The Japanese General Election of 2012: Sometimes, Lucky is Better than Popular

by Joshua Tucker on December 27, 2012 · 5 comments

in Campaigns and elections,Comparative Politics,Election Reports

Continuing our series of election reports, the following is a post-election report of the 2012 Japanese General Election from political scientists Steven R. Reed (Chuo University), Ethan Scheiner (UC Davis), Daniel M. Smith (Stanford University), Michael Thies (UCLA). It is quite a bit longer than some of our previous reports, so interested readers are especially invited to click through to the full post.

*****

A more detailed series of analyses of the election will appear later in Japan Decides 2012 (Palgrave, 2013), a volume edited by Robert Pekkanen, Steven Reed, and Ethan Scheiner in which contributions by each of the authors of this post – as well as many others – will appear.

On Sunday, December 16th, Japan’s Liberal Democratic Party, which had been swept from power in August 2009 after more than half a century of dominance, roared back with a landslide of its own.  Entering the election with only 118 of 480 seats in the House of Representatives, the lower house of the National Diet, the LDP emerged with a stomping majority of 294, and together with its long-time coalition partner, New Kōmeitō, surpassed the two-thirds threshold needed to override vetoes from the upper house, the House of Councillors (where the coalition lacks a majority—at least until the 2013 upper house election).  The incumbent Democratic Party of Japan (DPJ), which had taken power by winning more than 300 seats in 2009, retained only 57 this time, just barely managing second place after three difficult years in government.

In a previous post, Christina Davis did a terrific job explaining why the ruling party was so unpopular, and considering the policy implications of the LDP’s return.  Here, we will examine the mechanics of how the LDP found its way back to power.

How did the LDP do it?  How was the party able to reclaim the support of a fed-up Japanese electorate and reverse the 2009 debacle?  Well, it didn’t, actually.  A better description of this latest reversal of fortunes (see Figure 1) is that the LDP was in the right place at the right time—the default winner in a chaotic election.

More specifically, the LDP was the beneficiary of a mammoth loss of popularity of the DPJ, which, rather than generating a wave of support for the LDP, prompted huge numbers of abstentions and the at least temporary disappearance of Japan’s Duvergerian two-party system.  Divisions among the non-LDP core of parties allowed the LDP to win many more seats than it had any right to expect under a heavily plurality-based electoral system.

The Japanese Electoral System and the Changes in Votes and Seats between 2009 and 2012

To be sure, the 2012 election was a harsh referendum on an unloved incumbent government.  In most plurality-based electoral systems (absent a severe partisan gerrymander) even a small uniform vote swing can be amplified into a very large seat swing, because a small change in votes is enough to change who wins and who loses in many districts.  This sort of swing is what happened in Japan in 2009, and it is part of the story in 2012 as well.

But the seat swing in 2012 was exaggerated by two facts.  First, Japan does not use a pure plurality-based electoral system.  The electoral system for the House of Representatives is a two-tiered hybrid:  300 seats are filled by plurality rule in single-member districts (SMDs), and another 180 seats are filled in 11 proportional representation (PR) blocs.  Voters cast two votes—one in each “tier”—and ever since the first use of this electoral system in 1996, a handful of small parties have consistently won a combined 40 or more seats via the PR tier.  These parties also run candidates in SMDs, occasionally winning a seat here and there, but nearly always sopping up (collectively) 15% or so of the SMD votes.  The presence of these candidates, even if they rarely can win in the SMDs, lowers the winning threshold for big-party candidates and can increase the vote-to-seat amplification even further.  The second fact is that, with that threshold lowered, sometimes third parties can even win seats if they are locally strong.  For the first time in Japan, a regionally based third party—the Japan Restoration Party (JRP)—emerged and thrived in 2012.  Table 1 shows the distribution of vote shares and seat shares over the past four general elections. 


The LDP won 294 seats in 2012, compared with only 119 in 2009, a gain of 175.  But in the PR tier, where Duvergerian incentives are absent, and where voters are therefore expected to vote sincerely, the LDP’s gain over 2009 was only 1.6 percentage points of the vote, and only two seats!  This means that the LDP’s 2012 landslide consisted almost entirely of the extra 173 SMDs (from 64 in 2009 to 237 this time) that it won without any real increase in (even relative) popularity from its lowest ebb three year ago.  Undoubtedly, some voters swung back to the LDP in 2012, but overall the party actually lost votes from 2009.  Given the lack of a major vote increase for the LDP, how did it swing so many districts?

There are two reasons.  First, voter turnout declined from a record high (for the current electoral system) of 69% nationally in 2009 to a record low of 59.3% in 2012—and came especially at the expense of support for the DPJ.  The pro-DPJ swing in 2009 won that party a lot of votes and seats from floating, unattached voters (many of whom had plumped for the LDP in 2005).  In 2012, many of these voters became disillusioned with the DPJ, but were unwilling to support the LDP or any other party, so they stayed home. At the district level, that decline hurt the DPJ much more than it hurt the LDP.  The hard-core LDP (and Kōmeitō) vote showed up, and was more consequential this time because so many casual voters dropped out. Amazingly, there were nearly 2 million fewer votes cast for LDP candidates in 2012 than in 2009.  But because abstentions increased by more than 10 million this year, the LDP’s SMD vote shares improved, from 38.7% in 2009 to 43% in 2012.

The second reason for the LDP’s conversion of 43% of the SMD vote into 79% of the SMD seats was the success of the so-called “Third Force” parties.  Third Force parties won a significant number of seats this time (see the shaded cells in Table 1), but more important for the DPJ, they muddied the DPJ’s claim to be the 2nd viable party in the SMDs, and split the non-LDP vote in a way that made it possible to hand the LDP as many as 75 seats it probably shouldn’t have won.

Duverger Unhinged

Aside from the massive set of seat swings in this election, the most striking feature of the results is the shift—at least for the time being—away from the (Duvergerian) two-party system.

Duverger’s Law (Duverger, 1954) states that plurality electoral rules in SMDs should produce two-party competition because voters will avoid wasting their votes on non-viable candidates and parties will avoid wasting their resources by supporting non-viable candidates.  But the logic behind Duverger depends on the assumption that voters know which two candidates are the viable two (Cox, 1997).  Over the past several Japanese elections, the question of which two parties were the viable contestants had become settled.  In 2009, 95% of SMD seats were won by either the LDP or the DPJ, and those two parties finished first and second in 85% of districts.

So normally, the basic choice set for voters would come down to: vote for the candidate of the incumbent government or vote for the candidate of the main opposition party (or abstain, on which more below).  Because the opposition LDP had 57 years of history as a national party, nearly all of it in government, with strong local and candidate-based organizations, it (or its electoral ally, Kōmeitō) was viable everywhere.  The LDP lost most of its SMD seats in 2009, but therefore also had dozens of capable former incumbents anxious to return to the Diet.

The Success of the Third Force Parties

This time, the incumbent government was hugely unpopular, and had just suffered a bitter, very public, party split.  In district after district, the assumption that the DPJ candidate was clearly one of the two with the most realistic chance of winning the seat was under attack.  In some districts, the stronger claim was made by a TPJ incumbent defector.  In others, a new party (the JRP) with strong roots in local politics was well organized and more than credible.  And in still others, parties that had entered the Diet in past elections via the PR tier fielded experienced candidates who were eager to take advantage of the DPJ’s fall from grace and challenge for SMDs to call their own.  In addition to the DPJ, then, Japanese voters in most SMDs had 3rd, 4th, and even 5th choices in this election.  So instead of the multiplicity of choices splitting the anti-incumbent vote, it split the anti-LDP vote.  (The LDP and Kōmeitō, meanwhile, assiduously avoided competing with one another.)

The most successful Third Force party was the JRP, created by popular Osaka mayor Tōru Hashimoto, and led during the election by Shintarō Ishihara, who resigned his post as governor of Tokyo to lead the party.  The JRP was successful for the same reason that most “third parties” are successful in plurality based systems—it had a strong regional base (think about the Parti Québécois in Canada, Lega Nord in Italy, or the Scottish Nationalists, Plaid Cymru, or Sinn Féin in the UK, although the JRP evinces no ethnic or separatist characteristics).  The JRP was centered on the Osaka region in Central Japan, and won nearly all of its 14 SMD seats there.  The party did field PR lists in every region, and managed to take 40 PR seats (22.2%), second only to the LDP, but its success waned as the distance from its Osaka base increased (Figure 2).


The second of the Third Force parties, Your Party (YP), led by former LDP Diet member Yoshimi Watanabe, was also modestly successful.  YP first ran candidates in 2009, and increased its Diet delegation this time from 8 to 18 seats.  Before the election, the JRP and YP discussed the possibility of merging, but this idea was ultimately rejected when Ishihara’s group joined the JRP.  Nevertheless, the JRP and YP only ran candidates against each other in 28 SMDs, and cooperated in the districts where they were not in direct competition.

The last of the Third Force parties, the TPJ, was centered around a group of Diet members who split from the DPJ in July 2012 over the decision to raise the national consumption tax.  The defections did lead to some “coordination” problems for the DPJ.  In the 39 districts in which the DPJ ran a new candidate against an incumbent defector, no DPJ candidates won, and in only two did the DPJ candidate even finish second.  The defectors themselves didn’t do any better.  Of the 67 DPJ defectors who ran for a new party in 2012, only the rebel leader Ichirō Ozawa won an SMD seat.

Despite the JRP’s success, the Third Force candidates did not actually win many more SMDs than usual.  The sum of LDP/Kōmeitō and DPJ single-member district seats in 2012 was still 273 (91%), down a little from 2009, but about even with 2005.  In the PR tier, however, parties other than the LDP, DPJ, and Kōmeitō took a combined 71 seats (39.4%), up from just 17 PR seats (8.4%) in 2009.  Many voters who chose Third Force parties in PR apparently failed to fulfill the Duvergerian expectation to vote strategically in the SMDs, and stuck with those parties in both tiers.  The DPJ had lost its claim to be the obvious SMD alternative to the LDP.

The result was a 2-way, or 3-way, or even 4-way split in the non-LDP vote in many SMDs, which handed the LDP many seats it might otherwise have lost.  Had the DPJ been able to unite the votes cast for all DPJ and Third Force candidates, the LDP would won 75 fewer SMDs and Kōmeitō 2 fewer.  Put together, these lost seats would have pulled the LDP down to 219 seats – rather than the 294 it did win—far less than a legislative majority.  Rather than a two-thirds majority (325 seats), the LDP- Kōmeitō coalition would have had 248 seats, a much more modest win.  In the event, the DPJ won 57 seats.  Combined with the Third Force, it would have had 138.  Combined with the 77 LDP & Kōmeitō seats that might have switched with coordination, the DPJ (or other united non-LDP party) sum would have been 215—no longer a devastating result.  The lack of coordination, a reflection of the failure of Duverger’s Law when voters cannot agree on who are the two most viable candidates, turned a 248-215 seat advantage into a 325-139 shellacking.  (Throughout this discussion, we ignore the small number of seats that went to the perennial also-rans, the Japan Communist Party and Social Democratic Party, as well as a couple of other tiny fringe parties and independents.)

Put differently, the trend toward a two-party system, so clear through 2009 was reversed, at least temporarily, in 2012. And the trend may not resume soon because voters may have trouble deciding whether the best way to defeat the LDP is to vote DPJ or to vote JRP, a classic non-Duvergerian equilibrium situation.

Duverger’s Collapse Has Mostly Been in Urban Areas 

Overall, then, Duvergerian disequilibrium in the SMD tier explains how the LDP turned a mere victory in terms of votes into an epic rout in terms of seats, but the story is more nuanced than that.  The LDP’s base had always been in rural areas.  Prior to 2009, of the 100 most rural SMDs in the country, the LDP never won fewer than 74—and the party stayed afloat in 2009 by winning 42 of the rural SMDs, even when the country was otherwise swinging en masse to the DPJ.  The 2012 election saw a reversion to LDP rural supremacy.  The LDP took 82 of the 100 most rural SMDs.  The LDP also won 70 of the 100 most urban SMDs and 85 of the 100 that were “mixed” (not overwhelmingly rural or urban). It is worth noting, however, that the LDP’s rural wins were much more convincing.  Many more of its rural winners than its non-rural winners took home at least 45% of the vote (see Table 2).


Part and parcel of these differences is the disparity in the number of candidates contesting races.  Duvergerian convergence to two per district took its first step back since the introduction of SMDs in 1996.  In 2009, the effective number of candidates per district across Japan was 2.26.  In 2012, it jumped to 2.99.  The biggest increases came in non-rural districts.  In 2009, the average effective number of candidates in rural districts was 2.17; in 2012 it rose to 2.52.  In mixed districts, the number increased from 2.18 in 2009 to 3.0 in 2012.  And in urban districts the 2009 average of 2.44 swelled to 3.44 in 2012.

Abstentions and Support for the Third Force Parties Were Directly at the Expense of the DPJ

Skeptical readers might raise two questions at this point.  First, mightn’t 2012 simply be the result of a correction to undo the wave that went against the LDP in 2009?  Second, isn’t it possible that abstentions and votes for the Third Force parties were actually at the expense of the LDP, rather than the DPJ, and therefore actually prevented even greater LDP success in 2012?  We find little evidence to support these possibilities.

We run a pair of OLS models in which the dependent variable is the (raw) vote swing from 2009 to 2012 for a given party.  The first model looks at the vote swing for DPJ candidates and the second model looks at the vote swing for LDP candidates (see Table 3).  

The results suggest, first, that, although there was a clear counter-wave to the swing for the DPJ in 2009, there was no such swing in favor of the LDP.  The negative and significant coefficient on the Previous Vote Swing variable in the DPJ model indicates that the DPJ lost more votes in districts that it had gained a particularly large number between 2005 and 2009.  However, the magnitude of the coefficient is smaller and nonsignificant in the LDP model, suggesting that there was no correction back to the LDP in 2012 for votes that went newly against the LDP in 2009.


The results also suggest that votes for the Third Force parties came very much at the DPJ’s expense, with a much smaller effect on votes garnered by LDP candidates.  As Table 3 shows, each vote won in a given SMD for a candidate of any one of the Third Force parties was associated with a loss of nearly half a vote for the DPJ in SMD balloting between 2009 and 2012.  In contrast, each vote for the JRP was associated with a much smaller loss of .21 votes for the LDP; and votes for YP were associated with a loss of .11 votes for the LDP.  Votes for the TPJ had no statistically discernible relationship with the vote swing for the LDP.  (We should note that Matthew Shugart presents tentative evidence that votes for the JRP may have actually hurt the LDP—thus, helping the DPJ.  However, as Shugart points out, his analysis does not take into account the 2009 DPJ vote base or include changes in votes received from 2009.  In future research, we intend to investigate this issue further.)

The other stark finding in the regression analysis is the effect of the turnout decline.  For every extra abstainer in a district, the DPJ candidate lost .89 votes.  In other words, for every 10 extra abstainers the DPJ lost nearly 9 votes!  The effect on LDP candidates is much smaller: .13, and it is only significant at the .07 level.  The implication is that huge numbers of 2009 supporters of the DPJ simply stayed at home in 2012; whereas a much smaller share of abstentions appeared to be at the expense of the LDP.

Summary 

To summarize, we see the return of the LDP as attributable to three inter-related phenomena.  First, obviously, the incumbent DPJ became enormously unpopular.  Second, as a result, turnout declined precipitously, and hurt the DPJ disproportionately—after all, its 2009 triumph had required the votes of a lot of unattached voters.  Third, the discredited DPJ even lost its claim to be one of the two main parties in many single member districts.  Since no other party assumed that mantle—with the exception of the JRP in the areas around Osaka—the non-LDP vote was scattered, and the LDP reaped the benefits, winning many seats it would have been unlikely to have grabbed in a (Duvergerian) two-party setting.

Cited Works:

Cox, Gary W. 1997. Making Votes Count: Strategic Coordination in the World’s Electoral Systems. New York: Cambridge University Press.

Duverger, Maurice. 1954. Political Parties: Their Organization and Activity in the Modern State. London: Metheun.

{ 5 comments }

Patrick Dunleavy December 27, 2012 at 1:52 pm

Although we don’t have district-level data, and there are grounds for believing that in Japan national data may overstate fragmentation at district level a bit, none the less the Effective Number of Parties in terms of votes for the results shown above is
3.83 in the plurality rule contests; and
6.33 in the PR contests (where voters could choose as they wished, free of artificial, wasted vote constraints)
Surely this outcome must raise a huge issue for previous analyses by the authors, especially by Stephen Reed from long back and in many different publications, and by Ethan Schein most recently in his article ‘The Electoral System and Japan’s Partial Transformation: Party System Consolidation Without Policy Realignment’ in Journal of East Asian Studies 12 (2012), 351–379, which looked forward confidently to the ineluctable assertion of Duverger’s Law in the Japan context.
Equally, it must raise a fundamental question mark over the strongly apparent conviction that runs thru the piece above, that somehow the 2012 result is a case of Duvergerian exceptionalism. Arguments that the 2012 result reflects ‘contamination’ of a DL effect by the PR element really look weak when exactly the same levels of contamination applied previously but did not prevent substantial apparent two-bloc polarization at the last election.
It really is about time that American political scientists analyzing overseas countries’ politics stopped seeing things thru ethnocentric lenses and started realizing that there are good arguments now that Duverger’s Law does not apply in any advanced industrial country with established liberal democratic arrangements, except the USA. You can see this argument set out in an accessible and populist way in my recent blog ‘Duverger’s Law is a dead parrot’ at http://blogs.lse.ac.uk/europpblog/2012/06/20/duvergers-law-is-dead/
And you can see it explained in a detailed way with new tools for analyzing party competition, in P. Dunleavy and R. Diawakar, ‘Analysing multi-party competition in plurality rule elections, Party Politics, advance online at http://ppq.sagepub.com/content/early/2011/09/09/1354068811411026 (Sorry – Pay site)
My own view is that Duverger’s Law is now clearly a ‘degenerative research programme’ in Lakatos’s terms – one that increasingly relies on overtly ideological expositions and convictions in order to survive, and that generates no new or interesting research propositions.

Ethan Scheiner December 27, 2012 at 7:50 pm

A very thoughtful response by Patrick Dunleavy!

I certainly was incredibly surprised by the extent of the breakdown of 2-partism in Japan this year. The evidence for the consolidation of the party system around two parties in Japan prior to 2012 was just so overwhelming – and based on such a systematic, over-time trend. It was, if anything, Duverger’s dream scenario. It will be interesting to look more systematically (going well beyond simply the fact that the DPJ was super unpopular) at why the consolidation of non-LDP forces around the DPJ fell apart so spectacularly. This latter question is, in my view, the really interesting one surrounding the election, and one that I have not yet seen or arrived at a good answer to.

(I should add that our comment about contamination was not meant to be a major point in the discussion here. I am a committed de-emphasizer of the impact of contamination. The coordination problems for the non-LDP crowd in this election goes far beyond any sort of contamination issue.)

I wholly agree with the assertion that Duverger has had significant problems in a large number of cases. This is something that Rob Moser and I analyze in detail in our book ELECTORAL SYSTEMS AND POLITICAL CONTEXT (Cambridge, 2012). http://www.cambridge.org/de/knowledge/isbn/item6869944/?site_locale=de_DE

Nevertheless, the past pattern in Japan under its current system was consolidation. I would be shocked if it did not begin the slow steady process away from the current fragmentation again.

Richard Katz December 30, 2012 at 12:39 pm

Your usual excellent work. But your data on the urban, rural, mixed split raises a biq policy and electoral question–especially when one compares your data for the overwhelming LDP loss in 2009, when it won only 42 rural seats, 12 mixed seats and 10 urban seats. It lost a lot of its rural base compared to 2005, but was almost totally wiped out in the tw0-thirds of SMDs from urban and mixed areas.

So, this poses for the LDP the classic dilemma: does it match its policies and campaign rhetoric leading up to the July 2013 Upper House election (and beyond) to consolidate and mobilize its rural base, or does it shift policies and rhetoric to attract swing voters in mixed and rural seats? This assumes that there is a trade-off on issues like TPP, and perhaps public works and the consumption tax, that show a rural-urban split. Any ideas on this?

Do you have any data on voter policy preferences along your urban-mixed-rural split beyond the obvious ones?

Mike Thies December 30, 2012 at 10:46 pm

No decent exit poll data yet, but I think some might become available soon.

I think we learned in the 2007 and 2010 Upper House elections that the key to winning is the rural single-seat districts. So if I had to guess, I’d say the LDP will pander to rural voters at least until after this summer’s election. I think that means at least 3 things:
1. Promising public works – more pork for the farms
2. Delaying any action on TPP
3. Maybe even dragging their feet on the court-mandated redistricting, although if they can get the DPJ to go along with any redistricting plan, that would give them cover.

Patrick Dunleavy December 31, 2012 at 9:25 am

This is a response to Ethan. First, many apologies for spelling your name wrong – and many thanks for the reference to your forthcoming book. We will try to get a copy for review on LSE Review of Books (www.LSEReviewofBooks.com), and if it bears on UK or Europe on our other blogs at Europp (www.Europp.eu) and British Politics and Policy (http://blogs.lse.ac.uk/politicsandpolicy/). I hope that’ll help you spread awareness.
One of the problems with a ‘degenerative research program’ is that folk defend the core hypothesis only, and seek to protect it against the failure of auxilliary predictions by arguing that more time or more evidence will somehow revalidate it. So I was reminded by your closing comment (“Nevertheless, the past pattern in Japan under its current system was consolidation. I would be shocked if it did not begin the slow steady process away from the current fragmentation again”) of Gordon Tullock’s famous prediction that other plurality rule countries (like Canada and India) would get to 2 party politics like the USA, but it would take 200 years.
My interpretation of Japan is that we have a decaying one party dominant system, and that is all we’ve ever had (see my Chapter ‘Rethinking dominant party systems’, in Matthijs Bogaards and Françoise Boucek (eds), Dominant Political Parties and Democracy: Concepts, Measures, Cases and Comparisons (London: Routledge, 2010), pp. 23-44. See also Francoise Bouck’s great new book on ‘Factional Politics: How Dominant Parties Implode or Stabilize’ (Basingstoke: Palgrave Macmillan, October 2012). The extraordinary weakness of the DUJ in government and in 2012 highlights only the fragility of the slightly fantasized partisan consolidation in the late noughties. Observers who are looking confidently for something are all too apt to find it.

Comments on this entry are closed.

Previous post:

Next post: