Via The Monkey Cage:
But as we noted in that post, any specific seat estimate comes with a lot of uncertainty. Any forecasting model is a simplification of reality and will not make perfect predictions, which makes uncertainty inevitable. As a result, it can be more useful to convert estimates to probabilities. This enables us to answer questions such as, “What is the chance that the Democrats retake the House?”
In short, what we do is lots of simulations based on our model’s results. (In statistical terms, we are sampling from the error distribution of the model.) The uncertainty underlying the model means that simulations will generate a range of forecasts. The question is, how many of those simulations forecast that the Democrats would win 218 or more seats?
The answer is very few. Just over 1 percent, in fact. This is a testament to the fact that current conditions in the country, and the presence of so many Republican incumbents, make it hard for the Democrats to pick up many seats. In order for that forecast to change measurably in the Democrats’ favor, the economy needs to grow more rapidly or President Obama needs to become more popular, or both. A few more Republican retirements and strong Democratic challengers wouldn’t hurt, either.