A lecture on populism

This is the most important story in Taiwan electoral politics in recent years. It’s too bad it took me so long to realize what was going on.

The 2020 surge in youth turnout

It isn’t often that I am floored by a single chart, but a few months ago I saw a chart that made me question a lot of things I thought I understood about Taiwanese politics. It turns out that youth turnout was sky-high in the 2020 election. I know some people will say, “Well, of course it was. I could have told you that.” Let me caution you. It wasn’t obvious at all. I don’t care about your personal anecdotes. I heard them all in 2016, and youth turnout wasn’t sky-high then. 2020 was qualitatively different from 2016, and it isn’t obvious why.

Here is the chart that shocked me. In 2016 and 2018, turnout for people in their 20s was somewhere around 55-60%. As the age of voters increased so did turnout, peaking at over 70% among voters in their late 60s. There was a huge gap in turnout among young and old voters. 2020 doesn’t look like that at all. More than 70% of young voters turned out. Turnout increased with age, but not nearly as much as in the past. There is an enormous gap between the lines for young voters in 2016 and 2020, and only a small one among older voters.

Before I discuss these numbers in more depth, let me tell you where they come from. This is NOT an ordinary random sample public opinion survey. This is a study by my friend Chuang Wen-jong 莊文忠 and our intellectual godfather Hung Yung-tai 洪永泰 commissioned by the Central Election Commission, which has a legal mandate to investigate any potential gender disparities. It is normally strictly forbidden to look at actual voting records due to Taiwan’s strict privacy laws. However, for this special purpose, the CEC allowed them to look at some voter rolls. They took a sample of precincts, and then they took a sample of voters from each of those selected precincts. For each voter, they collected age, sex, and whether the voter had showed up to vote. (They did not record how each voter voted. That is not recorded on the voter rolls or anywhere else.) They eventually recorded data for over 137,000 eligible voters, and this should be a pretty darn good representative sample. It is reasonable to have doubts about telephone surveys. Some people don’t have telephones, some won’t answer unfamiliar numbers, some will hang up when they realize it is a survey, and some won’t give you honest or accurate responses. None of those are problems here; you will rarely find higher quality data than this. The full report was published on the CEC website. The CEC commissioned similar studies after the 2016 and 2018 elections. I really wish we could go back in time and see data from 2012, which (I think) was a fairly “normal” election, and 2014, immediately after the Sunflower movement. However, voter rolls are destroyed within a few months after each election, so that is impossible.

Since the official purpose of the report is to study gender differences, we should first look at the differences between men and women. Young women have somewhat higher turnout – roughly 5% higher – than young men. The gender gap shrinks with age, until turnout is roughly the same for voters in their late 60s. Among the very old, men vote at higher rates than women. Overall, turnout in the 2020 presidential election was 76.7% for women and 73.2% for men. We saw similar trends in 2016 and 2018.

I don’t have much to say about gender right now. Eventually I’m going to try to figure out if young women vote differently than older women. Newcomers to Taiwanese politics are always shocked that women are about 5% more pro-KMT than men since the much-publicized gender gap in the United States favors the more progressive party. My suspicion is that older women are much more conservative than younger women (ie: the age difference for voting behavior is much larger for women than men), but I don’t have any hard evidence of that right now. This topic will have to wait for another time.

On to youth turnout. Wen-jong has thoughtfully given us a chart comparing each 5-year cohort’s turnout rate in 2016 and 2020. There’s a big gap for voters under 35.

This chart forces me to rethink some basic assumptions about how Taiwanese politics works. For about a decade, lots of people (usually those who sympathize with smaller parties) have enthusiastically been talking about mobilizing young voters. I have generally dismissed such ideas. I have always assumed that trying to mobilize young voters is a fool’s errand. You can dump a lot of resources into these efforts and it will look productive because the politically motivated youth are highly visible. However, overall youth turnout was pretty miserable, so I’ve always thought that expending too much effort on youth votes was a was of resources. Besides, due to declining birth rates, there aren’t actually all that many youth votes to be gained. Much better to focus your energy on the more numerous older voters who might actually show up and vote.

However, it turns out that you CAN mobilize young voters. In fact, maybe I’ve gotten it backwards. Maybe the way to think about it is that older voters will reliably show up in all sorts of elections, whereas younger voters might vote or they might stay home. If that’s the case, the rational thing to do might be to focus a disproportional amount of energy on the youth vote where you might be able to produce a significant difference.

We know that there are more voters in the older age groups, so how much difference does this surge in youth turnout make? As a careful scholar with a narrow mandate, Wen-jong has wisely declined to tackle this topic. I’m a lot more irresponsible, especially here on my blog, so let’s give it a whirl.

The government publishes population data for every month. The number of citizens over 20 is not exactly the same as the number of eligible voters, but it’s close enough. So we have a pretty good estimate for the number of voters in each five-year age group in January 2020. Multiplying these numbers by Wen-jong’s turnout estimates, we can get an estimate of actual votes for each age group in 2016 and 2020. (Since this is a quick and dirty exercise, I didn’t bother to get 2016 estimates for eligible voters even though everyone was four years younger then. It shouldn’t change things too much.)

age	pop	2016 votes	2020 votes	gap
20~24歲	1511939	876925	1099180	222255
25~29歲	1609628	906221	1155713	249492
30~34歲	1591102	916475	1137638	221163
35~39歲	1960348	1152685	1372244	219559
40~44歲	1977952	1234242	1408302	174060
45~49歲	1773496	1172281	1289332	117051
50~54歲	1812826	1292545	1395876	103331
55~59歲	1827740	1363494	1473158	109664
60~64歲	1658994	1282402	1365352	82950
65~69歲	1385806	1110031	1147447	37417
70~74歲	809388	641035	671792	30757
75~79歲	609088	454989	480570	25582
80歲 & up	822702	496089	515011	18922
Total	15724025	12899413	14511615	1612203

According to this, there were about 1.6 million extra votes in 2020 because of higher turnout. These were highly concentrated among the younger voters. In 2016, people 39 and under only made up 29.9% of the electorate, but this group accounted for 56.6% (about 912,000) of the “extra” turnout. As a result, the share of the under 39 group increased to 32.8% of the electorate in 2020. Nearly a million extra young votes seems like a lot.

But let’s go further. We know from quite a bit of survey evidence that young people were much more likely to vote for President Tsai. Let’s try to estimate how all those extra youth votes affected the final election tallies.

I’m going to use some numbers from the Taiwan Election and Democratization post-election survey. Surveys always ask respondents who they voted for, and post-election surveys always find inflated numbers for winners. (This is true every year, regardless of the winner’s party.) This year was no different. In the actual election, Tsai beat Han 57.7% to 38.6%. In the survey data, Tsai’s victory was an enormous 67.2% to 27.9%. To account for this, I inflated Han’s support by a factor of 1.38, Tsai’s by 0.86, and Soong’s by 0.88. This yields estimates for each of the age groups as follows:

Age group	Soong%	Han%	Tsai%
20-29	3.3%	22.7%	74.0%
30-39	6.6%	27.8%	65.5%
40-49	6.1%	39.5%	54.5%
50-59	2.9%	48.1%	49.1%
60&up	3.3%	45.8%	51.0%

Remember, these are fairly dirty estimates suitable only for blogging purposes. A serious methodologist writing a serious academic paper would massage them in several different ways, and they’d certainly all be different (but maybe not by all that much).

What does this imply about the final vote totals? If you take the 2020 votes and the “extra” votes from the previous table and multiply them by the vote shares in this table, you get the following:

Age group	All votes Tsai	All votes Han	All votes gap	“Extra” Tsai	“Extra” Han	“Extra” gap
20-29	1668702	512158	1156544	349110	107149	241961
30-39	1644687	698353	946334	288798	122627	166171
40-49	1469929	1064253	405676	158625	114847	43778
50-59	1407430	1379799	27630	104487	102436	2051
60&up	2129846	1912653	217193	99674	89510	10164
total	8320593	5567216	2753377	1000694	536568	464126

Tsai won every age group, but she didn’t win older voters by very much. In this estimate, Tsai beat Han by 2.7m votes. More than three-fourths of this margin (2.1m) came from voters under 40. If we look at the “extra” votes, the gap is even more stunning. Compared to what would have happened if people had voted at 2016 levels, Tsai’s margin of victory was 464,000 larger. Of this margin, 88% (408,000) came from “extra” voters under 40.

If turnout had been the same in 2020 as it was in 2016, Tsai would still have won. All that extra youth turnout didn’t ultimately affect the outcome. However, the narrative might have been a bit different. Tsai got 56.1% in 2016 and surpassed that with 57.1% in 2020. Without that extra 400,000 votes from high youth turnout, some of the news coverage would have noted that her vote share had slipped a bit. “Tsai wins, but is less popular!” Han might have cracked the psychologically important 40% threshold. It wouldn’t have made THAT much of a difference, though. After all, it was a landslide either way.

However, there have been elections in which an extra 400,000 votes would have been crucial. Chen Shui-bian won the 2000 election by just over 300,000 votes and the 2004 election by under 30,000. Ma Ying-jeou won the 2012 election by a fairly comfortable 800,000 votes, though one can imagine the narrative would have been significantly different if that margin had been cut in half.

The extra youth turnout didn’t end up making that much of a difference in the presidential election, but what about the legislative election? I know there will be some readers who want to count all those extra 400,000 votes as mobilization triumphs for the New Power Party, Taiwan People’s Party, or some other small party. Let me remind you that there were plenty of small parties around in 2016. My guess is that a disproportionate number of these newly mobilized youth votes supported small parties, but it was probably less than half of the total. I suppose I could use the TEDS party vote estimates the same way I used the presidential estimates, but I’m not that brave. I can swallow a 3% error for an estimate of 55%; it’s a lot harder to feel good about a 3% error for an estimate of 6%.

Anyway, the election outcome wasn’t decided by the party list seats. Legislative outcomes turn on the 73 single seat districts. In this election, the DPP-led coalition won a legislative majority by winning 50 of the 73 SSDs. So how many of those did the green side win because of higher youth turnout?

It’s hard to tell, but let’s do some really shaky calculations. Dividing the “extra” gap by 73, higher turnout produced an advantage of just under 6000 votes for each legislative district. However, that is for the presidential election. It is reasonable to think that a fairly high number of young voters cast their legislative district vote for a small party rather than for one of the two main candidates. I’m going to assume that any green candidate in a normal sized district (ie: not tiny Penghu) who won by less than 4000 votes owes their seat to the surge in youth turnout. Here’s the list:

Lai Pin-yu (New Taipei 12, won by 2780 votes)

I was surprised to find that there weren’t more tight races; I had assumed there would be three or four of these districts. The blue side won a couple of tight races, but the green side didn’t have many squeakers this time. There were a few other districts in which the race might have been close enough if ALL the extra youth votes went to the green candidate.

Huang Hsiu-fang (Changhua 2, won by 4836 votes)

Chuang Ching-cheng (Taichung 5, won by 5272 votes)

Freddy Lim (Taipei 5, won by 5416 votes)

Chiang Yung-chang (New Taipei 8, won by 5597 votes)

Kao Chia-yu (Taipei 4, won by 6706 votes)

Of these, Chuang Ching-cheng and Kao Chia-yu might have been the most vulnerable. Both of them ran in very large districts, with almost 100,000 more valid votes than some of the smaller districts (such as Huang Hsiu-fang’s, Lai Pin-yu’s and Freddy Lim’s districts). There were presumably a higher number of “extra” youth votes in Chuang and Kao’s districts, and I think it is plausible that the surge in youth turnout changed the outcome in these two districts.

In the end, the surge in youth turnout probably didn’t fundamentally alter the election outcome. It was, after all, a landslide. However, in a different year with a much closer election, it absolutely could have been decisive. I will never again dismiss efforts to increase youth turnout.

(Caveat: The number of young voters is about to dramatically decrease. There were 1.51m people in the 20-24 age cohort. There were 1.25m in the 15-19 age cohort and 1.01m in the 10-14 age cohort.)