Uncertainty and vote for third party candidates

[buritobr]

One could expect that third party candidates would do well in elections in which the winner is certain, when polls in the eve predict a large margin for one of the two major candidates, and do bad in close elections, in which the winner is uncertain. In close elections, voters would be more willing to cast the strategic vote for the lesser of the two evils.

By observing the vote for third party candidates in the post-WWII elections, we can see...

When the winner was not known on Monday evening
1948: 5,38%
1960: 0,73%
1968: 13,83%
1976: 1,91%
1980 (polls were predicting 3% margin for Reagan): 8,24%
2000: 3,75%
2004: 1,01%
2012: 1,74%
Average: 4,57%

When the winner was already known on Monday evening
1952: 0,49%
1956: 0,66%
1964: 0,48%
1972: 1,80%
1984: 0,67%
1988: 0,98%
1992: 19,54%
1996: 10,05%
2008: 1,54%
Average: 4,02%

So, the results did not fit into this theory. Vote for third party candidates was a little bit BIGGER when the election had uncertainty.

In 1948, Wallace had many votes even in states in the Northeast where the race between Truman and Dewey was close.
In 1968, the other Wallace, Nixon and Humphrey had almost 1/3 of the votes each one in some Southern states. Even though, southern conservatives did not decide to vote for Nixon in the last minute.
What is more weird is that liberals were free to vote for Nader in 1996 and 2008 without having risk to influence the result of the election, but he performed much better in 2000.

[buritobr]

Everyone knew Dewey was going to win in 1948, so your chart is kind of off.

Even if I remove 1948 from the group of 1960, 1968, 1976, 1980, 2000, 2004, 2012 and include it in the group of 1952, 1956, 1964, 1972, 1984, 1988, 1992, 1996, 2008, the averages will not be very different.