The Annette Lu effect: Less than she thinks

What impact did Terry Gou dropping out and Annette Lu entering the race have? If one looks only at the Tsai-Han matchup, there is basically no change before and after the September 17 announcements. Here are a few recent before and after polls from the same pollsters. (I didn’t list the Formosa poll, which was conducted on Sept 16 and 17 – so half before and half after Gou’s announcement.)

pollster	date	Tsai	Han	gap
UDN	9.9	44.0	33.0	11.0
UDN	9.21	45.0	33.0	12.0
.
Apple	9.9	45.1	33.3	11.8
Apple	9.22	44.4	32.9	11.5
.
PinView	9.05	39.8	28.3	11.5
PinView	9.12	37.4	30.5	6.9
PinView	9.18	37.6	27.1	10.5
.
Green Party	9.04	47.6	33.8	13.8
Green Party	9.13	48.1	35.2	12.9
Green Party	9.17	49.1	32.3	16.8

This isn’t a whole lot of data, and two of the post-decision surveys (PinView and Green Party) were conducted in the immediate aftermath of the announcements, when people were still adjusting to the shock. Those two surveys show a small bump for Tsai, but that could be a statistical blip. The two surveys that gave people a few days to cool down (UDN and Apple) show almost no impact. The overall trend is stability.

 

Unlike Terry Gou, Annette Lu has no reasonable chance of winning the presidential election. However, including her in the race does have a noticeable impact. When Lu is included, support for both Tsai and Han goes down noticeably. So far as I can tell, the only survey that has asked both the Tsai-Han and then the Tsai-Han-Lu matchup is PinView.

	date	Tsai	Han	Lu
PinView	9.18	37.6	27.1
PinView	9.18	35.4	25.2	6.4

This survey shows Lu having roughly the same effect on Tsai and Han; both of them decline about 2% when she enters.

Let’s turn to Apple’s past three surveys:

	date	Tsai	Han	Lu
Apple	9.9	45.1	33.3
Apple	9.17	37.7	27.5	4.7
Apple	9.22	44.4	32.9

Apple’s graphic in today’s newspaper puts all three of these results in the same time series, but the middle question was different. The survey from 9.17 asked about a three-way race, not a two-way race. Small differences in survey questions can have big effects on the results, and this certainly did. Tsai’s support in the 9.17 survey is about 6% lower than the other two, and Han’s support is about 5% lower. Keep in mind that this poll was conducted immediately after Gou’s announcement, so I think part of these drops in support are just voters trying to figure out what just happened.

The most interesting result from the Apple poll is about Lu’s 4.7%. Who are those people? I didn’t think that Annette Lu had so much popularity. Actually, she doesn’t. Those people are protest votes. The pundits all assume that Lu’s support should come from the deep green part of the spectrum, since Lu is representing the Taiwan nationalist fundamentalist Formosa Alliance party and other deep greens who detest President Tsai. In fact, her 4.7% aren’t a bunch of cranky old ideologues. She gets 4.9% among green party identifiers, 5.4% among non-identifiers, and 4.1% among blue party identifiers. Apple further clarified that she did even better among the light greens and light blues than among the deep greens and deep blues, though they didn’t provide any concrete numbers for this. Most stunning, among respondents aged 20-29, Lu gets 11.0% support. These are people who don’t have any memories of her as VP, much less as Taoyuan County magistrate, legislator, democracy activist, or feminist pioneer. It’s hard to escape the conclusion that people who answer “Lu” aren’t so much expressing support for her as trying to avoid support for Tsai or Han.

Yesterday’s United Daily News poll did not ask about the three-way race. Instead, they also included potential NPP candidate Huang Kuo-chang in their survey.

		Tsai	Han	Lu	Huang
UDN	9.21	45	33
UDN	9.21	39	31	5	7

UDN didn’t provide any breakdown of these results, but there is an interesting difference with the PinView and Apple results. In those, Lu (perhaps unexpectedly) seemed to take nearly as much support from Han as from Tsai. In this poll, Tsai clearly drops by a lot more (6%) than Han (2%). Of course, it is possible that Lu is pulling equally from both major candidates, but Huang’s support comes overwhelmingly out of Tsai’s ledger.

 

So what effect would Lu’s presence on the ballot actually have? This preliminary evidence suggests a much smaller effect that one might expect. While 5% is nothing to sneeze at, these data suggest that those voters are protest voters, not really Lu supporters. They could easily decide to switch elsewhere or simply stay at home. Moreover, even assuming she does get that 5%, it seems to be split fairly evenly between people who would otherwise support Tsai and people who would otherwise support Han. Han probably benefits a bit from Lu’s entry, but the effect is tiny.

If you are surprised by this conclusion, let me note that I am too. I thought Lu would clearly hurt Tsai, and there would be a clear incentive for the KMT to do whatever necessary to get Lu on the ballot.  I actually had to revise the article as I was writing it because the data didn’t say what I thought they did. It could be that I am jumping to conclusions too quickly from too little data. For now, however, the polls say that Lu (a) gets quite a bit of support and (b) that it doesn’t matter very much.