This year is particularly difficult to assess because of the obvious political passion of the new Tea Party Movement. Whether one agrees with the current Administration and Congress or not, the last two years have been consequential, with nearly a trillion dollars of stimulus spending and a complicated and copious health care reform bill. This, coupled with the pain of nearly 10% unemployment that seems stubbornly fixed and a depressed housing market has dramatically changed the political landscape. A president who once enjoyed an approval rating of nearly 70% has seen that approval sink to below 50%.
Typically most Republicans will vote for Republicans, most Democrats will vote for Democrats. The questions are: How motivated these partisans are to vote,? What is the self-identification with the parties? Toward which party will the unaffiliated voters break? These questions make it difficult to model who will be the likely voter this year.
Rasmussen was the was the most accurate poll in predicting the presidential outcome in 2008, and is consequently highly respected. RealClearPolitics.com trys to average out poll biases by publishing a running average of major polls. Nate Silver of fivethirtyeight.blogs.nytimes.com is a statistician who during the last election cycle generated accurate predictions using multiple polls and averaging the results of thousands of computer simulations.
Although I yield to these full-time organizations and professionals for their noble work in making these assessments, I wonder how predictive a simpler approach might be. The Gallup Organization is over 70 years old and has a long track record with respect polling. Whatever its virtues and deficiencies the Gallup poll has existed for a long time.
One important bellwether for mid-term elections is the generic preference poll: what fraction of likely voters prefer an unnamed Democratic or Republican candidate. The horizontal axis below is the Gallup-measured Democratic advantage in the generic poll, for the mid-term elections since 1950. If this value is 2, it means the Democrats are preferred over Republicans by a 2% margin. The vertical axis is the margin, either positive or negative, that the Democrats earned in the actual House vote. For example, the value of 1 means that the Democratic margin in the actual vote for the House of Representatives is 1%.
A linear fit to these points is y = 0.8901 x – 0.0354, where x is the Democratic advantage in the generic poll and y is Democratic advantage in the House vote. The 0.8901 slope indicates that, in general, Democrats slightly under perform their poll numbers. Also note that the correlation coefficient sqaured, R2, equals 0.88, meaning that 88% of the observed election-to-election variations can be accounted for with the linear fit.
As of this writing, Gallup is uncertain as to its likely voter model. If there is a low turn out, this usually means proportionately more Republicans turn out, the generic poll value is -17% (the negative sign means the Democrats poll less well than Republicans). With a high turn out election, Gallups likely voter model produces a smaller -11% value.
Given the high energy level on the part of Tea Party members (likely to vote for Republicans), it is not clear that a high voter turnout is to the advantage of Democrats this year. Nonetheless, if we use the -11% value, the simple linear model suggests that Republicans could expect to gain 64 seats for a total of 242. Republicans need 218 to take the House. The 95% confidence limits suggested by the model predict a range of gains from 57 to 70 seats. If we used the -17% value for the likely voter generic preference for Democrats, the gain for Republicans would be an enormous 73 seats, for a total of 251, with 95% confidence for a gain of between 64 and 78 seats.
At present, the RealClearPolitics.com average gives Republicans a certain 220 seats with with 37 toss ups. Apportioning the toss ups evenly would give Republicans 239 seats for net gain 61 seats. Nate Silver predicts a Republican 51-seat gain. The proposed linear model clearly is more favorable to Republicans than more sophisticated ones. Since the Republicans have never before had a generic poll in their favor by more than 7%, we are predicting outside the range that this simple linear model has experienced.
There is still one week before the election. We will update the predictions using the latest possible generic poll and see how well this first-order model performs.
There is much that is opaque to the casual outside viewer of polls. In baseball, a spectator is not always able to tell that whether a well hit ball to the outfield has enough power to make it out of the park. However, astute students of the game will watch the outfielders who have a better perspective and likely a keener eye. By watching how they respond to the ball, one can get a indication of where the ball will go. If the outfielder starts to run toward a particular point, one can be assured that the ball will remain in the park. If the outfield doesnt move much, he likely sees that the ball will fly over the fences. Whatever the public and private polls are indicating, Democrats are rapidly pulling funding out of key districts, indicating they are retreating to protect their core in close raises where additional spending could have an effect. This is certainly indicative of what they believe is happening.
This entry was posted on Sunday, October 24th, 2010 at 8:49 am and is filed under Politics, Social Commentary. You can follow any responses to this entry through the RSS 2.0 feed.
You can leave a response, or trackback from your own site.
Predictions for the House Election 2010
This year is particularly difficult to assess because of the obvious political passion of the new Tea Party Movement. Whether one agrees with the current Administration and Congress or not, the last two years have been consequential, with nearly a trillion dollars of stimulus spending and a complicated and copious health care reform bill. This, coupled with the pain of nearly 10% unemployment that seems stubbornly fixed and a depressed housing market has dramatically changed the political landscape. A president who once enjoyed an approval rating of nearly 70% has seen that approval sink to below 50%.
Typically most Republicans will vote for Republicans, most Democrats will vote for Democrats. The questions are: How motivated these partisans are to vote,? What is the self-identification with the parties? Toward which party will the unaffiliated voters break? These questions make it difficult to model who will be the likely voter this year.
Rasmussen was the was the most accurate poll in predicting the presidential outcome in 2008, and is consequently highly respected. RealClearPolitics.com trys to average out poll biases by publishing a running average of major polls. Nate Silver of fivethirtyeight.blogs.nytimes.com is a statistician who during the last election cycle generated accurate predictions using multiple polls and averaging the results of thousands of computer simulations.
Although I yield to these full-time organizations and professionals for their noble work in making these assessments, I wonder how predictive a simpler approach might be. The Gallup Organization is over 70 years old and has a long track record with respect polling. Whatever its virtues and deficiencies the Gallup poll has existed for a long time.
One important bellwether for mid-term elections is the generic preference poll: what fraction of likely voters prefer an unnamed Democratic or Republican candidate. The horizontal axis below is the Gallup-measured Democratic advantage in the generic poll, for the mid-term elections since 1950. If this value is 2, it means the Democrats are preferred over Republicans by a 2% margin. The vertical axis is the margin, either positive or negative, that the Democrats earned in the actual House vote. For example, the value of 1 means that the Democratic margin in the actual vote for the House of Representatives is 1%.
A linear fit to these points is y = 0.8901 x – 0.0354, where x is the Democratic advantage in the generic poll and y is Democratic advantage in the House vote. The 0.8901 slope indicates that, in general, Democrats slightly under perform their poll numbers. Also note that the correlation coefficient sqaured, R2, equals 0.88, meaning that 88% of the observed election-to-election variations can be accounted for with the linear fit.
As of this writing, Gallup is uncertain as to its likely voter model. If there is a low turn out, this usually means proportionately more Republicans turn out, the generic poll value is -17% (the negative sign means the Democrats poll less well than Republicans). With a high turn out election, Gallups likely voter model produces a smaller -11% value.
Given the high energy level on the part of Tea Party members (likely to vote for Republicans), it is not clear that a high voter turnout is to the advantage of Democrats this year. Nonetheless, if we use the -11% value, the simple linear model suggests that Republicans could expect to gain 64 seats for a total of 242. Republicans need 218 to take the House. The 95% confidence limits suggested by the model predict a range of gains from 57 to 70 seats. If we used the -17% value for the likely voter generic preference for Democrats, the gain for Republicans would be an enormous 73 seats, for a total of 251, with 95% confidence for a gain of between 64 and 78 seats.
At present, the RealClearPolitics.com average gives Republicans a certain 220 seats with with 37 toss ups. Apportioning the toss ups evenly would give Republicans 239 seats for net gain 61 seats. Nate Silver predicts a Republican 51-seat gain. The proposed linear model clearly is more favorable to Republicans than more sophisticated ones. Since the Republicans have never before had a generic poll in their favor by more than 7%, we are predicting outside the range that this simple linear model has experienced.
There is still one week before the election. We will update the predictions using the latest possible generic poll and see how well this first-order model performs.
There is much that is opaque to the casual outside viewer of polls. In baseball, a spectator is not always able to tell that whether a well hit ball to the outfield has enough power to make it out of the park. However, astute students of the game will watch the outfielders who have a better perspective and likely a keener eye. By watching how they respond to the ball, one can get a indication of where the ball will go. If the outfielder starts to run toward a particular point, one can be assured that the ball will remain in the park. If the outfield doesnt move much, he likely sees that the ball will fly over the fences. Whatever the public and private polls are indicating, Democrats are rapidly pulling funding out of key districts, indicating they are retreating to protect their core in close raises where additional spending could have an effect. This is certainly indicative of what they believe is happening.
This entry was posted on Sunday, October 24th, 2010 at 8:49 am and is filed under Politics, Social Commentary. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.