It’s August and time to start thinking about the Presidential election though nothing really matters until after Labor Day, in our opinion, especially in a race that is so tight. The latest Gallup tracking poll of registered voters has the President at 46% and Governor Romney 45%.
Fitting. A dead heat in the hottest summer on record.
For historical context, we looked at every Presidential campaign since 1936 and the polls from the Gallup interactive site on or around August 1st of every election year. We then compared them to the final margin of the victory in the popular vote in each of past nineteen Presidential elections.
There are many caveats about comparing polls across various Presidential campaigns. For example, in the days of smoke-filled rooms some of the candidates probably had not yet won their party’s nomination and were not even the presumptive nominee in early August. The lack of uniformity in processing and sampling may also distort the comps as some polls used registered voters and while others likely voters. Furthermore, the science, art, and process of polling and sampling has evolved over the past 70 years.
Nevertheless, it’s all we’ve got and close enough for government work. After all, we’re not writing a Ph.D. dissertation! Here’s what we found:
1) The current race appears to be one of the tightest at this point in the campaign;
2) There have been only four reversals where one candidate led the other in early August and lost the popular vote (highlighted in purple in the tables): 1948, Truman and Dewey; 1960, Kennedy and Nixon; 1988, Bush 41 and Dukakis; and 2000, Bush 43 and Gore.
3) Polls are volatile moving into the fall and public opinion can break big in the last few weeks before the election. In 1980, for example, President Carter closed the gap and went ahead of then Governor Reagan in late August and held a 6 point lead at the end of October among registered voters. Reagan surged in the last few weeks taking 50.7 of the popular vote to President Carter’s 41 percent.
The upshot? Very little predictive power in the data cuz’ “this time is different”, right?
(click here if tables are not observable)