Following up on our earlier review of the 2012 Japanese House of Representatives election, the graph below shows the patterns of two-party competition in the nominal tier, consisting of 300 single-seat districts, won by plurality. The graph plots a district’s winner’s vote percentage (vertical axis) against that of its runner-up (horizontal axis), showing the four most common district dyads. If a district featured two parties getting all of the votes, its marker would be on the diagonal line where the top two vote shares sum to 100%. There are no districts on this pure two-party line, though it is immediately obvious that almost all of those that are were close to it were won by the Liberal Democratic Party (LDP), with the Democratic Party of Japan (DPJ) second.
As if we needed more evidence of the slaughter suffered by the DPJ, the graph makes clear how much worse it could have been. Look at the two local-regression (“lowess”) curves, which simply track the relationship between first and second vote shares in districts won by the LDP or DPJ (in both cases, regardless of who the runner-up was). Note how the DPJ line sags in the zone where its candidate has won a district with a vote percentage in the low 40s, and the runner-up was in the 30s. By contrast, the LDP curve becomes almost flat at about 48% throughout a range of second candidate shares running from around 25% to over 40%. ((The difference between the means of winning vote percentages for these two parties is significant (p=.02).)) This, combined with some other evidence in this graph that I will discuss below, suggests that the DPJ actually benefited from the presence in this election of a third-party challenger, the Japan Restoration Party (JRP). A substantial share of the districts that the DPJ held were those in which the JRP was a strong third, and possibly where the JRP siphoned enough votes from the otherwise surging LDP to prevent the latter’s victory in the district.
Further supporting the JRP’s squeeze on the LDP from the right of the political spectrum is the cluster of districts in the lower left of the data plot in which the LDP defeated a JRP runner-up by a big margin, but with a winning vote percentage of under 50%. These are districts in which the DPJ finished third; the data I am using do not let me distinguish how many of these were already held by the LDP before the election and how many were DPJ defeats. Nonetheless, we can see that the LDP was competing with the JRP in much of the country even as it was competing to displace the ruling DPJ overall.
In fact, in nearly a fifth of the districts (19.3%) the top two candidates were LDP and JRP; these include 47 won by the LDP and 11 won by the JRP. By contrast, the DPJ won only 27 seats, 25 of which saw the LDP in second place. There were 140 LDP wins where the DPJ was second. If we include districts where the winner or runner-up was LDP-allied New Komeito or DPJ-allied Peoples New Party, we have 172 districts with a dyad that was DPJ (or ally) vs. LDP (or ally), representing a mere 57.3% of all districts. Thus this election seems to have been much less a straight fight between competing governing formations than were those of 2003, 2005, and 2009. ((It is unclear whether we can even speak of two big (potential) governing formations anymore. Maybe Japan now has a dominant LDP, as before, but this time flanked by an ideological right as well as various center-left parties.
More on the New Komeito. The characteristics of its districts are interesting. It won 9 seats; it did not lose any that it contested. Only three of these wins were against the DPJ as runner-up. Others were in districts where the runner-up was the Japan Communist Party (2), the Tomorrow Party of Japan (2), and one each for the New Party Nippon and Your Party. This pattern shows why the New Komeito is a good ally for the LDP: the sorts of district where it has appeal are evidently quite different from the Japanese norm. These might, therefore, be districts the LDP would have trouble winning on its own, so it makes sense to stand aside for an ally. Obviously, without the New Komeito, the LDP would still have a confortable majority. However, it would not have a two thirds majority, unless it could have won a majority of these New Komeito seats on its own.
Further emphasizing the LDP dominance in this election is that of 41 districts in which the winner had over 60%, 36 were won by the LDP. But it could have been worse for the DPJ. Almost exactly 10% of the districts (29) were decided by margins of under five percentage points, and ten of these were DPJ wins over the LDP. On the other hand, 15 were LDP wins over the DPJ; had the DPJ held on to just six of these, it would have prevented the LDP-Komeito alliance from winning the overall two-thirds majority in the House of Representatives.
Finally, I will leave you with some raw statistics about each of the three parties discussed here: LDP, DPJ, and JRP.
Vote share of winning candidate (mean, median):
LDP: 48.8, 46.7
DPJ: 43.1, 44.1
JRP: 43.7, 40.8
[Note: these numbers have been corrected]
Margin of victory (mean, median):
LDP: 21.8, 18.9
DPJ: 11.8, 7.5
JRP: 12.1, 10.0
Obviously, these again drive home the point that the LDP was strong enough most places to win a two-candidate race (though many contests were actually three-way, as the LDP’s big margins attest), while the DPJ and JRP both relied for their few seats on winning contests with relatively strong third candidates, and did so with narrower margins.
Reminder: all of the above points–and the data set that I am using–consider only the single-seat districts; in addition to 300 of these, there are 180 seats allocated by list PR. The PR seats are non-compensatory, and are allocated in several regional multi-seat districts or “blocks” as they are often called in writing on Japan. In the PR seats, the LDP won 57 seats on 27.8% of list votes; New Komeito 22 on 11.9%; DPJ 30 on 15.5%; JRP 40 on 20.5% (data from Wikipedia.)
The data analysis, including the graph, are mine, but would have been impossible were it not for the generosity of Professor Ko Maeda of North Texas University, who made the raw data on the election available on his website.