Generic poll numbers are only one metric used in predicting elections. Some political scientists use presidential approval ratings in their prediction models, and there is some evidence to support a link between presidential approval and midterm outcomes. Leading up to the 1998 elections, President Clinton's approval rating was high, in the low-to-mid-60s. That year, Democrats bucked the midterm election trend, gaining four House seats. By contrast, President Bush's approval was at 38 percent ahead of the 2006 elections, when Republicans lost 30 House seats.
Presidential approval itself depends on other factors, particularly the economy. Blumenthal notes, "When the economy's bad, presidential approval is low, and the 'out' parties tend to do better. Those things are all related." With Obama's approval rating in the mid-40s and unemployment nearing ten percent, these figures point to a good election year for Republicans.
Even with many indicators pointing in the same direction, prognosticating remains an uncertain science. Rasmussen emphasizes that presidential approval ratings and generic ballot poll results are only "general indicators of mood." He continues, "Anybody who wants to try and delve deeper and say you can take these broad indicators and project precisely how many seats somebody will gain or lose is missing the point."