What to make of the presidential polls?

As anyone who knows a bit about Taiwanese politics will tell you, media polls should be taken with a pinch of salt. Such is the wild variation from one polling organization to another, you could be forgiven for thinking that something smells fishy. I don’t want to go down that route, but will simply note what most people already know: media organizations in Taiwan are (and have always been) pervaded by partisan interests. Although actual ownership of news organizations by political parties was outlawed several years ago (one of CSB’s achievements), various forms of influence continue as strongly as ever. Media are generally not independent from partisan interests, and this is reflected in opinion polls. For as long as I can remember, the variation between polls across the blue-green divide suggests something systematic rather than random. Namely, blue-friendly media consistently over estimate blue candidates’ support, and under-estimate green candidate support. The opposite is probably also true, just that the number of green-friendly media outlets is much smaller.

These ruminations were prompted by Michael Turton’s post on the current poll confusion. In the latest polls, Apple has Ma ahead by 10 points, Now News by 7, TVBS by 6 and China Times also by 7.  These outlets lean towards blue to a greater or lesser extent. By contrast, the strongly green-leaning Liberty Times has Tsai ahead by 2 points. In each of these polls, the percentage of self-reported undecided voters and don’t knows is rather high, on average 20%. Looks like grim news for Tsai, but the Chengchi University/ XFuture election market (which doesn’t have a measure of undecideds) has her trading 10 points higher than Ma and predicts a Tsai win. Being confronted with such a big spread is one reason many Taiwan observers (and voters) put little stall in pre-election polls.

Individual polls are actually fairly consistent; but whether they are consistently right or wrong is open for debate. See for instance the poll numbers published by the blue-leaning TVBS over the last four months.

In fact the blue-friendly media polls over the last four months all say pretty much the same thing. The figures below are the mean levels of support for each candidate (and the proportion of don’t knows) over all the polls by each organization taken since August. All three blue-leaning media have Ma at around 40%, Tsai around 34%, Soong around 11% and don’t knows around 15%. Mean levels of support in the green Liberty Times polls have Soong at a similar level, and Tsai around the same but in fact a bit lower on average. The difference between Liberty and the other polls is in the number of undecideds, which is seven or eight points higher than in the blue media polls. And that seven or eight points is exactly the difference between the level of support found for Ma in the blue polls and the green poll (40% vs. 32%).

TVBS

       UDN

    China Times

   Liberty

Ma

39.9

      40.3

39.9

32.4

Tsai

34.5

      33.0

34.8

31.8

Soong

12.3

      11.3

10.5

12.0

DK

13.4

      15.3

15.0

21.8

Isn’t it unusual to find so many undecided voters? Why should the blue and green media samples find such different proportions of don’t knows? Back in the day there was a theory that green support was systematically under-estimated in polls because people were worried about revealing their preferences for fear of reprisals. That is no longer a factor. A more prosaic answer is that there are non-random sampling errors. The bottom line is that we can’t read too much into the polls and an educated guess is probably as accurate. Either way, these numbers, which are widely publicized in Taiwan, can be spun by each side. Ma can say that he isn’t so unpopular after all, and Tsai can demonstrate her competitiveness while still urging the mobilization of every last green supporter (something the DPP is good at). And, if ~20% of the electorate really is undecided, there are still enough votes out there for each camps’ campaign to make a difference.   

Mail me at jonathan.sullivan@nottingham.ac.uk, follow me on Twitter @jonlsullivan, or access my papers at http://jonlsullivan.com

Advertisements