Thursday, 18 September 2014

Republicans' and Democrats' views on Republicans and Democrats

It being late 2014, time to start thinking about the next U.S. presidential election is of course rapidly approaching. With only a year-and-a-half to go, all the bells, whistles and melodrama of that event are now practically imminent. And while it is not yet clear which Democrat and which Republican will battle it out on the final furlong for presidency, a number of individuals have been put forward as potential competitors.

Helpfully, several of these individuals featured in a recent poll where YouGov asked Americans to voice their opinion on a selection of prominent U.S. politicians. The list of politicians was twelvefold, comprising nine Republicans and three Democrats. The nine Republicans were: Rand Paul, Chris Christie, Jeb Bush, Mike Huckabee, Marco Rubio, Ted Cruz, Scott Walker, and Rick Perry. The three Democrats were: Hilary Clinton, Elizabeth Warren, and Brian Schweitzer.

In each case, respondents were required to state whether they had a "very favourable", "somewhat favourable", "somewhat unfavourable", or "very unfavourable" opinion of the politician under scrutiny; they were also given the option of replying "don't know". The results are broken down along a number of dimensions, one of which happens to be party identification (Republican, Democrat, or Independent). We can therefore get some insight into the following questions. Do politicians who are loved by members of one party tend to be hated less or hated more by members of the other? And conversely, do politicians who are reviled by members of one party tend to be adored more or adored less by members of the other?

The most obvious answer is that if an aspiring candidate is loved by members of one party, she will surely be hated more by members of the other, while if she is reviled by members of one party, she will surely be adored more by members of the other. However, a not-implausible alternative is that objectively better candidates are liked more by members of both parties, meaning that the candidate who is seen as ideal by members of one party will be the one who is seen as least-bad by members of the other.

But according to YouGov's numbers, it is the obvious answer that is correct. The two charts below display, respectively for Republican politicians and Democratic politicians, the relationship between politicians' average rating from Republican respondents and their average rating from Democratic respondents. Average rating was computed by: first assigning '5' to "very favourable", '4' to "somewhat favourable", '2' to "somewhat unfavourable", and '1' to "very unfavourable"; then calculating a weighted average; next dividing by the percentage of respondents who gave a response (as opposed to saying "don't know"); and finally re-scaling as a percentage of the rating of top-rated politician among the particular party's respondents.

The correlation for the first graph is .85 (p < 0.01); the correlation for the second graph is .97 (p > 0.1). A rather important caveat is that an inference from n = 9 is at best questionable, while an inference from n = 3 is bordering on criminal! Incidentally, the reason for dividing by the percentage of respondents who gave a response is to make the figure conditional on having heard of the individual. While some respondents may respond "don't know" because they genuinely hold a neutral opinion, it seems likely that most do so because they have not actually heard of the person in question. For example, in the case of Brian Schweitzer, nearly 70% of both parties' members replied "don't know", presumably because (like me), they didn't know who he was. In terms of the unconditional ratings, Mr Huckabee is the top-rated among Republicans, and Mrs Clinton is still the top-rated among Democrats.

No comments:

Post a Comment