The xfuture market is, as we all know, reputed to be more accurate than surveys. The xfuture market knows all. Bow down before the amazing predictive powers of xfuture.
I’m on record as not being a big fan of the futures market here in Taiwan, but rather than explain why I think it is flawed from a theoretical perspective, I’m just going to bury them with their own numbers. Welcome to fantasyland.
Most people are looking at the presidential election market. Now Tsai 51.2%, Ma 42.1%, Soong 8.9% looks fairly crazy to me. Tsai might win, but by 9%?? However, it seems there are some people who think this could happen. So instead, I have turned my gaze to the other election. (Yes, there is another election.) I looked at the vote markets (not who will win) for all 73 district races. The following table shows the leader in each race and their margin over the closest competitor as of tonight. I have ordered the table from the blue camp’s best to the green camp’s best districts. For example, New Taipei 11 (Xindian City) was the KMT’s best district, with Luo Ming-tsai 羅明才 leading Kao Chien-chih 高建智 by a margin of 64.4-35.3%.
New Taipei 11 | KMT | 29.1 |
Taipei 8 | KMT | 28.0 |
Lianjiang | KMT | 24.9 |
Miaoli 2 | KMT | 24.6 |
Taichung 2 | KMT | 24.2 |
New Taipei 9 | KMT | 19.3 |
Taipei 7 | KMT | 18.0 |
Jinmen | New | 10.3 |
Hsinchu City | KMT | 10.0 |
Taipei 1 | KMT | 9.2 |
Taichung 5 | KMT | 8.4 |
Taipei 6 | KMT | 7.8 |
New Taipei 8 | KMT | 7.3 |
Changhua 3 | KMT | 6.6 |
Taichung 3 | KMT | 5.7 |
Changhua 4 | KMT | 5.5 |
Nantou 2 | KMT | 4.4 |
Nantou 1 | KMT | 4.1 |
Hualien | KMT | 3.2 |
Changhua 2 | KMT | 1.8 |
Taoyuan 6 | KMT | 1.0 |
New Taipei 6 | DPP | 0.5 |
New Taipei 4 | DPP | 0.8 |
Taipei 4 | DPP | 0.9 |
Taipei 3 | DPP | 1.0 |
Penghu | DPP | 1.1 |
Kaohsiung 3 | DPP | 1.3 |
Yunlin 1 | DPP | 1.6 |
Chiayi 1 | DPP | 1.7 |
New Taipei 10 | DPP | 2.2 |
Miaoli 1 | DPP | 2.4 |
Ilan | DPP | 3.2 |
Taichung 4 | DPP | 4.1 |
Taipei 5 | DPP | 4.3 |
New Taipei 12 | DPP | 4.3 |
Taoyuan 1 | DPP | 4.6 |
New Taipei 5 | DPP | 5.8 |
Chiayi City | DPP | 5.8 |
Pingtung 2 | DPP | 6.3 |
Jilong City | DPP | 6.7 |
Taoyuan 4 | DPP | 7.1 |
New Taipei 3 | DPP | 7.4 |
New Taipei 7 | DPP | 7.6 |
Kaohsiung 9 | DPP | 7.8 |
Taoyuan 5 | DPP | 8.0 |
New Taipei 1 | DPP | 8.1 |
Hsinchu County | DPP | 8.1 |
Kaohsiung 1 | DPP | 8.1 |
Taoyuan 3 | DPP | 8.9 |
Kaohsiung 6 | DPP | 9.1 |
Changhua 1 | DPP | 9.1 |
Taitung | DPP | 9.4 |
Taipei 2 | DPP | 9.8 |
Kaohsiung 2 | DPP | 11.0 |
Kaohsiung 8 | DPP | 11.3 |
Taichung 6 | DPP | 11.6 |
Taoyuan 2 | DPP | 13.1 |
Kaohsiung 5 | DPP | 15.0 |
Taichung 1 | DPP | 17.1 |
Tainan 4 | DPP | 18.7 |
Chiayi 2 | DPP | 19.5 |
Taichung 7 | DPP | 20.0 |
New Taipei 2 | DPP | 20.5 |
Yunlin 2 | DPP | 21.5 |
Pingtung 3 | DPP | 22.4 |
Tainan 3 | DPP | 24.9 |
Taichung 8 | DPP | 25.5 |
Kaohsiung 4 | DPP | 27.4 |
Tainan 5 | DPP | 31.5 |
Tainan 1 | DPP | 34.0 |
Kaohsiung 7 | DPP | 35.1 |
Pingtung 1 | DPP | 35.2 |
Tainan 2 | DPP | 38.5 |
You may have noticed that there are a lot of DPP districts. According to the xfuture market, the DPP will win 52 of the 73 districts. The KMT wins only 20, and the New Party wins the other one. The DPP is predicted to win 5 of 6 in Taoyuan, 9 of 12 in New Taipiei City, and 4 of 8 in Taipei City. Just for reference, they did not win 18 of those 26 seats four years ago; they won a mere two. You know that phrase the DPP loves to repeat about how the KMT can’t cross the Zhuoshui River? The prediction market takes it literally. The DPP wins everything south of the river, and that even includes Penghu and Taitung. There are crazy predictions up and down this list. The DPP is supposed to win Jilong City by a comfortable 6.7%. If that happens, I’ll eat my pink Tsai Ing-wen flag. Maybe the most incredible result is one that gets the winner right. In Taoyuan 6, the KMT is predicted to win by a mere 1%. That could happen, but only if you change the “1%” to “30%.”
Oh but wait, we’re not done yet. There are also 34 party list seats to distribute. According to xfuture, five parties will get more than 5% and be eligible to win seats. The five are the KMT, DPP, PFP, TSU, and … the Green Party!!! I was already skeptical at the TSU, but then you tell me the Green Party is going to get 5.3%. What can I say to that? I’ve already committed to eating my pink flag; I guess if the Green Party goes over 5% I’ll have to promise buy a bicycle and ride it to work for a month. Translating the xfuture numbers to seats, you get KMT 12, DPP 15, PFP 3, TSU 2, and Green Party 2.
The six aboriginal seats restore some measure of sanity, as the DPP’s lone candidate somehow is not favored to win a seat.
Adding up all the seats, we come up with:
KMT | 35 |
PFP | 5 |
New | 1 |
Ind | 1 |
DPP | 67 |
TSU | 2 |
Green | 2 |
If you assume that the first four groups are blue and the last three are green, the two camps have 42 and 71 seats, respectively.
So xfuture is predicting not just a DPP victory, but a DPP landslide of nearly unimaginable proportions. President Tsai wouldn’t just have a majority to work with, she would have a large enough majority that she would be able to ignore most dissent from within her own party. The two parties would effectively be switching positions, with the DPP becoming the party in total command of the government and the KMT completely unable to exercise any check on the DPP’s power. (At least not through legitimate elected institutions.)
By the way, don’t forget that the xfuture market is even more accurate than surveys!
December 22, 2011 at 2:58 am |
I was afraid of that. Back to my usual pessimism!
Great post though.
December 22, 2011 at 3:29 am |
When I read I’m just going to bury them with their own numbers, I thought, being a researcher, by saying so, you must have made a scientific analysis by comparing the past numbers with past results to show that the Xfuture is not that accurate as they claim.
I was expecting a scientific comparison, and ready to accept it if it does show that their credibility is not yet established.
Stunningly, your entire work is to compare their current numbers against your own past impression. The entire article just a personal impression like “how I feel it should be.”
I can’t imagine how that is gonna beat Xfuture’s clams, which is backed by real, confirmed data.
December 22, 2011 at 3:39 am |
You’re right. The technical term for what I did is the “smell test.” Their results just don’t smell right to me.
December 22, 2011 at 4:14 am |
Does 71 to 42 seem plausible to you?
December 22, 2011 at 4:44 am
【Does 71 to 42 seem plausible to you?】
According to XFuture, if the sum in contract is not close to or equal to 100, means it is still fluctuating.
【what I did is the “smell test.”】
Sorry that I was misled. Could have mentioned it in the beginning, ‘cos you can’t bury data with smell.
December 22, 2011 at 3:38 am |
I find the legislative predictions extremely optimistic too for the DPP. However, based upon my long-going observation of the prediction market, some of the seats in your list can be subjected to temporary spikes. For example, Jilong have long been KMT leading by some 5~7%, which I think is reasonable. The current DPP spike might be effected by the KMT’s candidate’s involvement in the Yu Chang case (same can be said for Kaohsiung 2). I believe eventually it will correct itself on these particular case, as it did for Kaohsiung 9, where after the Chen-Chiu debate, Chen’s number spiked, & than drop back to something reasonable.
Another thing I believe is worth mentioning is that if we are to really compare the accuracy of the market vs. the surveys, we would need to have polling every single ridings to determine which one is more accurate. Since surveys don’t do that, I can only rely on my observations of past market vs polls result (from the municipalities to the many by-elections), & the market are more accurate, with the exception of Global View Polling, which we all know is now gone.
Doesn’t change the fact that the results here look fishy. I don’t think the presidential vote is too weird though, but maybe that just my bias. Would it not be nice to have such an overwhelming victory for the green camp for once though?
December 22, 2011 at 6:01 am |
Just curious, is there any data around from the market’s predictions on the presidential & legislative races in 2008?
December 22, 2011 at 6:18 am
I don’t have any hand-on information right now, but I recall they got almost all of the 2009 County Executive predictions rights with the exception of Hsinchu County. I also recall they got almost all of the legislative by-election right with the exception of Huanlien (which they predicted a Hsiao victory) & one other I don’t remember. As for the for Five Municipalities Election in 2010, I think they got Taipei city wrong, but most polls & even Global View got that one wrong too, plus there was the bullet.
I didn’t start frequenting xfuture when the 2008 Presidential Election was happening, but I heard they got the presidential vote percentage almost spot on, same goes for Global View (which is why I respect this particular poll). I don’t know about their 2008 legislative predictions though, but they do have them.
December 22, 2011 at 3:47 pm |
I couldn’t find any data on earlier races on the xfuture site. They seem to have taken it all down. Admittedly, I didn’t look very hard. They did predict the 2008 presidential race fairly accurately (thanks Echo for that reference).
December 22, 2011 at 4:19 am |
The market has spikes, and some of the data points might have been affected by spikes today. However, that’s probably true at any point you look at the market. The overall picture, however, is not created by these surges; xfuture is simply wildly optimistic about the DPP’s chances.
You are right to point out that we have no polls from individual districts. My point is simply that the xfuture data looks so outlandish that it is probably worse than no data at all.
December 22, 2011 at 6:04 am |
As nice at it would be if these predictions came true, they do seem to be a bit…much. I do think at the very least it reflects public perception on the outcome, which will heavily influence turn out. I’d imagine the Pan-Green voter base is energized and excited at the moment, while their Pan-Blue counterparts are rather dejected just given how poorly Ma’s campaign has gone in relation to Tsai’s.
December 22, 2011 at 3:38 pm
My guess is that you are exactly right. This reflects a gap in energy. Or maybe in energy among net users.
December 22, 2011 at 4:47 am |
【My point is simply that the xfuture data looks so outlandish that it is probably worse than no data at all.】
Unfortunately, other than a wishful thinking, you don’t have nothing to back that statement.
December 22, 2011 at 3:58 pm |
Exactly correct. A 71-42 legislature is so far outside what I consider to be the range of plausible results (based on my two decades of intense observation of elections here) that I don’t think any hard evidence is necessary. To me, this is something like Soong and Lin’s insistence that they are going to win the presidential election. The argument is so unrealistic that the only effect is to erode the credibility of the source.
Of course, this sort of argument based on ridicule only works if the audience shares my reaction: that 71-42 is just not plausible. I haven’t heard any predictions of that sort from even the most optimistic DPP fanatics.
We can certainly revisit this after the election.
December 23, 2011 at 1:48 am
Ok, Garlic, lets assume that your smell is correct, that the current situation couldn’t be that far off your past impression.
Then, let me point out another blunder you are committing here.
This issue of xFuture credibility in this thread of discussion arises from a discussion in which we don’t agree on their predictions on how Tsai and Ma perform. I argued that it looks more reliable in the past PRESIDENTIAL elections, which is backed by real data.
Now, you don’t agree with that, so you pulled out LEGISLATIVE elections to argue that their prediction is unreliable, by doing so dismiss the credibility of xFuture.
This sounds like a trick to me. You should have known that the nature of president election is very different from that of legislators. Where’s their prediction on PRESIDENT elections in this smell picture of yours ?
Back to the legislator prediction. As Rust points out, it fluctuates. If you watch their values in the morning and again in the afternoon, you get different results. In areas that two candidates are different with something like 80%:20%, the fluctuation might not be significant. But if it was like 30%:34%, 40%:42% … etc, daily fluctuation might change the final results significantly. This uncertainty is more severe in areas of less interest, ‘cos less people will participate in the prediction, resulting in a smaller sample size.
That is the nature of this method. So simply picking one point and make a conclusion outta its credibility is a wrong way of evaluating it. You will need to understand the method more and do a more thorough analysis to prove your point, even it’s just a smell test.
December 22, 2011 at 5:08 am |
I think Rust is right. The problem is that individual constituencies are subject to random spikes more than something like the presidential election is.
Let’s see how the market’s looking in early January.
December 22, 2011 at 11:54 am |
if it is really “Tsai 51.2%, Ma 42.1%, Soong 8.9%” then the DPP winning a strong majority in legislative should happen.
the presidential, legislative constituency and party votes results in 114 next year for the individual party ( KMT and DPP ) will be very close. within 3% range.
if DPP can win the presidency, they will also come close to winning the legislative imo.
December 22, 2011 at 3:44 pm |
I certainly agree that the DPP could win the legislature. I’ve been saying that for a couple of months. However, a 71-42 landslide seems outlandish to me. Even a 51-42-9 presidential race probably wouldn’t produce such a landslide, because the PFP isn’t running many legislative candidates. All those Soong votes would probably vote KMT in the legislative races, most of which are straight blue-green contests. For the green camp to win 71 seats, Tsai’s vote probably needs to be closer to 58%.
December 22, 2011 at 11:10 pm
DPP to win 5 of 6 in Taoyuan, 9 of 12 in New Taipiei City, and especially 4 of 8 in Taipei City seem way off the mark. DPP could win most of the 50-50 seats in Central and South.
i took a quick look at 2008 legislative results for Taipei City. KMT won very convincingly in all seats expect District 2. DPP need 9% vote swing ( that is a lot ) to win the next closest district this time. now there is the factor of PFP candidates in a few Taipei City districts. if they fare well into double figures, those KMT safe seats might become 50-50.
December 23, 2011 at 12:51 am |
Ballet & Bullet, a Nottingham University blog on politic that recently been focusing on Taiwan’s presidential election, have a post that might help answer our question.
http://nottspolitics.org/2011/12/22/how-far-should-we-trust-the-chengchixfuture-election-market/
I cannot read the journal though! At least I cannot find this particular journal in my university database. Anyone care to take a look?
December 23, 2011 at 4:32 am |
I also cannot read the journal, but I think this following article could be similar judging from the abstract (by the same authors): http://www3.nccu.edu.tw/~ctung/file/xFuture/Taiwan%20election-200911.pdf
December 23, 2011 at 3:12 am |
I want to see Nathan riding the bike.
December 23, 2011 at 3:38 am |
Actually, so do I. Maybe Santa will bring me a Green Party miracle for Christmas.
December 23, 2011 at 4:05 am |
See the trend in Yahoo online polls (Tsai 59 : Ma 29 : Soong 10):
http://tw.news.yahoo.com/2012election-poll/
To my knowledge, Yahoo readers are more middle to light-blueish. The trend has been kept for two weeks. The value is almost a perfect match to the “who will win” values in two contract groups of the xFuture ( Tsai 67 : Ma 27 : Soong 4 ).
Certainly, both Yahoo and xFuture focus on netters.
December 23, 2011 at 4:18 pm |
I think you aren’t looking at this the wrong way so put on your tin foil hat as I explain.
The wealthy DPP supporters are quite leery of overt financial support of their candidates as it could easily effect their private lives and financial interests. So manipulation in a thinly traded market(think penny stocks), would be a useful use of their resources. It allows the candidate to point to something to cheer his voters on while demotivating his opponent’s. If anything you should be taking advantage of such situations for personal financial gain.
Hence we see Ma going to Chiayi to kiss the ring of the guy in charge of most Chinese internet gambling or some such BS.
No one is paying attention to the legislative elections much less betting heavily on them. Therefore a wealthy DPP supporter could easily manipulate a market for his preferred party with little or no repercussion to himself or his business interests in China or Taiwan.
December 23, 2011 at 8:28 pm |
This market doesn’t use money. So wealth doesn’t really matter.
December 23, 2011 at 4:59 pm |
So, after looking at the map, current seat proportions and taking into account the ‘smell’ of this election campaign season I have come to the following conclusoion (after consulting with MT):
KMT – 60 seats
DPP – 40 seats
Aboriginals – 6 seats (all KMT)
PFP & Independents – 7 seats
What do you think Nathan? I can see the DPP regaining seats they lost in the south in 2008 but I can’t see them breaking through in places like Ilan, Hualien, Taidong, Taipei City, Miaoli, Taishang voters or aboriginal voters.
My guess is if the KMT loses established seats it will be mostly in Taichung City, Changhua and maybe a few in New Taipei City.
Your thoughts?
December 23, 2011 at 8:31 pm |
My guess is that things haven’t changed much since I wrote this about a month ago.
December 23, 2011 at 8:41 pm
How about my district of Taichung 6? There have been rumblings that incumbent Huang is really up against it against the DPP’s Lin Jia-lung. From what I’ve seen he started his campaign really late and Lin has grabbed most of the best advertising spots. Also, Ma came to his office twice but each time to see a local Patriarch rather than Huang who some KMT supporters feel has not been sending enough pork back home …
December 24, 2011 at 3:35 am
About Taishung 6, I give the edge to Lin Jia-lung. We all know he ran for mayor before in 2005 against Hu, & despite failing, decided to stick around to build up local support. Additionally, he welcomed Su Chia-chuan to run for mayor despite him having stayed for some seven years already. Both of these facts earn him a lot of kudos.
Now, Su did not win this district. Hu won by about 10,000 votes, but Hu is now way less popular than he was back than. Lin is also a much better candidate in comparison to Huang. I heard that Su Chia-chuan also mentioned a while ago that, according to a DPP poll, Lin is winning Huang by a comfortable margin. Not a big lead, but a comfortable one outside the MoE.
Considering the above, I think Lin will win.
December 24, 2011 at 2:36 am |
Ben & Michael:
I think your conclusion is a little too pessimistic. It might be the case that I am too much of an optimistic instead though, but anyways, here’s my predictions:
Ridings (leaning + safe):
Green: 16 + 19 = 35
Blue: 13 + 14 = 27
Toss-up: 11
Details in Frozen Garlic format here:
https://docs.google.com/document/d/1-kQ8vnCWMU4oKnGjsrvg9akzXtTriZp96JOlRs5oVJg/edit
Party List:
DPP: 17
KMT: 14
PFP: 3
So adding them all ups with the aboriginal seats (6), we got:
Green: 52
Blue: 44
?: 11
Since I think Tsai is going to win the Presidential Election, chance are those toss-up are going to end up in the green camp. Therefore the DPP have a greater chance of getting a majority than the blue camp.
Do you guys think my riding predictions are reasonable? I actually think since all the reader of this blog know a thing or two about elections in Taiwan, maybe we can make our own prediction market!
Who knows? Maybe putting together our opinions can be rather accurate.
December 24, 2011 at 10:02 pm |
Hey Rust,
Thanks for your feedback and opinions. I still think you are a little optimistic but this is one election where it feels like anything could happen so who knows, maybe you’ll be right! I think there is a desperate need in Taiwan for an independent, credible and reliable polling company rather than another prediction market.
December 24, 2011 at 11:38 pm |
There was one, the Global View (遠見雜誌). But it was forced to shut down forever (and the manager,(戴立安) left the company) in September when the poll showed Tsai led by 5-7%.
Note that some time around May, the GV started publishing only rough data, keeping the final data available for subscribers only. People suspected that the GV did that ‘cos Tsai and Ma are compatible, or even Tsai was leading, unlike all other blue media that show Ma led by a large margin. People who paid to see the internal result reported that Tsai led by 7%. And Dai also mentioned in an interview that Tsai led by 5%.
So how accurate the GV poll is? See the polls for 2008 president election listed in this wiki. When other polls show an error of 5~10%, their errors is ~2% — on both blue and green candidates.
Note that the poll data shows up only in the English version of wiki. The Chinese version of that page doesn’t have the poll data.
January 1, 2012 at 10:19 pm |
The market might be a bit optimistic towards the DPP..winning Hsinchu County by 8%? That’s highly questionable to me although DPP currently holds the seat.
Personally I’m expecting 50~52 seats for the DPP and a close Tsai victory in presidential election.
Rust’s idea of making our own market is great, however we only got 3 days left before all discussions on polling are banned.
January 3, 2012 at 12:59 am |
The latest press release from xFuture
http://www.taiwannews.com.tw/etn/news_content.php?id=1803207
KMT 55 DPP 52 Rest 6
Very different from what you posted then, FG. Maybe the market does adjust.
January 3, 2012 at 1:27 am |
I haven’t looked at the numbers since I wrote that post, but I doubt it. I think they are quoting different numbers than I was. I was looking at their prediction for each individual race and then adding up the results. They have another market for overall prediction. If this were a mature market, those two things would roughly match. When I wrote the post, they weren’t close. As several people (notably Echo) pointed out, the individual races are the most unreasonable looking. They have fewer participating and lower turnover. I think only the most passionate people are bothering with individual races and they are voting their hearts, not their brains. While the individual races are extreme examples of the market bias, I think the overall market probably suffers from the same bias, just to a lesser extent.
January 3, 2012 at 1:36 am |
Maybe next time I’ll read the article before writing a reply. It looks like the numbers might have moved a little and pushed several races slightly over the border. However, it still doesn’t look to me like those individual races will add up to their 55-52-6 prediction.
January 3, 2012 at 11:16 pm |
If you look at the predictions for the 73 district races in their press release, you get 24-44-2-3 (KMT, DPP, IND, unclear). The INDs are both blue as is one of the unclear races (Jinmen). If you split the other two between the two camps, you get a blue-green balance of 28-45. That’s a lot closer to what I would consider a realistic best-case scenario for the DPP, but still not quite there.