Is it still Blue North, Green South?

In the beginning, the DPP was an urban party. The Tangwai protests were concentrated in the cities, as were most of the successful opposition politicians. Outside the cities lay vast rural voting populations that the DPP had an extremely difficult time with. Places like Miaoli, Taichung, Changhua, Yunlin, Chiayi, Hualien, and Penghu Counties were places that DPP politicians went to lose. The pro-DPP media lamented that these places were “democracy deserts.” There were two exceptions: the DPP did very well in Yilan and Kaohsiung Counties. These two rural districts were so different from everywhere else that they were labeled as “sacred democratic land.”

The great political scientist Samuel Huntington once noted that revolutions are born in the cities but mature in the countryside. So it was with the DPP. Over the course of the 1990s and 2000s, the DPP slowly built its strength out in those former democracy deserts by cultivating a generation of politicians who could speak the local jargon, eroding the KMT’s traditional patronage system, and nationalizing political competition. By the late 2000s, the DPP had become stronger in rural areas than urban areas.

Sometime during Chen Shui-bian’s presidency, pundits started speaking of a very different north-south divide. Any urban-rural differences paled in comparison to the regional differences between the blue north and the green south, with the Chuoshui River as the dividing line. This north-south divide has featured prominently in our understandings of Taiwanese political geography for the past decade.

However, we may be in the midst of another grand shift. The third regional pattern may see the DPP improve its standing in the north and in the cities but see losses in central and southern Taiwan. Granted, these shifts are just beginning to emerge, and I may be jumping the gun.


Back in the late 1980s and early 1990s, the DPP was much more successful in urban areas than in rural areas. To illustrate how this changed over time, let’s look at some differences between the 1992 and 2004 Legislative Yuan elections. Before the 2010 administrative reforms, most urban areas featured a city surrounded by a county of the same name. The counties varied in levels of urbanization, but all of them were far less urbanized than their neighboring city. So let’s compare how the DPP did in six cities and counties in 1992:

city DPP % county DPP %
Taipei City 35.6% Taipei County 26.0%
Hsinchu City 32.6% Hsinchu County 38.2%
Taichung City 39.4% Taichung County 24.8%
Chiayi City 50.1% Chiayi County 22.9%
Tainan City 38.2% Tainan County 38.6%
Kaohsiung City 35.4% Kaohsiung County 38.9%
All 6 cities 36.8% All 6 counties 30.0%

And here is the same table for the 2004 legislative election (the last to use the old SNTV electoral system):

city DPP % county DPP %
Taipei City 34.7% Taipei County 35.8%
Hsinchu City 25.7% Hsinchu County 36.9%
Taichung City 32.0% Taichung County 35.7%
Chiayi City 30.2% Chiayi County 45.0%
Tainan City 41.4% Tainan County 44.6%
Kaohsiung City 41.4% Kaohsiung County 42.9%
All 6 cities 35.9% All 6 counties 38.8%

In 1992, the DPP did better in the urban areas by 6.8%. In 2004, it was better in the rural areas by 2.9%. That’s a (massive) swing of 9.7%.

What happened? In the early democratic era, elections were far more localized. The KMT could run candidates that were good at talking to rural Taiwanese voters in rural areas, and these candidates were largely unaffected by the KMT candidates speaking a very different language to urban (and often mainlander) audiences in other areas. KMT local factions mobilized the resources of the local governments to contest elections on a patronage logic. Since control of the national government wasn’t really ever at stake – everyone knew the KMT would still be in charge – voters were free to support whoever promised a new road, helped out dealing with government bureaucracy, or simply slipped them NT500 on election eve. DPP candidates lacked access to these sorts of resources and had to compete on ideology. However, since power wasn’t actually at stake, ideological appeals had limited impact. During the 1990s and 2000s, the DPP slowly eroded these KMT advantages. They cultivated politicians who learned to talk about local development as well as democracy and independence. They pushed reforms to make it more difficult for local factions to use local government budgets or farmers associations credit unions as campaign slush funds and to crack down on vote buying. Perhaps most importantly, elections slowly became nationalized. With the introduction of direct presidential elections, everyone started to think of politics more in national terms. This became more urgent after the DPP won the 2000 presidential election, smashing the old assumption that national power wasn’t really at stake. The urban-rural divide was never quite as large in more nationalized elections as in more localized elections. If you compare the first national election, the 1994 governor election, with the 2008 presidential election, the last before the 2010 administrative reform, you find less than half as much of a shift. In 1994, the DPP was 2.1% stronger in the six cities, while in 2008 it was 1.8% stronger in the six counties, for a total swing of 3.9%. However, in the early 1990s, legislative elections were more important in defining how people thought about politics; by the late 2000s, people thought almost entirely in terms of presidential politics.

city DPP % county DPP %
1994: All 6 cities 41.4% 1994: All 6 counties 39.3%
2008: All 6 cities 41.5% 2008: All 6 counties 43.3%


By the end of the Chen Shui-bian era, everyone had noticed the growing north-south gap. Let me show the emergence of this gap using both legislative and executive elections.

North of Chuoshui R. DPP % South of Chuoshui R. DPP % gap
1992 LY 29.8 1992 LY 34.4 4.6
2004 LY 33.8 2004 LY 40.9 7.1
2012 LY 40.0 2012 LY 54.5 14.5
1994 governor 38.1 1994 governor 42.2 4.1
2012 president 37.4 2012 president 51.2 16.8

In the early 1990s, the DPP was only 4-5% stronger in the south than in the north. However, by the 2012 elections, this gap had more than tripled in size.

Again, I think this has everything to do with nationalizing politics. Taiwan identity has always been stronger in the south, but that didn’t necessarily translate into votes for the DPP when national power was not at stake. Moreover, when the electoral contest became more nationalized, the parties could not craft different messages for different voters to the same extent. The things you said in Taipei and Hsinchu mattered in Chiayi and Pingtung. Personalities also mattered. In the 1990s, Lee Teng-hui led the KMT. As the first native Taiwanese president, Lee had a very strong pull on voters with nativist sympathies. When he ran in 1996, he did extremely well in the south. Even when he was not on the ballot, voters could tell themselves that the KMT was a diverse party with people like Lee in powerful positions, thus muddying the parties’ positions. That changed when Lee left office and the KMT was taken over by much more Taipei-oriented politicians, such as Lien and Ma. On the other side, Chen Shui-bian was a son of the south.


So if phase 1 is the urban-rural divide and phase 2 is the north-south divide, are we entering into phase 3 of Taiwan’s political geography? It’s far too early to give a definitive answer, but there are some hints in the presidential polling that echo some of my subjective gut feelings.

You should be very careful in drawing conclusions about regional support from polls. My rule of thumb for subgroups is that I don’t trust any number based on fewer than 200 respondents. Most surveys have about 1000 total respondents, but they cut the sample into five, six, or even seven groups. Some of those groups are inevitably going to have fewer than 200 respondents, and often it is far less than 200. While pundits love to scream about how someone’s support has cratered in central Taiwan between this poll and the poll two weeks ago, I usually just ignore those numbers as noise caused by small sample size.

That doesn’t mean all is lost. While I don’t trust any single poll, we are fortunate to have multiple surveys from the same pollster. If we see the same trends again and again, we might have a bit more confidence in the results. That is, by combining the results of multiple surveys, we can increase the number of respondents in each group and obtain an estimate that isn’t garbage. My favorite two pollsters, TVBS and Tai Li-an (Formosa), have each published four polls in the last two months. I’m going to ignore changes over time and take the averages of regional breakdowns in these two series to see what sorts of patterns are emerging. In both of them, I look at the Tsai-Han head-to-head race. Note that TVBS and Formosa define their regions slightly differently, so we have to look at each pollster’s results independently.

In the charts below, Tsai’s and Han’s results are presented in red and orange, respectively. For context, I have also put data for how the DPP and KMT did in each region in the 2008 and 2016 presidential elections. You will notice that the 2008 and 2016 lines are almost exactly parallel, reflecting the fact that each party was above or below its national average in each region by almost exactly the same amount. The main exception is the KMT in 2016 in the Taoyuan area, where Chu underperformed. Tsai did not overperform in this region; rather, Soong (whose running mate was from this region) did quite well. You should feel free to ignore Chu’s poor performance in Taoyuan as an idiosyncratic feature of that election. The underlying regional patterns were basically the same in 2008 and 2016; both were still in the blue north / green south world.

One final caveat: Do not interpret poll results as being identical to vote numbers. That is, 40% in the polls does not mean that a candidate will get 40% in the election. Turnout matters, and undecided voters can break disproportionately one way or the other. If the red line for Tsai’s poll result overlaps with the green line for the DPP’s 2008 election result, that does not mean that Tsai will get the same percentage of votes that Hsieh got in that region. The point here is to look at regions of comparative strength and weakness, not to suggest that Tsai is at 2008 or 2016 support levels.

The chart for Tsai’s support in TVBS polls is the most striking of the four charts. Tsai’s strongest areas are still in the south, but the difference between the north and the south is dramatically smaller. Relative to the DPP’s previous support, Tsai is underperforming a lot in the south, underperforming a little in central Taiwan, doing quite well in Taoyuan, and doing extremely well in Greater Taipei.  Instead of a dramatic north/south divide, we have a much flatter pattern with Tsai doing about equally well in Taipei and the south while getting around 5% less support in the Taoyuan and Taichung areas.

The Formosa polls show roughly the same pattern, though with minor differences. Compared to the TVBS results, the Formosa polls show Tsai doing a bit better in Taichung and a bit worse in Taoyuan. Formosa breaks Greater Taipei apart into Taipei City and New Taipei City, and there is a big difference between the two. Tsai is overperforming in New Taipei by roughly the same amount as in Taoyuan and Taichung. However, in Taipei City, she is massively overperforming expectations. As in the TVBS polls, while Tsai is strongest in the south, this is also where she is underperforming the 2008 and 2016 elections by the greatest margins.

Overall, there is a pretty clear picture of Tsai as a creature of Taipei. The further south she gets, the less popular she is (relative to previous performance).

The charts for Han present a faded mirror image of the ones for Tsai. That is, you can see the same trends in reverse, but those trends seem a bit milder. Han is doing better in the south than in the north, especially compared to the 2008 election results. He is also clearly underperforming in Greater Taipei, specifically in Taipei City. In absolute terms, Han is stronger in the north than in the south, but in relative terms it is the opposite.

I usually think that elections are more about the incumbent than the challenger, but Han is such a powerful character that it is tempting to think that he is as responsible as Tsai for this new, flatter political geography. Just as Tsai is the most Taipei-oriented leader the DPP has had in years, Han is far better at speaking to southern voters and local factions than Lien or Ma ever were.


Who cares if the political geography is shifting? A vote is a vote, no matter where it is cast. For one thing, a shifting political geography reflects a shifting set of governing priorities. More practically, these shifts might have a major impact on the legislative races. While legislative races still have a local component, the vote results in the past three legislative elections have closely followed the presidential results. If the presidential race shifts, it will pull up or drag down legislative candidates in different areas.

The DPP had a large advantage in the south. If Tsai underperforms in the south, it probably won’t matter too much. The DPP has a large enough cushion that it can probably absorb some losses and still maintain majorities in the legislative races. However, as you move northward to central Taiwan, that is no longer the case. In central Taiwan, the DPP had a slim advantage in 2016, but prior to that the KMT generally had the upper hand. The DPP is probably looking at moderate-sized declines in its support in central Taiwan, and that might well tip the balance back to the KMT. (This certainly helps explain why the DPP seems scared about its two seats in Yunlin.) New Taipei has traditionally looked similar to central Taiwan, but Tsai seems to be a bit more popular in New Taipei than in the Taichung area (relative to previous results). This difference is small, but it might be crucial. Depending on where the tipping points eventually fall, it is possible that the DPP could lose seats in central Taiwan while retaining them in New Taipei. Retaining seats in Taoyuan might be more challenging, since the DPP starts from a lower baseline of support in Taoyuan than in New Taipei or Taichung. It still seems strange to me that Taipei City could possibly a tossup area. Taipei has been a blue stronghold for so long, and it was disorienting to see it as nearly a 50-50 city in 2016. I had expected that the DPP’s Taipei seats would be among the easiest for the KMT to win back, but these results suggest that may not be true.

I am still thinking through the impact of the shifting political geography on the legislative races. For now, I simply want to make the point that the regional patterns might not be the same in 2020 as they were in 2016.

6 Responses to “Is it still Blue North, Green South?”

  1. cassambito Says:

    Could the geographic shift be partially explained by progressive vs. conservative preferences of the electorate?

    Perhaps urban areas tend to be more liberal, and that’s how Tsai’s drawing this new kind of support, whereas rural areas folks tend to like Han’s conservative discourse.

  2. cassambito Says:

    (sorry, hit send button by mistake. Continuing..)

    If the above hypothesis is correct, Tsai might be in trouble because a significant part of DPP’s newly gained urban support may not vote on election day if they’re young.

    • frozengarlic Says:

      The progressive / conservative hypothesis seems plausible, but we already knew that young people prefer Tsai so the regional story doesn’t really suggest any additional vulnerability.

  3. Red Says:

    In addition to Han being popular with the south (by KMT standards,) there is also the conservative issue, as mentioned above. Many southern greens are socially conservative and aren’t happy with the Tsai and DPP push for gay marriage – i.e., presbyterian Greens. Also, many such southern deep greens consider Tsai insufficiently pro-independence. They are the William Lai greens. They would never flip blue, but their enthusiasm in coming out to the polls to vote for Tsai may be muted.

  4. Peger Says:

    With respect to central Taiwan, it’s probably worth noting that the DPP already underperformed there in terms of legislative seats in 2016 – Changhua 1, Taichung 2 & 8, and both Nantou seats went to the KMT in spite of Tsai’s strong performance in those districts. Winning DPP candidates underperformed her in a number of districts in the region as well, and I’d imagine they’re in a lot of danger in January.

  5. Jerry Says:

    Thanks for the interesting analysis once again!

    The only thing I am slightly skeptical about are the conclusions drawn from Tsai’s support in the south from TVBS polls. Even though TVBS tends to do fairly well in national polls, they have consistently underestimated DPP support in southern Taiwan in previous presidential elections. If we average the last four TVBS polls right before the 2012 presidential election, they had Tsai ahead by 8% in 雲嘉南(Yunlin, Chiayi, Tainan) and 5% in 高屏澎(Kaohsiung, Pingtung, Penghu). On election day, Tsai was ahead by 15% and 10% in these two regions, respectively.

    I also believe that Taiwan is in the middle of another regional political shift (Taichung seems to be very pro-Han compared to the north), but guess we need more data to back the argument up.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: