For fun (yes, while the rest of Getbig is banging models in their private jets I do this sort of stuff) I looked at a simple economic model for its ability to predict the result of presidential elections.
The model is simple: if seasonally adjusted unemployment is June isn't lower than it was in March (i.e., if unemployment did not decrease in the first quarter) of the election year, then the incumbent party will lose.
Year March June
1948 4.0 3.6
(decrease)---> Incumbent party wins (D-Truman)
1952 2.9 3.0
(increase)----> Incumbent party loses (R-Eisenhower)
1956 4.2 4.3
(increase)----> Incumbent party wins (R-Eisenhower)
1960 5.4 5.4
(unchanged)----> Incumbent party loses (D-Kennedy)
1964 5.4 5.2
(decrease)----> Incumbent party wins (D-Johnson)
1968 3.7 3.7
(unchanged)-----> Incumbent party loses (R-Nixon)
1972 5.8 5.7
(decrease)-----> Incumbent party wins (R-Nixon)
1976 7.6 7.6
(unchanged)-----> Incumbent party loses (D-Carter)
1980 6.3 7.6
(increase)-----> Incumbent party loses (R-Reagan)
1984 7.8 7.2
(decrease)-----> Incumbent party wins (R-Reagan)
1988 5.7 5.4
(decrease)-----> Incumbent party wins (R-Bush)
1992 7.4 7.8
(increase)-----> Incumbent party loses (D-Clinton)
1996 5.5 5.3
(decrease)-----> Incumbent party wins (D-Clinton)
2000 4.0 4.0
(unchanged)-----> Incumbent party loses (R-Bush)
2004 5.8 5.6
(decrease)-----> Incumbent party wins (R-Bush)
2008 5.1 5.6
(increase)-----> Incumbent party loses (D-Obama)
As you can see, the model accurately predicts 15 of the last 16 elections going back to 1948, which is as far back as the Department of Labor statistics go. [1]
Only Eisenhower bucked the trend by winning re-election in 1956. This probably has to do with his being a very popular man, with his approval rating in his first term never dipping below 60% (plus, the increase in unemployment was very little). [2]
With the recent June unemployment data in, we can extend the model to November:
Year March June
2012 8.2 8.2
(unchanged)----->Incumbent party loses (R-Romney)
An Eisenhower-style, popularity-based outcome is precluded by the fact that Obama's approval ratings are much lower than Eisenhower's were, with the latest weekly average presidential rating at 47%.
Inductive generalizations are never certain and it is impossible to calculate the precise probability Obama will win/lose given the small amount of cases under consideration. But, with 94% of the elections adhering to the model and with it obviously being non-random (we aren't flipping coins here - there's an obvious reason why the data predicts outcomes so well), I think it's safe to say there is a "moderately high" inductive probability that Obama will lose re-election.
1.
http://data.bls.gov/timeseries/LNS140000002.
http://online.wsj.com/public/resources/documents/info-presapp0605-31.html3.
http://www.gallup.com/poll/116479/barack-obama-presidential-job-approval.aspx
