Japan’s first virtual prediction market failed to forecast the Liberal Democratic Party’s landslide victory in the Sept. 11 election, though it did as good a job as most opinion surveys, says economist Hiroshi Yamaguchi.
Between Aug. 11 and Sept. 10, Hatena, a blog-hosting company, sponsored “Hatena General Election,” an online game that allowed participants to bet and trade virtual money on political parties as if they were stocks on a stock exchange. (See JMR’s “Japan Begins Web ‘Trading’ of Political Party Stocks” for background.) During that time, 1,163 users participated in the exchange, performing a total of 15,459 trades.
There was initial concern that Hatena’s virtual election market might run up against Japan’s Public Office Election Law, which has restricted use of the Internet during election campaigns. However, trading continued as originally intended until Sept. 10.
Ultimately, Hatena users predicted that the LDP would get 50.86 percent of the vote, their coalition partner Komeito 8.37 percent, and the principal opposition party, the Democratic Party of Japan, 31.35 percent.
In the actual election, the LDP did considerably better: it won 61.67 percent, or 296 of the 480 available seats. The Komeito came closest to the game’s prediction, winning 6.46 percent (31 seats), but the DPJ did much worse, obtaining only 23.54 percent (113 seats).
One source of the discrepancy is the fact that Japan’s lower house of Parliament is composed of two kinds of districts: 300 single-seat districts in which the candidate with the most votes wins, and 180 proportional representation districts, in which seats are awarded according to the percentage of votes received by each party. Hatena’s predictions are closer when one only considers the proportional districts. The prediction for the DPJ’s total (31.35 percent) comes very close to the actual result of 33.89 percent, but the LDP forecast (50.86 percent) is considerably higher than the actual 42.78 percent of votes received.
Hiroshi Yamaguchi, chief economist for The Japan Center for International Finance and a specialist on predictive markets, said it’s too early to say how well Hatena General Elections works, based on only one election. Still, he told JMR:
“My assessment is that Hatena General Election seems to be as good a predictor as opinion polls. But it is not a replacement; it works rather as a complement … The problem lies in whether the market mechanism is truly functional or not. A major concern of us before Hatena’s attempt was that Japanese people might not be well accustomed to ‘markets,’ especially virtual-money markets. In this sense, now I am very pleased to see that Hatena General Election showed the market mechanism worked fairly well. In my opinion, the prediction market approach has good potential as a new method to predict political events (and other issues).”
Like Hatena General Election, opinion polls also underestimated the extent of the LDP’s win. On his blog, Yamaguchi names two representative opinion polls: a Sept. 4 poll taken by The Asahi Shimbun that predicted the LDP would take 53.1 percent of the seats, and a Sept. 3 prediction by Prof. Masayuki Fukuoka that the LDP would win 51.7 percent.
That may have been intentional. According to Nozomu Nakoaka, a former senior editor of the business magazine Weekly Toyo Keizai, the major newspapers may have deliberately toned down the findings of their opinion polls prior to the election. He wrote:
Just two days before the election, a small private meeting was held at the center of Tokyo, where journalists, diplomats, scholars, government officials and others attended and freely exchanged their own views about various topics. This time, naturally, the main subject was about the election. One participant, who is well acquainted with newspaper editors, said that the newspapers intentionally reported the conservative forecasts because if they report the most likely results, it would surprise the readers greatly and might influence their voting behaviors. Needless to say, the most likely outcome would be a land-slide victory for the LDP. In a sense, the opinion polls made a correct prediction. Other participants agreed with his comment and analysis.
About David Jacobson
