Pennsylvania's 8-Year Coincidence?

I'm not particularly good at math, but one of the mathematical insights I understand is that, given a large enough sample, the unlikely becomes ...
| April 6, 2010

I'm not particularly good at math, but one of the mathematical insights I understand is that, given a large enough sample, the unlikely becomes likely. As in: It's unlikely that anyone in particular will win the lottery, but if enough people play the lottery it becomes likely that someone will win.

That's how I've always felt about Pennsylvania's 8-year revolving governorship. Since 1954, Democrats and Republicans have traded the governorship every eight years. Since 1970, this has meant 5 consecutive two-term governors, each with the opposite party affiliation of his predecessor. As a result, you get analysis like this from political scientists Terry Madonna and Michael Young:

The statistical evidence makes a compelling argument that the cycle is no fluke. Mathematical probability tells us that 14 elections beginning in 1954 are extremely unlikely to produce the alternating eight-year cycle simply by chance. The probability is less than 0.000141% that this string of 14 gubernatorial elections could have happened simply by coincidence. Put differently, the odds are more than 5000 to 1 against getting such an alternating string of election results, unless something meaningful has been occurring to produce the pattern. This is solid and persuasive statistical evidence.

But, there are 50 states. It seems quite likely that purely by chance some state would, over some period of time, have some quirky pattern in its gubernatorial election results, whether it was this one of some other equally strange one. If one American wins the Powerball jackpot at 195,249,054 to 1 odds, that doesn't prove "something meaningful has been occurring to produce" his series of correct numbers. He just guessed.

On the other hand, today Stateline (which is where I discovered the link above) offered the most plausible explanation of the pattern that I've seen:

Perhaps most alarming for Democrats, who have held the governorship for the last eight years under former Philadelphia mayor Ed Rendell, the streak has almost always favored the party that's not in power in Washington. Only in 1982, when Republican Dick Thornburgh was elected to a second term as governor while Ronald Reagan was president, did Pennsylvania voters choose not to separate their state and federal chief executives by party.

Have Pennsylvania voters really just been checking the power of the party in control of the presidency?

On its face, this theory makes some sense to me. For a long time, Pennsylvania's presidential voting has fairly closely mirrored the nation as a whole. In 1976, for example, Jimmy Carter won 50%-48% nationwide and 50%-48% in Pennsylvania. So, Pennsylvania is a good candidate for a state that would swing back and forth with the national mood or one where voters have a partisan ambivalence that inclines them to check the power of any one party.

On closer inspection, though, I'm less sure. The key years to look at, I think, are the ones where there was an open seat -- most incumbent governors win reelection regardless of the political environment or who is in the White House.

Democrat Ed Rendell was first elected in 2002. But, voters across the country clearly weren't trying to check President Bush's power that year. Republicans won 22 of 36 governor's races (and also scored well in congressional races).

If Pennsylvania really was checking Bush's power, its voters must have been thinking fundamentally differently than those elsewhere. That would be quite strange, since, as I just indicated, Pennsylvania has a notable tendency in presidential elections to vote like the country as a whole. I'm willing to bet that President Bush's approval rating was well above 50% in Pennsylvania in November 2002. Also worth noting: If Mark Schweiker runs for his own term in 2002, the 8-year cycle probably ends.

In other years, the theory makes more sense. A Republican was elected governor in 1994, when voters were interested in rebuking President Clinton. A Democrat won in 1986, which was a fairly good year for Democrats (although Republicans actually made notable gains in gubernatorial races in other states and there wasn't really that much Reagan fatigue). If memory serves, 1978 and 1970 were relatively neutral political years.

So, maybe voters' desire to check the power of the president's party explains some of the results, but probably not all of them. That leads me to a good point that Madonna and Young made:

It is, in fact, most likely that the eight-year cycle is caused not by a single factor, but by a multiplicity of factors. Collectively, they may explain the cycle, but none alone is sufficient to do so.

And, there's a good chance that the biggest factor was chance.

Josh Goodman  |  Former Staff Writer
mailbox@governing.com