Nate Silver has a great post mapping out how each potential vice presidential pick affects Mitt Romney's chances of winning the electoral college. That had me wondering: have VP picks actually affected their tickets' performance in their home states? Silver has written a bit about this, and concluded that past running mates produced a gain of only 2.2 percent in their home states.

To figure that out, Silver looked at the difference between the ticket's performance in the veep's home state and their performance nationally. He then calculates that same gap for elections before and after the veep pick was on the ticket, finds the differences between the gap in the VP year and those years, and then average the differences together. I thought it might be more informative to see if the swing in veep picks' home states - that is, the difference in performance relative to the last election - can be explained solely by the swing in the national vote totals.

For example, the Obama-Biden ticket did better than Kerry-Edwards did in Delaware, but it also did better than Kerry-Edwards did nationwide. To see if Biden helped, the data point that's most useful is the difference between national swing from 2004 to 2008 and the state swing in Delaware between those elections. Looking at the data since 1948 - the cutoff is basically arbitrary, but World War II seems like as fair a point as any to start from - that gap is usually positive, suggesting that VP picks are helping their tickets in their home states:

The red line is Republican VP picks, the blue line Democratic picks. If a data point is above the grey dotted line, the swing was bigger in the VP pick's home state than nationally. The mean difference difference in a veep's home state is 5.09 percentage points; the median is 4.305. That is, a veep pick gains his running mate around 4-5 points. That's a bit bigger than Silver's estimate, but still not much.

What's more, that average is heavily weighted by a few weird elections. In 1952, Democratic running mate John Sparkman made big gains in his home state of Alabama, but only because the last time around the state was won by segregationist third-party candidate Strom Thurmond. But there are some real cases where a pick helped. LBJ won Texas for Kennedy in 1960, Geraldine Ferraro allowed Walter Mondale to do slightly less terribly in New York in 1984, and Dick Cheney gave George W. Bush an even huger margin in Wyoming in 2000. And there are real busts, like Jack Kemp, whose home state of New York saw an even stronger anti-Republican swing in 1996 than the rest of the country.

Another way of looking at this is to see how much of swings in veep picks' home states are explained by national swings. The national swings explain most of it - 63.14 percent - but still leave plenty of room for outliers, suggesting that veep picks can make a difference:

Click for a bigger version of the plot. Republican running mates are represented by red circles, Democrats by blue circles. The line is what you'd expect the state vote swing to be given the national vote swing. It has a slope of 1.0338, meaning the state vote swing is expected to be 3.38 percent bigger than the national vote swing - that is, a veep pick gets you 3.38 additional points in their home state. Points above the line outperformed this average, and points below it underperformed it. As you'd expect, LBJ in 1960 way outperforms expectations, and Jack Kemp in 1996 underperforms. So too, oddly, does Walter Mondale in 1976.

The bottom line: veep picks matter, and can swing their home states. But national trends still matter most.