When Counterfactuals Go Bad

Larry Sabato, Kyle Kondik and Geoffrey Skelley started off with a good idea: illustrate the regularities of Congressional and down-ballot elections with a counterfactual. 

But, oh, it just goes so horribly wrong.

Imagine, for a moment, if Sen. John McCain (R) had somehow won the presidency in 2008. How might the country be different?

See, there's the thing. John McCain was clobbered in 2008. We're not talking Bush/Gore, where we can easily imagine the outcome of the presidential race changing on its own; McCain lost by seven percentage points, and a 365-173 electoral college margin.

So Sabato (et al.) go on to spin this out as a "what happens next" thing, where Democrats do well in 2010 because there's a Republican in the White House...and wind up with a nonsense close:

But winning the presidency has its downsides, too — if not directly for the person who holds the office, then certainly for his or her party.

Now, explain to us again why so many senators, members of the House, governors and state legislators work hard to elect their party’s presidential nominee?

Okay, everyone -- spot the flaw? Yup:  Sabato is excluding the initial election from his analysis of the overall effects of winning the presidency. That is: Democrats going into 2008 had a 51-49 majority in the Senate. They're now at 54-46. Yes, they lost several seats in 2010, going from 59 to 53 -- but that was smaller than the landslide they had in 2008, going from 51 all the way up to 59!  (Plus Specter switching parties to make it 60, which doesn't happen if McCain is in the White House; minus Scott Brown, which also doesn't happen under President McCain).

If John McCain wins in 2008, it almost certainly means that Republicans win a whole bunch of Congressional seats they in fact lost. Yes, they would then have taken a hit in 2010 instead of winning a landslide...but we don't really know, overall, which would be better for them.

A few things...

You can imagine McCain doing seven points better and Republicans otherwise not benefiting, but that's not the world we live in; Democrats don't have a downballot landslide coupled with a Republican victory. Nor is "imagine McCain doing seven points better" really plausible if other events are held constant -- McCain and the Republicans lost badly in 2008 because of the economy (and other things that made George W. Bush and the GOP very unpopular). It's possible to imagine the close ones changing -- 1960, 1968, 1976, 2000 -- but 2008 wasn't going to be a good Republican year unless there was a very different 2008.

And again, if there's a 2008 that's different enough that McCain wins, then we're also talking about a very different 2010. Sure, if all you know is "Republican in the White House," it's safe to predict that, all things equal, Democrats gain seats.

But it's not safe to predict which side wins Congress if all you know is the presidential party in a midterm. After all, Democrats held both Houses of Congress in 1962, 1966, and 1978; Republicans did the same in 2002, and retained the Senate in 1982. Nor is it safe to assume that the president's party is the loser over the whole cycle, including the election in which the president wins. For example: Republicans lost control of the House in 1954, the first Eisenhower midterm...which gave them 2003 seats, or four more than they had before Ike was first elected.

And while some of this is about blame being given to the party in the White House when things to wrong, another significant part of it is just opportunity. Presidents who bring fewer Members of Congress in with them (JFK, for example) tend to have relatively good midterms; one of the reasons that Obama's midterm losses were so severe, in terms of number of seats lost, was that 2008 (and 2006) had been so good for the Democrats, leaving lots of easy targets for the out-party.

Again: the main point here, which is that midterm elections are bad for the party of the president, is perfectly fine. But this counterfactual example of it is more likely to misinform than to illustrate it.