Is marriage equality a cleavage?

Last week, my colleague Wu Yu-shan gave a stimulating talk about changes in political cleavage structures around the world. Most of the talk was about the rise of pre-material cleavages (ie: nationalism) in western industrial democracies, but he also had something to say about Taiwan. He believes that we are seeing the rise of materialist (ie: a left-right cleavage) and post-materialist (ie: marriage equality and environmentalism) cleavages in addition to the old nationalism cleavage. In Taiwan’s political science world, Wu is the major voice staking out this position. The opposite view, that national identity is still basically the only cleavage that matters, has most recently and forcefully been voiced by Chris Achen and T.Y. Wang in their forthcoming edited volume, The Taiwan Voter. It is hard to overstate the importance of this debate. Depending on whether you believe Taiwan has one or multiple important political cleavages, you might come to different conclusions on many of the most central questions facing Taiwan today. Does the KMT need to change its position on China, or is returning to the 1992 Consensus a viable option? Will the NPP be able to encroach on the DPP’s pool of voters? Will it be able to appeal to voters that the DPP cannot? Did the 2016 election mark a fundamental break with the past, or is it merely a deviance from a well-established pattern? Should President Tsai push for marriage equality? Why isn’t President Tsai aggressively pushing for admission to the United Nations under the name “Taiwan”? This question of one or many cleavages gets right to the heart of our understanding of how Taiwanese politics work.

During his talk, Wu presented a fascinating graph, taken from a story on Commonwealth Magazine’s website. In this post, I want to explore what we should and maybe should not learn from this graph.

This graph plots legislators’ positions in the political space along two dimensions. The X axis is the Independence-Unification dimension (with independence on the left), while the Y axis is support or opposition to marriage equality (with support at the top).

I don’t understand exactly what the authors did to produce this graph, but I’ll do my best to explain the methodology. The authors looked at Facebook data from each legislator. They used the two party chairs as anchors, examining people who followed both the party chair and the legislator. (Note: I don’t understand exactly how they used these overlapping followers. However, they presented this part in detail, as if they believed it was the most important thing for us to know.) They examined the “likes” on various posts and put that data into a factor analysis model. The purpose of factor analysis is to condense many variables into a smaller number. If you start with n variables, the model calculates a matrix to multiply each variable by to produce n new variables that are completely uncorrelated to each other. However, these n variables are not equally useful. Some have a lot of explanatory power, while others have almost none. Typically, we throw all the variables that account for less than 1/n of the total variance in the data. They have kept two dimensions, though they did not report how much explanatory power each one had or how many variables cleared the 1/n threshold. The final challenge in factor analysis is naming the new variables. Remember, the algorithm has simply produced new variables that are orthogonal to each other; it doesn’t care what went into them. The researcher typically looks at the coefficients that were multiplied with the original variables and decides on a name. Factor analysis has the veneer of cold, objective data analysis, but interpreting it is actually highly subjective. At any rate, I’m going to assume that the authors made reasonable assumptions and inferences in handling the data. For example, I’m going to assume that the dimensions are appropriately labeled. I’m also going to mostly ignore the possibility that Facebook likes and followers don’t necessarily mirror a legislator’s own positions or even the preferences of that legislator’s constituents.

What are we supposed to see in this graph?

I suspect the first thing people will notice is the position of the two party chairs. Tsai Ing-wen is fairly distant from her party median on both dimensions. On the IU axis, she is in the center of the political spectrum. This looks reasonable; most of us think of her as a moderate on identity and nationalism. The Y axis suggests she is also a bit out of touch with the rest of her party on marriage equality. She is noticeably higher on the plot, suggesting she is a stronger supporter of marriage equality than the average DPP legislator. I think this also fits in with the conventional wisdom. There are a few DPP legislators who are more stridently in favor of marriage equality than Tsai, but there are also a lot of hesitant legislators terrified of angering their socially conservative constituents. So Tsai is moderate on China and somewhat progressive on marriage equality. Hung Hsiu-chu’s position is rather more surprising. Hung is widely known as an extremist on national identity questions. Yet here she is smack dab in the center of the KMT caucus. Further, she has made several statements that indicate she is more pro- marriage equality than the average KMT legislator, yet here she is, again, right in the middle of the KMT caucus. These data suggest that Hung Hsiu-chu is not an extremist. She is actually a nearly perfect representation of the average KMT legislator!

The second thing people might notice is how lonely Jason Hsu looks up at the top of the graph. He is the only KMT legislator firmly in the pro- marriage equality camp. Reporters love to interview him on this topic, and this gives the impression that there is a significant wing favoring marriage equality in the KMT. Nope. Not according to this plot.

Third, there is a relationship between the two dimensions. In the DPP, there seems to be a tradeoff. Extreme nationalists tend to be social conservatives, while social progressives tend to be moderate on identity. Why does someone choose to be in the DPP? It is one or the other. I don’t know why it isn’t both, but it doesn’t seem to work that way. The same relationship also exists to a lesser extent in the KMT. Social progressives are slightly more moderate on identity.

Fourth, the NPP is all located in roughly the same position (though Hsu Yung-ming is slightly less progressive and more nationalist than the other four). It is socially progressive but moderate on nationalism. I think this will surprise many people. The common perception is that the NPP is extreme on both dimensions. Here it simply looks like an extension of the progressive wing of the DPP.

I think those are the obvious things we are supposed to see. What are some of the less obvious things?

First, this is a two dimensional plot, giving the impression that there are two equally important cleavages in Taiwan. However, the second dimension isn’t necessary. A vertical line perfectly separates the blue and green camps.

The authors did not report the eigenvalues of the two factors, which indicate how much of the variance each factor accounts for. We don’t know that the second value was at least 1/n or that the first dimension wasn’t several times as powerful as the first. Maybe instead of a square box, this graph should have been flattened into a short and wide rectangle like this to give a better sense of the actual political space:

If you think about the plot this way, one of the takeaways is the extent to which the DPP has captured the middle ground and the KMT has been pushed back into the far right. I’ll bet the KMT held much more of the middle ground in 2008.

Second, look at that cluster of DPP legislators in the top half of the graph. Notice anything about them? How about if I list all the DPP legislators higher than the top KMT legislator (roughly from top down):

尤美女 You Mei-nu, party list

鄭麗君 Cheng Li-chun, party list

林靜儀 Lin Ching-yi, party list

蔡培慧 Tsai Pei-hui, party list

林淑芬 Lin Shu-fen, New Taipei 2

鍾孔炤 Chung Kung-chao, party list

段宜康 Tuan Yi-kang, party list

邱泰源 Chiu Tai-yuan, party list

吳焜裕 Wu Kun-yu, party list

陳曼麗 Chen Man-li, party list

Kolas Yotaka, party list

蔡英文 Tsai Ing-wen, president and party chair

余宛如 Yu Wan-ju, party list

何欣純 Ho Hsin-chun, Taichung 7

蘇嘉全 Su Chia-chuan, party list

施義芳 Shih Yi-fang, party list

徐國勇 Hsu Kuo-yung, party list

吳思瑤 Wu Si-yao, Taipei 1

That’s 14 party list legislators (of 22 total) and 3 district legislators (of 51). Lin Shu-fen is the only district legislator occupying a clearly pro- marriage equality position. This radically changes the way I look at this chart.

For one thing, as the party chair, Tsai Ing-wen had the final say on the composition of the party list. She seems to have packed it with social progressives. So while she might be somewhat out of favor with gay rights activists for her current tepid stance, most of the strong voices in favor of gay rights in the legislature are there because she put them there.

From another point of view, if you only consider district legislators – the ones who actually go out and win votes – the DPP and the KMT don’t look all that different. The two big parties both cover roughly the same portion of the Y axis. The DPP may be slightly more progressive, but the difference isn’t all that great.

Ignoring the DPP list legislators also makes the NPP stand out. They now occupy a distinctive space on the political spectrum (assuming the second dimension is important). They are basically the only politicians who take a clear pro- marriage equality position before the voters.

One way to think about this is that elected politicians are socially conservative, and this social conservatism probably reflects a cold strategic judgement that full marriage equality is too radical for the electorate to swallow. A different way to think about it is that Lin Shu-fen, Huang Kuo-chang, Hung Tzu-yung, and Freddy Lim all won district elections while occupying this part of the political space, so maybe there wasn’t a marriage equality penalty in 2016. It certainly didn’t seem to hurt the other major politician in the top half of the chart, Tsai Ing-wen. It could be the case that (a) there are plenty of socially progressive voters, or (b) the second dimension simply doesn’t matter. Of course, it could also be the case that the cleavage simply hadn’t fully emerged in 2016.

Still, that vertical line perfectly dividing the space is a major problem for the idea that the second dimension matters. I’ll be more open to the idea when that line needs to be drawn at a 60 degree slope. To me, it looks as though there is still one dominant cleavage line in Taiwanese politics, and it isn’t marriage equality. However, this debate is far from settled.