Monday, September 08, 2008

The History of Bounce

Once again, I went trekking to one of my favorite archives, the records at Gallup for poll results during Presidential elections. There were no pre-convention poll results available for 1936 and 1940, so I had to begin with 1944. But there are some interesting numbers to look at, in terms of what value a post-convention “bounce” has for a candidate. The polls are for pre-both conventions, and after both conventions.

1944: Pre-conventions FDR 47, Dewey 45
Post-conventions FDR 47, Dewey 45
[ no bounce ]
Election result FDR 53.4%, Dewey 45.9%

1948: Pre-Conventions Dewey 49, Truman 38
Post-conventions Dewey 48, Truman 37
[ no bounce ]
Election result Truman 49.6%, Dewey 45.1%

1952: Pre-Conventions Eisenhower 59, Stevenson 31
Post-Conventions Eisenhower 50, Stevenson 43
[ Stevenson 21 point bounce ]
Election result Eisenhower 55.2%, Stevenson 44.3%

1956: Pre-Conventions Eisenhower 61, Stevenson 37
Post-Conventions Eisenhower 54, Stevenson 41
[ Stevenson 11 point bounce ]
Election result Eisenhower 57.4%, Stevenson 41.2%

1960: Pre-Conventions JFK 50, Nixon 44
Post-Conventions JFK 47, Nixon 47
[ Nixon 6 point bounce ]
Election result JFK 49.7%, Nixon 49.6%

1964: Pre-Conventions Johnson 76, Goldwater 20 (yes, 20 percent support)
Post-Conventions Johnson 65, Goldwater 32
[ Goldwater 23 point bounce ]
Election result Johnson 61.0%, Goldwater 38.5%

1968: Pre-Conventions Nixon 40, Humphrey 38
Post-Conventions Nixon 43, Humphrey 28
[ Nixon 13 point bounce ]
Election result Nixon 43.4%, Humphrey 42.7%

1972: Pre-Conventions Nixon 53, McGovern 37
Post-Conventions Nixon 61, McGovern 33
[ Nixon 12 point bounce ]
Election result Nixon 60.7%, McGovern 37.5%

1976: Pre-Conventions Carter 53, Ford 36
Post-Conventions Carter 51, Ford 40
[ Ford 6 point bounce ]
Election result Carter 50.1%, Ford 48.0%

1980: Pre-Conventions Reagan 37, Carter 34
Post-Conventions Reagan 39, Carter 38
[ Carter 2 point bounce ]
Election result Reagan 50.8%, Carter 41.0%

1984: Pre-Conventions Reagan 53, Mondale 39
Post-Conventions Reagan 56, Mondale 37
[ Reagan 5 point bounce ]
Election result Reagan 58.8%, Mondale 40.6%

1988: Pre-Conventions Dukakis 47, GH Bush 41
Post-Conventions GH Bush 48, Dukakis 44
[ GHW Bush 10 point bounce ]
Election result GH Bush 53.4%, Dukakis 45.7%

1992: Pre-Conventions Clinton 56, GH Bush 36
Post-Conventions Clinton 51, GH Bush 42
[ GHW Bush 11 point bounce ]
Election result Clinton 43.0%, GH Bush 37.5%

1996: Pre-Conventions Clinton 53, Dole 36
Post-Conventions Clinton 53, Dole 37
[ Dole 1 point bounce ]
Election result Clinton 49.2%, Dole 40.7%

2000: Pre-Conventions Gore 47, GW Bush 46
Post-Conventions Gore 47, GW Bush 45
[ Gore 1 point bounce ]
Election result GW Bush 47.9%, Gore 48.4%

2004: Pre-Conventions Kerry 49, GW Bush 47
Post-Conventions Kerry 48, GW Bush 47
[ GW Bush 1 point bounce ]
Election result GW Bush 50.8%, Kerry 48.3%

2008: Pre-Conventions Obama 45, McCain 45
Post-Conventions McCain 48, Obama 45
[ McCain 3 point bounce ]

Looking at these results, the following observations may be made:

In the sixteen elections for President between 1944 and 2004, in twelve of the sixteen the candidate leading after the conventions won the election. The conventions gave the aggregate bounce to the candidate who lost the election nine times, to the winner twice, and there was no aggregate bounce on two occasions. The 1988 election is the only time in the sixteen elections that the candidate who was trailing before the conventions went on to take the lead after the conventions on a bounce, and go on to win the general election. Of the nine times where the losing candidate enjoyed the best convention bounce, in six elections the trailing candidate was 17 points or more behind in the polls before the elections, and the bounce simply tightened the race. On average, the candidate leading in the polls after the conventions enjoyed election results 2.3 points higher than his poll support after the conventions, while the trailing candidate enjoyed election results 2.9 points higher than his poll support after the conventions.

Gallup was used exclusively for this examination, because only Gallup has consistent public results dating back to 1944, and only Gallup has maintained the same polling methodology for all of its election polling. When other poll agencies are similarly complete and consistent, I plan to start using their results as well.

UPDATE - What does it all mean?

Blogger does not let me attach Excel charts, which is a shame, because the graphic track is interesting. What happens on that track for the most part, is the candidate who wins the election tends to have strong (48%+) support during the end run of the race, while the loser, well, he is either far behind or slipping in his support (meaning he would have to spend a lot of effort shoring up his base). Timing is critical, in other words, and peaking too early really does kill your campaign.

Of course, I have to throw in some caveats. For a reasonable level of confidence, you really need 32 or more samples of data, and this gives only half that many, and there are a number of apparent outliers already. Truman appeared to come from way behind in 1948 to win, the only such case of the sixteen elections examined here. Also, 1944 and 1948 showed no apparent bounce from the conventions, but the 14 elections after them all showed a bounce effect. In most elections, the candidate ahead when the conventions began won the election, but in 2000 and 2004 the election winner was behind in the polls both before and after the conventions. It is all but impossible to know if there is significance this year in McCain's sudden lead after the convention, the first time since 1988 that a candidate went into the party conventions trailing but came out from the conventions with a clear lead.

1 comment:

Mr. Right said...

Hi, DJ.

One way to overcome the inability to post an Excel chart on Blogger, is to turn it into a jpg file.

Using a screen capture could do the trick nicely if you can't convert it directly, and you can crop the screen capture if you need to using another application, such as Photoshop.

If you really just can't do it, e-mail me and I'll try to help you out if I can. It's not all that hard, really.

All the best!