According to this piece from yesterday's Journal, it's going to be a long, long time before unemployment returns to its pre-recession levels:

The U.S. has shed 7.2 million jobs since the recession began in December 2007. How long will it take for the economy to replace them? ...

In addition to replacing 7.2 million lost jobs, the economy needs an additional 100,000 a month to keep up with population growth. If the job market returns to the rapid pace of the 1990s -- adding 2.15 million private-sector jobs a year, double the 2001-2007 pace -- the U.S. wouldn't get back to a 5% unemployment rate until late 2017, Rutgers University economist Joseph Seneca estimated. And that assumes no recession between now and then. "Even with some very optimistic assumptions, it's a long road back," Mr. Seneca said. ...

IHS Global Insight predicts the total number of jobs in the U.S. won't return to prerecession levels until 2013. And that doesn't account for the growth in the labor force, so it forecasts unemployment will be a painfully high 8.1% then. Getting back to the 5% unemployment rate that prevailed before the downturn isn't anywhere in the firm's 10-year forecast horizon. At the end of 2019, it puts unemployment at 5.75%.

Ugh. 8.1 percent unemployment in 2013 means a bit higher than that when Obama stands for re-election in 2012. In case you're wondering, the last time the unemployment rate was nearly as high just before an election was 1992 (7.6 percent in September of that year--probably the last unemployment statistic to sink in before Election Day), which didn't work out so well for the incumbent. Other high election-year unemployment rates in recent decades: 7.5 percent (September 1980) and 7.6 percent (September 1976), which also don't appear to have helped the incumbent party's cause.

On the other hand, it's probably all somewhat relative. The 7.6 percent unemployment rate in September 1992 was pretty close to the peak of 7.8 percent from that business cycle, whereas 8.3 or 8.4 percent unemployment in September of 2012 would be substantially lower than the peak from this business cycle. By way of comparison, the unemployment rate stood at 7.3 percent in September 1984, but that apparently qualified as "morning in America" after a peak of 10.8 percent in 1982. 

P.S. See Judis for more on unemployment and presidential popularity.