness, or confess, as part of their efforts to keep their job?  How quickly did they do so, and how completely.  I am only scoring statements here that were made before any inkling that the person was planning to resign or not run for re-election; contrition after making that decision isn’t relevant to modeling the decision.  In many cases, the scandal broke so quickly there were no statements at all.  A half dozen people are rated 1 for denying the claims, blaming political opponents, and so on.
  • Misappropriations: Did the person do anything else wrong, either as part of or in response to the events of the scandal?  This covers secondary scandals that are spinoffs of the primary sex scandal, usually (but not always) financial in nature.  John Ensign, with his strange payoff, scores an 8 here.  Bill Clinton scores a 7 for his creative interpretations of the truth while under oath.  Half of the set has no hint of this, and scores a 1.

In the next section, I’ll take these gathered variables and try to build a functional model out of them.  Wu can serve as a good test case, since we can check what the model says he will do.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s