Picking up from the main Blog page, I provide a detailed explanation of how simply reading the polls as percentages does not give us enough information of the dynamics of the “isovotes” that underlie them.
Continuing the conversation…
In our current situation, however, the percentage increase may be good news for the Democrats, as long as the percentages are roughly equal from state to state, another factor that has to added to the mix. But that requires a number contingencies be in place. According to 2004 records, using figures appearing on both the Wikipedia pages for the Democratic and Republican parties there are 72 million registered Democrats, 55 million registered Republicans and 42 million persons who vote as Independents.
The discrepancy of 17 million more voters in favor of the Democrats appear to give them a degree of wiggle room the Republicans do not enjoy. The percentages are taken from the Gallup Poll Congressional Election Key Indicators for October 3, 2010. Mere numbers, however, do not tell the entire tale.
Let’s create a equation to see what the trends are really telling us:
- Total Number of voters = TV = 169 million voters
- Registered Democratic Party voters = DRV = 72 million voters
- Democratic votes = DV =No. of voters casting ballots for Democratic candidates
- DRVs who actually vote = AV = 43% of RPV using Gallup polling
- Independents who vote = IV = 35% of IV who vote for Democratic candidates
- Republicans who cross vote = RCV = 2% of RCV voting for Democratic cands (Gallup estimates)
Therefore, we have the equation for the Democrats:
(DRV x %AV) + (IV x %AV) + (RCV x %AV) = DV
Running the numbers using the assumptions above, this is how the equation works:
(72 mil DRV x .43 AV) + (42 mil IV x .35 AV) + (55 mil RCV x .02 AV) = 46.76 mil
If we follow the conventional perspective that the Republicans are more motivated this year, and estimate that Registered Republican Voters (RRV) who actually vote is going to be 46% (using Gallup estimates) and that the Independents who will vote for the Republican candidates is 45% (Gallup estimates) and the number of Democrats that will cross party lines is 3% (Gallup estimates), we will see what the number adds up to be:
(55 mil RRV x .46) + (42 mil IV x .45) + (72 mil DCV x .03) = 46.36 mil
So, one can see how the Republicans despite their 17 million registered voter gap, on a national basis, are easily within striking distance of the Democratic majority. Keep in mind, though, that I am using aggregate data and making assumptions on party membership data that is not just 6 years old, but does not include the significant shifts in voter demographics that occurred in 2008 that gave Obama the presidency and the Democrats strong majorities in both houses. It should also be noted that the Gallup polling results as of October 3, have the Republicans up by +3 points.
How do we calculate what the percentages mean? We know roughly how many people are going to vote, 46.8 million Democrats and 46.4 million Republicans, a total of 93.2 million voters. The percentage difference is 50.2% (D) to 49.7% (R). That’s a tiny difference of only 466,000 voters compared on a national scale. But that analysis is actually not correct, because these numbers represent the categories of voters, D, R, and I that will vote either Republican or Democratic. Tucked inside the partys’ totals are 14.7 million Independents who will vote Democratic and 18.9 million who will likely vote Republican in this election. That is much larger gap of 4.2 points in the favor of the GOP.
If the polls, though, are moving in favor of the Democrats, giving them a 3 point advantage, does that mean in all the races they will have a 3% greater chance to win? No, because the percentages are not equivalent to probability. It is important to remember is that these numbers are aggregated from polling results. A random sample (which means no one person has a greater chance of being picked than any other) of voters is selected are contacted and asked questions about their preferences. The sample of 33 surveys does not exclude internet polling, robocalls, or party-affiliated surveys.
In one respect the polling results do not have anything to do with real numbers of people casting ballots. The polls are a snapshot of the sample. The fact that they are randomly drawn is considered a statistical standard because it is a method to ensure the playing field is level. It’s good in theory, yet even if the sample is random, if the questions are weighted to one bias or another, the poll should not be considered a reliable picture, or statistically significant (and there are equations that can test for that significance) The percentages are just that, the percentage of respondents who answered one way or another to their questions.
That is why looking at the subpersonal isovotes is important. It also why it is so difficult to find out the content and nature of the isovote dynamics at work in any given individual.
Where does that leave us? Pollsters have developed all sorts of propriatary processes to tease out not only what the isovote structures contain, but also how likely are they to change through time as we move closer to the election, now less than a month away.
What we can say is that the slope of the poll percentages when aggregated does illuminate a picture of real movement. I suggest that the percentage is not as important as the change positive or negative and the speed of which those changes are being driven. Another factor we should look at is the relationship between registered voters compared to likely voters polled. Although likely voters is a standard metric surveyors use, I would submit that registered voters provide a more robust sample:
If you haven’t taken a couple of minutes to read Mark Blumenthal’s HuffPost Pollster article on how survey organizations choose “likely voters” I encourage you to read that now. Click here.
Nate Silver, founder of FiveThirtyEight, which was recently acquired by the New York Times, has a somewhat different take on the likely voter issue:
The likely voter adjustment. Throughout the course of an election year, polls may be conducted among a variety of population samples. Some survey all American adults, some survey only registered voters, and others are based on responses from respondents deemed to be “likely voters,” as determined based on past voting behavior or present voting intentions. Sometimes, there are predictable differences between likely voter and registered voter polls. In 2010, for instance, polls of likely voters are about 4 points more favorable to the Republican candidate, on average, than those of registered voters, perhaps reflecting enthusiasm among Republican voters. And surveys conducted among likely voters are about 7 points more favorable to the Republican than those conducted among all adults, whether registered to vote or not.
By the end of the election cycle, the majority of pollsters employ a likely voter model of some kind. Additionally, there is evidence that likely voter polls are more accurate, especially in Congressional elections. Therefore, polls of registered voters (or adults) are adjusted to be equivalent to likely voter polls; the magnitude of the adjustment is based on a regression analysis of the differences between registered voter polls and likely voter polls throughout the polling database, holding other factors like the identity of the pollster constant.
As once can see, I differ in my assumptions regarding the reliability of registered voters versus likely voters. Having this data selected out, on the assumption that registered voters are the most likely to vote, one can then look at the data as the trend lines:
Two points can be inferred from these trend lines, although with prudent caution. First is the obvious, that the blue line representing the upturn in Democratic voters has crossed the red Republican line in just the last week. The second point can be made from the statistic represented as the R² value. Using Excel’s trend line function, I tested each set of data for the line formula that would produce an R² value as close to the ideal of .9999 as possible. The polynomial formula produced the best fit and the highest value for both the Democratic and Republican data. I also tested the sample for ANOVA, which was significant at the p = 0.010 at the .05 level.
A comparison can be made, even though both R²’s are well below the ideal, that the relative strength of the Democratic upward trend is greater (for reasons not discernible in this one number) than the Republican’s downward trend. Perhaps this is an indication of the Democrats beginning to close the so-called enthusiasm gap moving from vacillating whether or not to vote, to clearly deciding to vote, but that is admittedly a guess. It could be simply an indicator that a growing number of Democrats who have been undecided have moved toward voting for their party’s candidates. They were going to vote regardless, but now they have begun locking themselves into voting for the Democratic candidates in their state.
The reason for the Republican decline is no clearer. One might infer that the very small R² is not particularly negative, that this is just a dip that in the next week will reverse itself and their hold on the the higher percentage will once more grow at a healthy pace, preventing the Democrats from being able to close the gap. In this scenario, it is possible to speculate that the GOP would win more seats than they have predicted. It can also mean, when we test this again in a week if the number and the slope continues to decline that the Republicans have peaked and will have to scramble to sustain the leads they have, the result being they may win fewer seats than they have predicted.
Having said that, I want to quote Nate Silver again regarding his comments on the element of uncertainty in the current political environment, and on this, I am in agreement with him. What he suggests is very much in line with my earlier post The Black Poll Wars:
But whatever the reasons, the dynamics of the battle for the House are much different this year than in the recent past. Viewed in this context, the uncertainty that our model implies should be viewed as a feature rather than a bug. We are certainly not afraid to make bold forecasts where they are warranted: for instance, our model was much quicker than others, after the financial collapse in September 2008, to figure out that John McCain had essentially zero chance of winning the Presidency. As I have written about extensively, it also makes some very certain-seeming forecasts in some individual races for Senate and governor, regarding candidates as 90 percent, 95 percent, or even 99 percent favorites in some contests that others regard as toss-ups.
Sometimes, however, a thorough and objective analysis of the data leads one to the opposite conclusion: that the competition is too sure of itself. Our model figures there is a very wide range of potential outcomes because that is the only responsible forecast. We’re not being meek or wishy-wishy: instead, we are firmly, boldly, affirmatively and happily embracing the uncertainty. This is not because of any intrinsic property of our forecasting model; rather, it is because of the particular set of circumstances on the ground this year.
A year ago when the Tea Party emerged during the August Congressional recess, they started off very strong. By the end of the month they had been so over-covered by the media that the American public, in my opinion, became saturated with their acerbic rhetoric and began to view them as tragic or annoying whiners rather than courageous patriots. The isovotes quietly moved back to the Obama agenda, and right under their noses allowed him by the end of the year to pass the health care reform act, and in rapid succession pass numerous other pieces of legislation that the Republicans for all their bluster in August found themselves completely unable to even filibuster.
This year we don’t have the Tea Party putting on displays of righteous indignation in front the of cameras in Town Hall Meetings, but Tea Party 2.0 has become a complex, if still immature, force in American politics. Their complexity, though, may be viewed by most outsiders as ambiguity. In the turbulence and uncertainty of the gloomy days of the post-Great Recession depression, the American electorate may end up deciding the risk is just too great to elect these people and grant them the license to experiment with an already rocking and leaky ship of state to move us into calmer waters. Their ideas for repairing it, all too often, seem grounded in a fantasy of an American that never really was, but which they are completely dedicated to creating as if it had been. Frankly, no one has the slightest idea if that will work.