Analyzing the Survival of Party Chairmen: "It's Not You, It's Us"

As you know, Michael Steele was deposed as chairman of the RNC a few short weeks ago in favor of the whimsically-named Reince Priebus.  In addition to his numerous public gaffes, Steele left the RNC in serious debt after the 2010 mid-term election.  Of course, Steele would counter that debt is standard for both the RNC and DNC after an election.  But Seth Masket asks the more relevant question: “what's the value in not spending absolutely everything during an election cycle? It's not like Republicans have nothing to show for it.”  Well said.  Shouldn’t the net gain of 63 House seats trump a few million dollars of debt and a handful of public gaffes?  This got me thinking (a couple weeks ago) about the tenure of party chairmen.  So yesterday, with some spare time, I collected data on the tenure of every RNC and DNC chairman since 1856 (and by “collected” I mean obtained it from Wikipedia). One thing I want to make clear about the data: the “tenure” of a party chairman blends instances in which he or she retired with those, like Steele, where he or she lost their reelection bid.  On the one hand, we would expect chairmen to retire or lose the nomination when an election goes poorly for the party and run for another term (and win) when an election goes well for the party.  Still, there are clearly a number of chairmen who retired during good times.  For example, Howard Dean stepped down in 2008 after Obama won the presidential election and Democrats added to their House and Senate majorities.  Indeed, winning presidents typically handpick the party’s next chairman.  In short, it might be worth another look to isolate those instances where the chairman actually lost a reelection bid, as Steele did.  Nonetheless, here some are some results from the tenure data I collected.

On average, DNC chairman serve for about 3.6 years while RNC chairman serve for about 2.5 years.  Though I knew chairmen were somewhat ephemeral creatures, I was surprised at how short their lifecycles really are (especially on the Republican side).  To visualize the tenure of DNC and RNC chairmen, the following two charts (using event history analysis) plot the cumulative likelihood of failure and the smoothed hazard function over tenure time.  The left  chart (failure) shows that Democrats have about a 50% likelihood of leaving their party’s chairman post at 3 years while Republicans have about a 75% chance of leaving.  (Democrats in orange and Republicans in green, the apparent default in Stata 9).  In the right chart (hazard function) we can see that, for Republicans, there is a sharp increase in the risk of failure between years 2 and 4.  For Democrats, the risk of failure exhibits less variability over time—replacement is greatest around year 4, but Democrats experience less risk of replacement or retirement.  All of this is nothing spectacular, we know that party chairmen serve short tenures.  Still, it’s perhaps interesting to visualize the trends and observe the difference between the two parties.

Next, to assess the question posed by Seth Masket, I coded data on the number of seats picked up in the House by each party since 1970.  We can’t use the corresponding Senate data for obvious reasons—with only 1/3rd of senators up for reelection we get a biased sample of the party’s electoral performance.  I ran a series survival analyses (Cox proportional hazards) on the risk of replacement using the seat pickup data as the key independent variable.  I also ran logit models on a dummy variable indicating whether the chairman survived after an election.  I modeled both the raw number of seats picked up during the prior national election cycle and the election-to-election first difference.  I estimated combined models (RNC and DNC) as well as separate models for each party.  The results—absolutely nothing.  There is no evidence that the party’s performance in an election has any effect on the chairman’s likelihood of continuing in his or her job.  For Democrats, the effect approaches statistical significance in a few of the models (by which I mean the associated p-values are in the mid to low teens); but this is still too far to draw any meaningful conclusions.  So, yeah, file this one under “unspectacular.”  Still, as I tell my research methods students, sometimes null results are interesting.

All in all this is somewhat unsurprising stuff.  More goes into the job performance and tenure of a party chairman than the raw number of seats his or her party wins.  Namely: (1) the president can install a party chairman after a victory and (2) I suspect there is a norm that chairman only serve a few terms.  So the old relationship adage applies to Michael Steele—“it’s not you, it’s us.”