Jay Cost: A Republican Meltdown? Part II

This week I critiqued the idea that, in the House races, the Republican Party is suffering from a meltdown in which the bottom is dropping out in the second-, third- and even fourth-tier races.

I argued that this hypothesis is underdetermined because there are other, equally reasonable but theoretically distinct, theories to account for the data. One of them is that the lower-tier races might be swinging because of relatively little candidate activity.

I argued:

The second- and third-tiers feature races where there is relatively little campaign activity. On the top-tier, candidates are unloading huge war chests to crowd out any other political information except what they wish to communicate. On the latter two tiers, this is not so much the case, which is not to say that there is no activity. There is just less. Might these responses be more affected by the national political environment at the time the poll was taken because of less campaign-related information? And, if that is the case, are they "real" responses? Or are they more akin to what we might call "non-opinions" that are drawn not from deep-seated feelings and orientations, but rather from top-of-the-head answers that simply sample the overall media take on the situation? If the latter is the case, then these results are unlikely to be stable because incumbents in these races will soon begin to unload their war chests to focus the voter not on the generalities of national politics, but national politics in the context of that incumbent. And, furthermore, since the poll was taken during the nadir of the GOP's position in the national press - should we not expect these numbers to improve, independent of candidate activity, just as top-of-the-fold headlines like "Were GOP Leaders involved in Illegal Sex Cover-up?" disappear from the newspaper?
Think of it this way. The top-tier races feature loads of money being spent. This money serves as an anchor that keeps voter opinions from drifting with the media tide. So, they were the least likely to change. In races where the candidates have not been as active, voters are still more inclined to go with the media flow. Hence, the decline. Insofar as Republican incumbents start spending money to anchor these voters -- those numbers will change, especially in light of the fact that Democratic challengers in those races tend to be under-funded and therefore unable to fully counteract incumbent campaign communications.

Also this week, I argued that the results of the recent poll in PA 04 were consistent with this theory. Incumbent Melissa Hart had spent little money thus far to really "anchor" voters - and, according to this theory, when she does, they will come her way. A few people did not like that argument. They accused me of "rationalizing the results," which, to an extent, is understandable. It is not appropriate to start applying theories before you have offered some initial testing of them - unless of course you are a Marxist or a string theorist, neither of which I happen to be.

So today, to rectify that problem - I am indeed going to test this "floating" hypothesis. The basic thesis is that the national mood is what is driving the bad poll numbers on the lower tiers. When Republican candidates engage, their numbers will improve because voters will have turned from the national climate to the local climate. I also expect other factors - specifically, incumbent tenure and district partisanship - to be influential independent of candidate spending. So, I have included control measures for these factors.

To test this hypothesis, I have collected the data from the most recent batch of the Majority Watch polls in Republican-held districts. These provide a nice mix of first, second, third and even fourth-tier races to investigate. It is a good, representative sample of the Democratic playing field. Importantly, I have excluded the races that have been directly affected by national GOP troubles - Abramoff, Foley, etc. The data in these races indicates that money is not having any kind of systematic effect, which is consistent with my broad hypothesis - because, for Republicans where the worst of the national climate is local, there is no solace in changing focus. Republican poll positions in this subset seem to be largely conditioned by district partisanship, which explains 63.5% of the variation within it. This also makes a good deal of sense in the light of the theory I am offering. The negative national mood that Republicans are facing is what is the major factor in these races - and the extent to which the local candidate is being harmed by it depends upon how generally amenable his/her electorate is to the Republican Party. Accordingly, it is appropriate to exclude these Republicans from the analysis, and not control for them. There seems to be a distinct causal process in these districts.

I test my hypothesis using a statistical technique called ordinary least squares regression. It is a method that offers an equation that indicates the extent to which the independent variables affect the dependent variables. It is a common tool in social science. Unfortunately, the use of this technique also requires me to deal with the remaining upstate New York seats that were not dealt with in the prior paragraph. Regression analysis offers a prediction of each result which is then compared with the actual result. The difference between the two is known as the "residual." This residual, for regression analysis to be valid, must be independent of any other factors. And, the basic regression model I have outlined so far does not have independent residuals. They are correlated with whether or not the observation is of an upstate New York district.

In other words, the Republicans are doing worse in upstate New York than they "should" be in the Majority Watch polls. All else being equal, being a Republican on the ballot in upstate New York reduces one's standing by about 5%. So - one way or another, upstate New York must be dealt with. The important issue is whether (1) the same causal process that is moving the numbers nationwide is moving the numbers in New York, and the latter is simply depressed for some reason, or (2) those numbers are being driven by some unique causal process. The evidence clearly points to the latter. The upstate New York numbers seem to be driven by a different process than the rest of the nation - and the process seems to be similar to what is occurring in the districts where GOP candidates are directly tied to the state of the nation. Namely, district partisanship alone explains 52% of the variation in these districts, where it only explains 0.9% of the variation in the rest of the sample. Furthermore - just like the "scandalous" districts, the tests I ran showed that my model was extremely powerful outside upstate New York and extremely weak in upstate New York.

Why might this be the case? The answer that instantly comes to mind is Mark Foley. The time around which these polls were taken was the time that scrutiny of Tom Reynolds was heating up. Just as Tom Foley helped to sink the Washington State Democratic delegation in 1994, perhaps Tom Reynolds was sinking GOP numbers in upstate New York. There might, of course, be a singular explanation for the variation in polling that unites both upstate New York and the rest of the nation, but this seems to me to be unlikely, given how divergent the power of my model is, given the similarity of these districts to the other "scandalous" ones, given the Reynolds speculation that was occurring during the polling. All in all, my intuition is that - like the other "scandalous" districts in the Majority Watch slate of polls from October - national events overwhelmed the normal causal process in upstate New York, thanks to the proximity of the national storyline. It therefore seems appropriate to exclude these cases as well, and that is what I have done.

Also - a jargon note. When pollsters discuss the "margin of error," they always mean it in relation to a certain level of confidence, usually 95%. When a lead is "outside the margin of error," that means that they are at least 95% confident that one candidate has a lead over the other candidate, i.e. the chance that normal statistical variation has produced the result is 5% or less. I am going to report similar confidence levels. This will tell you at what percentage I am confident that a factor's relationship with another is not reducible to normal statistical error.

My broad hypothesis admits of several deductions, each of which can be individually tested. Remember, the broad hypothesis is that activating the campaign systematically helps Republican candidates because it turns the voters away from national issues. I am calling it the "floating" hypothesis.

Deduction 1. Republican candidate spending is positively related to his position in the polls.

Intuition. As the Republican candidate unloads his war chest, he turns the voters away from the national context, which right now is not favorable to the GOP, and to the local context, which is more favorable.

Results. This deduction is confirmed. We can be 98.3% confident that Republican candidates' spending - when we control for district partisanship and incumbent tenure - has a statistically significant effect upon their position in the polls. For every $100,000 the Republican spends, his position in the polls increases by 0.14%. This model - which predicts Republican poll position based upon Republican spending, incumbent status and district partisanship - actually predicts 25.2% of all the variation in the Majority Watch polls in the sample.

I also found that Republican spending has a negative relationship with the Democrat's position in the polls (i.e. the more money the Republican spends, the less well the Democrat fares), but this result is statistically significant only at the 8.5% confidence level, which means that it is more-than-likely a result created by "noise." This might be due to the fact that Republican incumbents usually only directly attack their opponents when they sense real trouble - thus, Republican dollars enhance the Republican poll position rather than diminish the Democratic poll position.

Deduction 2. Democratic candidate spending is not helpful to his position in the polls, and is not harmful to the Republican candidate's position.

Intuition. Democratic spending at the local level helps to generate the local campaign, which will not enhance the Democratic candidate's position. Democrats do not do any better by spending money because, in so doing, they prompt the Republican candidate to engage - and therefore they shift the voter's focus from the national to the local. They thus "gain" nothing on the Republican.

Results. This hypothesis is confirmed, with some peculiar results. I found that the more money the Democratic candidate spends, the worse he or she does. This result was only significant at the 44.8% confidence level, so it must be chalked up to random variation. Nevertheless, it is opposite of my expectation. I expected that Democratic spending would yield a very slight, and statistically insignificant increase in the Democratic position. I was not expecting a slight and statistically insignificant increase in the Republican position.

What could explain this result? It is surely not that the Democratic message on the ground hurts Democrats. It is probably rather, a function of the broader thesis: as the Democratic candidate starts spending money, the Republican candidate will follow suit - thus shifting the focus of the electorate from the national landscape to the local campaign.

Equally peculiar was the finding that Democratic spending is consistent with an increase in the GOP position. Specifically, every $100,000 the Democrat spends, the Republican's position increases by 0.12%. And the result here is surprisingly significant - registering at 86.7% (though still not quite high enough to accept as true). Again - this is probably due to the fact that Democratic spending prompts a response from the Republican, and therefore an overall amplification of the local campaign.

Deduction 3. Overall spending in the race helps the GOP candidate, but not the Democratic candidate.

Intuition. Total candidate spending shifts the voter's focus from the national to the local- thus giving the Republicans a boost.

Results. This deduction is confirmed - and the results are the most robust. Every $100,000 in total spending enhances the Republican candidate's position by 0.10%. This result was found to be statistically significant at the 98.6% confidence level. 26.7% of the total variation in the polls is explicable by this final model.

On the flip side, total spending was found to have a slightly negative effect on the Democratic position. However, this result was significant only at 13.3%, which means that the most reasonable conclusion is that total spending does not help the Democrat (as opposed to actually hurting the Democrat). However - again, I am surprised by this result. I was expecting the Democrats to be aided in a statistically insignificant way. I was not expecting them to be harmed in a statistically insignificant way.

Conclusions

Overall - these results are consistent with my "floating" hypothesis. What we see is that - as the candidates become more engaged - the GOP position is enhanced. This is the case even for when Democratic candidates become engaged. Democratic spending is actually helpful to the Republican. The reason for this is not because Democratic ads implicitly endorse Republicans - the causal logic I offer here is that, as the Democrat engages, the Republican engages in turn. All in all, the district's attention shifts from the national to the local. This enhances the Republican position.

Let me close with two caveats. When I discuss a transition from the "national" to the "local" - I do not mean a transition from a discussion about Iraq to a discussion about bridges and roads. It means more than just bridges and roads. I mean "local" in the sense that Emory's Alan Abramowitz implies in a 1985 article on the effect of national issues in local House races: "It is often argued that elections for the House of Representatives are dominated by local issues. However, national issues can become local issues if these issues are raised by the local candidates."

This relates to a general complaint I have had of media analysis - they talk about the "nationalization" of local House elections. It is probably better understood as a "localization" of national issues. There are national issues being discussed in these races, but they are often in reference to the locality and the specific candidates on the ballot. The context of these national issues is local -- that campaigns are trying to tie national issues to local personalities.

This transition from the "national" to the "local" (or, perhaps better put, from the "abstractly national" to the "specifically local"), according to my analysis, helps the Republican Party. However - and here comes the second caveat - this analysis is only partially complete. We need to take a close look at individual voter attitudes to finish the job. The idea here is that money changes frames of political reference, which in turn enhances the Republican position. We have evaluated money's effect on the Republican position - which is consistent with the overall hypothesis, but not determinative of it. In an ideal situation, we would track changes in money to changes in orientations, and then changes in frames of reference to changes in Republican support. Unfortunately, we lack the data for this transition. What would complete the analysis is a survey of voter orientations in races where much money has been spent and where only a little bit of money has been spent. Unfortunately, Majority Watch does not really provide that kind of evidence. They only provide a "Voter Motivation Index," which does speak very well to psychological orientation.

Incompleteness aside, we have nevertheless found strong evidence that supports the "floating" hypothesis. All three of the deductions that flowed from the hypothesis were confirmed. This evidence supports the idea that the more money that is spent in congressional districts, the more the voters turn their attentions from the generalities of the national landscape to the specifics of the local race, the more the Republican position is enhanced.

Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.