Labor Day to Election Day patterns from past elections help put the current situation in context
GALLUP NEWS SERVICE
Revised Sept. 10, 2004
PRINCETON, NJ  The fundamentals of the presidential race did not change dramatically as a result of the Republican convention. George W. Bush had a slight lead over John Kerry among likely voters prior to the convention. He has now edged further ahead. Bush leads Kerry by 52% to 45% among likely voters, and by 49% to 48% among registered voters. With Ralph Nader included in the mix, it's Bush 52%, Kerry 45% among likely voters (Nader gets 1%), and Bush 48%, Kerry 46% among registered voters (Nader gets 4).
A question of keen interest to election observers focuses on the possibility of significant or dramatic change from this point forward. What is the probability that the election will change to the point where Kerry is ahead in the popular vote on Election Day?
"Gap Changes" Between Labor Day and Election Day
There have been 17 presidential elections since Gallup's election polling began in 1936. In each instance we can look at the difference in the margin between the Gallup Poll that immediately followed Labor Day and the final popular vote outcome.
RELATIONSHIP BETWEEN LABOR DAY POLL
ESTIMATE AND FINAL POPULAR VOTE OUTCOME ON ELECTION
DAY 


Republican candidate 
Democratic candidate 
Gap 
"Gap Change": 
2000 
% 
% 
pct. pts. 
pct. pts. 
Labor Day estimate (registered voters) 
41 
47 
6 
6 
Labor Day estimate (likely voters) 
44 
47 
3 
3 
Election Day outcome 
48 
48 
0 






1996 




Labor Day estimate (registered voters) 
35 
54 
19 
11 
Labor Day estimate (likely voters) 
36 
53 
17 
9 
Election Day outcome 
41 
49 
8 






1992 




Labor Day estimate 
42 
51 
9 
4 
Election Day outcome 
38 
43 
5 






1988 




Labor Day estimate 
49 
41 
8 
1 
Election Day outcome 
53 
46 
7 






1984 




Labor Day estimate 
55 
44 
11 
7 
Election Day outcome 
59 
41 
18 






1980 




Labor Day estimate 
37 
41 
4 
14 
Election Day outcome 
51 
41 
10 






1976 




Labor Day estimate 
40 
51 
11 
9 
Election Day outcome 
48 
50 
2 






1972 




Labor Day estimate 
61 
33 
28 
5 
Election Day outcome 
61 
38 
23 






1968 




Labor Day estimate 
43 
31 
12 
12 
Election Day outcome 
43 
43 
0 






1964 




Labor Day estimate 
32 
62 
30 
7 
Election Day outcome 
38 
61 
23 






1960 




Labor Day estimate 
47 
46 
1 
1 
Election Day outcome 
50 
50 
0 






1956 




Labor Day estimate 
52 
41 
11 
4 
Election Day outcome 
57 
42 
15 






1952 




Labor Day estimate 
56 
40 
16 
6 
Election Day outcome 
55 
45 
10 






1948 




Labor Day estimate 
47 
39 
8 
13 
Election Day outcome 
45 
50 
5 






1944 




Labor Day estimate 
45 
50 
5 
3 
Election Day outcome 
46 
54 
8 






1940 




Labor Day estimate 
40 
49 
9 
1 
Election Day outcome 
45 
55 
10 






1936 




Labor Day estimate 
45 
49 
4 
20 
Election Day outcome 
37 
61 
24 
The gap change between Labor Day and Election Day over these 17 elections has ranged from a low of 1 point (1940, 1960, and 1988) to a high of 20 points (1936). (The gap change represents the difference in the candidates' vote percentages around Labor Day and their final Election Day popular vote percentages.)
Here is another way of looking at the distribution of gap changes between those two milestones over the 17 elections between 1936 and 2000.
The median in the above distribution is 6 points. This means that half (eight) of the elections had gap changes between Labor Day and Election Day of less than 6 points, while half had gap changes of more than 6 points (one had a gap change of exactly 6 points).
The mean (mathematical average) gap change is 7.3 points, driven a little higher than the median by the high 20point gap of the 1936 election.
Based solely on history, that would put the predicted gap change for this year's election (comparing the Labor Day poll results with the actual popular vote on Election Day) in the six to sevenpoint range.
Shrinking or Expanding Gap?
In theory, that gap change, if it does occur, could be in both directions. Bush could gain six to seven more points on Kerry, or Kerry could gain six to seven points.
The historical analysis suggests that the change is more frequently in the direction of closing the gap, rather than expanding it, between Labor Day and Election Day. Of the 17 elections, the gap expanded in 5, while it shrank in 12.
Three of the elections in which the gap expanded involved Franklin D. Roosevelt. Beginning with the 1948 election, the Labor Day gap between the two majorparty candidates shrank in all but 2 out of 14 elections. The two postWorld War II elections in which the gap expanded were 1956 (Dwight Eisenhower gained steam on Adlai Stevenson as the fall campaign progressed) and 1984 (Ronald Reagan expanded his lead over Walter Mondale by Election Day).
Thus, if there is a change in the gap, the odds are higher  based on historical precedent  that Bush's lead will diminish than they are that Bush will move further ahead of Kerry.
Winning or Losing?
It is of some interest to look at the relationship between the Gallup Poll numbers for the candidates on Labor Day and the winner on Election Day.
Across the 17 elections between 1936 and 2000, a candidate who was behind in the Labor Day poll went on to win on Election Day in only 3:
 In 1980, Gallup's Labor Day poll had Jimmy Carter up by 4.
Reagan won by 10.
 In Gallup's 1960 Labor Day poll, Richard Nixon was up by one.
John F. Kennedy won by less than one percentage point.
 In Gallup's 1948 poll, Gallup's Labor Day poll had Thomas Dewey
up by eight points. Harry Truman won by five.
In other words, the candidate who is ahead on Labor Day usually wins the election, regardless of changes in the margins over the last two months of the campaign.
Bottom Line
Bush is ahead of Kerry by one to two points among registered voters and by seven points among likely voters in Gallup's Labor Day poll. Movement in the candidates' trial heat "gap" between Gallup's Labor Day polling and the popular vote on Election Day is common. The average movement in the margin has been roughly between six and seven points.
One example that bodes well for Kerry is the 1980 presidential election in which challenger Reagan moved the gap 14 points against the incumbent Carter.
Other races in which the gap shrank significantly included 1996, 1976, 1968, and 1964, although in none of these did the lead change between the Labor Day poll and Election Day.
There are exactly 55 days to go before the election on Nov. 2. Obviously, much can change between now and then. Indeed, hundreds of millions of dollars will be spent by both campaigns over the next two months in an attempt to change people's minds about the election. The news media will give extraordinary coverage to the candidates' every utterance. Three presidential debates and one vice presidential debate are planned at this point, and they could have a major effect on voter sentiment. Plus, Kerry is shaking up his campaign staff and we can expect some new and different strategies from his side of the ledger that could affect the race.
Still, a presidential race is in some ways similar to a football game. Certainly, much can change up to the final quarter  or even the final minute  of play. But in general, the team that is ahead at any point in the game has a higher probability of winning than the team that is behind. As the game progresses, the chances that the team that is ahead will maintain that lead become higher and higher. (And, the larger one team's lead at any given point in the game, the greater the probability that the team will go on to win.)
The same pattern occurs in a presidential race  particularly after the summer conventions. A candidate who is ahead has a higher probability of winning than a candidate who is behind, and the larger the margin for the candidate in the lead, the higher the probability that he will win.
The use of historical Gallup Poll data provides us with the ability to be a bit more specific at this point in the 2004 campaign. We can use the observed relationship between Labor Day Gallup Poll trial heat estimates and the final popular vote outcome from past elections to provide some crude context for the likelihood of a Bush or a Kerry victory this year.
What we find is a history of significant change between Labor Day and Election Day. In all presidential elections there has been some movement between Gallup's Labor Day poll and the final outcome on Election Day. Of course, the change in the race has historically moved in two directions; namely, the Labor Day gap gets bigger or the Labor Day gap closes. In other words, in this context, Bush could gain, or Kerry could gain. The general tendency is clearly toward a closing of whatever gap exists. Certainly then, the race is close enough at this point to suggest that while it is possible that Bush may maintain his lead or expand it, it is also quite possible that Kerry will gain and move into the lead himself.
Finally, it is important to note that even with 17 elections, we are dealing with a limited set of cases for precise statistical modeling. Polling techniques have changed over the years, and other changes in the ways poll estimates were made since 1936 mean the comparability of situations across the years is not exact. Nevertheless, it is instructive to use the historical data to provide the best possible empirical basis for assessing election probabilities. Otherwise, such assessments are left to speculation and informed guessing.