THE UNDERLYING ROOTS OF OPPOSITION FAILURE IN

JAPAN: CLIENTELISM + CENTRALIZATION = LOCAL OPPOSITION FAILURE

Ethan Scheiner Stanford University

1

Introduction1

Opposition parties’ ability to challenge ruling regimes is an integral component of representative

democracy. Democratic party theory suggests that in times of voter distress, credible alternatives will

challenge the existing order. Yet, democracies exist where, even during such distress, opposition parties

have difficulty selling themselves as credible challengers. What accounts for such failure? Party

competition is a function of many factors: socio-economic structure, economic strength, parties’

capacity to adjust their issue positions, clientelism, incumbency power, and the quality of parties’

candidates, to name a few. The comparative literature emphasizes the first few items but I focus on the

last, the role of individual candidacies, a factor given greater emphasis in the American politics literature.

Especially where party-generated, fixed list, proportional representation (PR) electoral

arrangements do not predominate, candidates are key to parties’ success. Many factors shape parties’

development of strong candidate pools, but parties’ strength at the local level is particularly important.

As Jacobson (1990) demonstrates in the American congressional case, parties holding many local

election seats can more easily find quality candidates for national legislative races and, in turn, are more

likely to find success in such races (58-60). However, under certain conditions, parties will be limited in

their capacity to develop strong candidate pools.

In particular, I argue that in political systems that are centralized and founded on clientelist

politics—systems where, in order to get their jobs done, local politicians must rely on the financial

graces of the central government—a Catch-22 arises. In such systems, parties that are not strong at the

national level will have great difficulty winning local office. Parties holding little strength at the local level

will have a hard time getting their message out and find few strong candidates to run for national office.

In turn, they will have difficulty winning national level seats. I use this framework to address the failure of

1 An earlier version of this paper was prepared for delivery at the 2000 Midwest Political Science Association and 2000 American Political Science Association annual meetings. The author gratefully acknowledges extremely helpful comments from John Aldrich, Shigeo Hirano, Yusaku Horiuchi, Shigeo Hirano, Melanie Hurley, Herbert Kitschelt, Steve Levitsky, Meg McKean, Bonnie Meguid, Scott Morgenstern, Tomoaki Nomi, Robert Pekkanen, David Soskice, Betsey Scheiner, Rob Weiner, Kuba Zielinski, and the WPS reviewer. Thanks to Kojima Aya, Ozawa Akira, Nakamichi Midori, and Tatsumi Mie for their assistance with the telephone interviews. Also, thanks to Meg McKean, Hironaka Yoshimichi, Kobayashi Yoshiaki, Shinada Suguru, Yanai Satoshi, Nakajima Kaze, Iigata Kôichi, Saitô Masamitsu and Niioka Tatsu, without whose introductions most of the interviews here would not have been possible. Great thanks to Susan Pharr for sharing the JEDS data set. Finally, thanks to the National Security Education Program, The Japan Foundation, Sanwa Bank, and Harvard’s Program on U.S.–Japan Relations for funding portions of this research.

2

the opposition to gain a foothold in Japan, where alternatives to the leading ruling party have been

unable to gain much ground.

In the postwar period, a number of cases of party competition failure have been apparent in the

democratic world. However, as new party success has arisen in numerous countries, including formerly

non-competitive party systems such as Italy and Mexico, a spotlight has shone even more brightly on

what is perhaps the clearest case of failure: despite its democratic trappings, Japan’s government has

been dominated by a single party, the conservative Liberal Democratic Party (LDP), throughout the

bulk of the postwar period.

The 1990s were seen as an opportunity to challenge the LDP. Voters were willing to support

the new opposition of the era, and the strongest new party outpolled the LDP in the PR portion of the

1995 House of Councilors (HC) election and came close in the PR sections of the 1998 HC and 1996

and 2000 House of Representatives (HR) races, but the opposition has won few seats in candidate-

oriented district races.2 As I indicate below, much of this weakness in district races is due to the

opposition’s inability to find strong candidates. Unlike the LDP, which dominates most localities, few

opposition politicians hold local office. Parties challenging the LDP, therefore, have only a shallow pool

of quality candidates for national office. To explain much of the opposition’s failure in Japan, then, we

must look to the local level.

What accounts for the opposition’s failure in local elections? I argue that it is in large measure

due to the combination of a clientelist political structure and local dependence on the central government

for funding. The combination of clientelism and local dependence on the center creates strong incentives

for (1) ambitious local politicians—who rely on the image of being able to pull in money and projects

from the center—to ally with the LDP, the party controlling the national purse strings, and (2) voters to

cast ballots in local elections for such candidates.

The Japanese Context

The LDP lost power in 1993. After the parliament (Diet) passed a no-confidence motion and

new elections were held, a coalition of former opposition parties, former LDP members, and new 2 The HR is the dominant branch. The HC and HR are both mixed systems where voters cast two ballots, one for a party in PR, and one for a district-level candidate.

3

parties formed the first non-LDP government since the LDP was created in 1955. Yet, by the end of

1994, the LDP was the leading party in a coalition government, and today is once again Japan’s leading

party, with no credible alternative in sight.

Nevertheless, there is substantial voter anger. The quest for party alternation in power fueled

much of the vote for new parties in 1993 (Kabashima 1994; Reed 1997). The deflated Japanese

economic bubble has dramatically weakened support for the LDP. LDP corruption, and the Socialist

Party’s willingness to abandon its traditional appeals and form a coalition government with its enemy, the

LDP, have led to great mistrust of Japan’s primary postwar parties. The LDP and Socialists had,

respectively, the support of nearly 45 and 25 percent of the electorate (i.e., not just those who voted) in

1958, but, by 1996, this support had dropped to approximately 20 and 2 percent, respectively. Yet,

even as voter independence rose, no strong challenger has mobilized this discontent for more than brief

periods.

From the 1960s until the early 1990s, Japan maintained a fairly stable national party system,

with, essentially, five parties. The LDP dominated the government and the opposition was made up of

the Socialists, the Democratic Socialist Party, Kômeitô, and the Communists. However, what began as

total LDP domination in the early years of the era gradually eroded as changing socio-demographics led

to a decline in LDP support and influence in the cities. In 1993, sitting LDP incumbents split from the

party to form two new parties: Shinsei and Sakigake. New party candidates were extremely successful

in the 1993 HR election, and, with the non-Communist opposition, formed a non-LDP coalition

government. The government’s most noted achievement was changing Japan’s electoral law from the

long-derided Single Non-Transferable Vote in Multi-Member District system (SNTV/MMD) to one

with 200 PR seats and 300 single member districts (SMDs). However, within a year coalition difficulties

led the Socialists and Sakigake to join the LDP in a new government.

The remaining members of the non-LDP coalition government formed the New Frontier Party

(NFP). With 178 HR members, the party appeared well positioned to challenge the LDP, which held

little more than 200. In the 1995 HC election, the NFP outpolled the LDP by 1.5 million votes in the

PR section of the ballot and took home three more PR seats than the LDP. However, it won nine fewer

4

seats than the LDP in the candidate-centered races in the prefectural district component of the HC.3

Similarly, the NFP won nearly as many votes and seats as the LDP in PR balloting in the 1996 HR

election, but did markedly worse than the LDP in the SMD component of the HR race, winning only

ninety-six SMDs to the LDP’s 169.

Party formation continued. Shortly before the 1996 election, moderate Socialist Party members

joined with a small group of new party members to form the Democratic Party of Japan (DPJ). By

1998, internal bickering led the NFP to disband into an array of splinter parties, and the DPJ was seen

as the primary opposition hope. In 1998 the party joined with a number of NFP splinters to form the

centrist, new Democratic Party of Japan (still DPJ). In the 1998 HC election, the party delivered a blow

to the LDP. However, despite winning only two fewer (out of fifty total) PR seats than the LDP, the

LDP won over twice as many district seats. In the 2000 HR election, the DPJ finished only three

percentage points behind the LDP in PR voting. Nonetheless, out of 300 SMDs, the DPJ could run

candidates in only 242, and won just eighty, nowhere near the LDP’s 177. The DPJ increased its seat

total, but still only won 127 (out of 480 total) seats to the LDP’s 233.

Japanese Opposition Failure: The Impact of Low-Quality Candidacies

What explains opposition success in PR races but failure in district ones?

A variety of factors have been used to explain the opposition failure in Japan. For example,

many have argued that the longtime strength of the Japanese economy, LDP policy flexibility, and the

LDP’s ability to make appeals that matched a very large cross-section of the Japanese public all

contributed to the party’s longtime dominance. However, such arguments cannot explain why the

opposition would do so well in PR balloting, but so poorly in district races.

3 The HC, like the HR since 1994, is divided into two tiers. One tier is a national PR district. Before 1983, this tier was one large SNTV/MMD district electing fifty winners at a time. From 1983 until 2001, the tier was a straightforward national PR tier, with each voter casting one ballot for a party. As part of a plan to take advantage of the fact that it had a larger number of attractive or, at least, well-established individual candidates, beginning with the 2001 HC election, the LDP changed the tier so that voters now cast a single ballot for a candidate, but the number of seats awarded to each party is determined by summing the votes awarded to all of the party’s candidates. The HC’s other tier uses SNTV/MMD with the multimember districts coinciding with prefectural boundaries. Only half the HC stands for re-election every three years, so these multimember districts only put up half their seats for reelection every three years as well. As a result, in any given prefecture, there is between one seat and four available to be contested in a given election.

5

Others, such as Cox (e.g., Cox 1997), argue that the opposition’s failure has arisen because of

coordination problems surrounding the number of opposition party nominees running in electoral

districts. Indeed, without looking at the data, one could surmise that the most convincing potential

explanation for the opposition failure in district races would probably be a failure to coordinate: the

opposition is splitting its vote among too many different parties. However, while the opposition was

harmed somewhat by coordination issues that allowed the LDP to win SMDs with fewer votes than the

combined non-Communist opposition in 1996, this problem was far less severe in 2000, occurring in

only twenty-nine out of the LDP’s 177 SMD victories. In short, opposition failure in recent years does

not appear to be due first and foremost to a failure of the opposition to cooperate. Indeed, my analysis

below suggests that the opposition’s failure had more to do with its difficulty finding “quality” or

experienced candidates.

Jacobson’s (1990) work on congressional elections in the U.S. argues that the “quality” or

experience of candidates is very important to a party’s success in winning seats. According to

Jacobson’s definition, “quality” candidates are those with substantial experience in governmental office.

For example, in the U.S., high level state politicians such as state assembly members would be high

quality candidates for congressional races because of the connections they have already established to a

particular constituency and because their time in office establishes them as having had sufficient

experience to justify voters’ faith in them. In American congressional elections, there is a substantial

advantage for experienced candidates over inexperienced ones. For example, according to Jacobson

(1990), experienced candidates challenging incumbents in American congressional races are four times

as likely to defeat the incumbent as an inexperienced challenger, and experienced candidates are also

more likely to win open seat races.4 Ultimately, Jacobson argues that much of the root of Republicans’

difficulty in House of Representatives’ races in the U.S. for many years was the relatively poor quality

(lack of experience) of the candidates who ran under the Republican banner in challenging incumbents

and contesting open seats (61-63).

4 Jacobson is sensitive to the fact that experienced contestants will be more likely to run when they have a good chance of winning. However, even when these other factors are controlled for, the experienced candidate does better than the inexperienced one.

6

Aldrich and Griffin’s (ND) work on the rise of the Republican Party in the U.S. South is even

more directly relevant a comparison to the Japanese case. The Democratic Party dominated the

American South for roughly a century. Aldrich and Griffin find that the Republican Party did

not make serious gains in U.S. congressional races in the South until it had first gained a

substantial number of seats in state legislative races. Moreover, similar to Jacobson’s analysis, they also

find that prior office experience was a key to success among Republican congressional hopefuls.5 The

rise of the Republican Party in the South is of course a fairly recent phenomenon and Jacobson also

notes that in fact the impact of challenger quality has increased in the United States as elections have

become more candidate-centered (1990: 51-57).

This is particularly suggestive for the Japanese system, which is so firmly founded on candidate-

centered politics. As I discuss in greater detail below, Japan maintains a heavily clientelist system and, in

the eyes of many voters, it is of course very important that one’s Diet representative be able to bring in

pork. At the same time, the personal connection is very important as well and, all else equal, voters will

be more likely to cast votes for candidates in whom they have particularly strong personal confidence or

affinity. For the most part, candidates develop this reputation through their prior careers (in particular in

public service), where they distinguish themselves as particularly capable or personally attractive.

Moreover, in many such careers, such candidates become more attractive by developing close personal

ties to those whom they hope to represent. Indeed, most national politicians in Japan have spent years

developing such relationships, to the extent that their personal bonds are often far more important than

party affiliation.

Therefore, it is clearly important for any party that wishes to find electoral success in Japan to

find a large number of “quality” candidates of these kinds. While a celebrity (such as Jesse Ventura in

the U.S.) might be termed a “quality” candidate, the heart of my definition of quality focuses, as

Jacobson does, on political experience. In Japan, the two most obvious means of gaining such

experience involve working in a national ministry or winning local elected office. As the governing party,

the LDP has a close relationship with the bureaucracy and therefore has an advantage recruiting talented 5 As I discuss in greater detail below, as a decentralized system, the U.S. offers fewer incentives for local voters and candidates to affiliate with a party merely because it controls the government at the center.

7

bureaucrats to run for office under the ruling party label. However, while at one time bureaucrats made

up a particularly large proportion of Diet members, national legislators today were more likely to have

gained experience in local political office.6

In Figures 1 and 2, I plot the proportion of HR and prefectural assembly seats won by the

opposition and LDP over the past few decades. As the Figures indicate very clearly, the opposition has

won very few prefectural assembly seats over the years, while the LDP has held a near monopoly. As a

result, the opposition has no doubt been particularly disadvantaged in its efforts to find quality or

experienced candidates to run for national office. The events of recent years—particularly surrounding

the NFP and DPJ, the leading opposition parties of the past ten years—illustrate the disadvantage the

opposition faces without a substantial pool of local legislators from which to draw. The new parties,

which have achieved quick success at the

national level by means of bringing in defectors from existing parties, have been unable to develop any

sort of local base to match their moderate success at the national level.7

Figure 1: Proportion of Legislative Seats Held by the LDP

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997

Year

Pro

po

rtio

n

Proportion National HR LDP

Proportion Prefectural Assembly LDP

Compiled from information in the Japan Statistical Yearbook (various years) and Yomiuri Shinbun’s Bun’yabetsu Jinmeiroku (various years).

6 More than twice as many former local office holders than bureaucrats were elected to the HR in 2000 (based on calculations from Miyagawa (2000)). 7 Another new party, the New Liberal Club (NLC) faced a similar problem in Japan in the late 1970s and early 80s. The party was never able to pull together a base of more than thirty-nine (approximately one percent of the nation’s total) prefectural assembly members throughout the country (Japan Statistical Yearbook), and, despite widespread popularity, it was never able to find many quality candidates to run under its banner at the national level.

8

Figure 2: Proportion of Seats Held by Non-LDP Parties

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997

Year

Pro

po

rtio

n

Proportion National HR Opposition

Proportion Prefectural Assembly Opposition

Compiled from information in the Japan Statistical Yearbook (various years) and Yomiuri Shinbun’s Bun’yabetsu Jinmeiroku (various years). *Note that a chunk of seat holders remain officially independent and therefore the totals won by the LDP and non-LDP parties sum to less than one.

After the 1995 local elections, the NFP only had representation in twenty-seven out of the

forty-seven prefectural assemblies, averaging three assembly members per prefecture, or just over five

assembly members in each prefecture in which it actually had any representation at all. This gave the

NFP a grand total of 4.8 percent (141 out of 2,927) of all of Japan’s prefectural assembly members.

This poor showing is all the more stark when compared to majority or near majority held by the LDP

and its conservative independent allies in nearly every prefectural assembly (with prefectural assemblies

averaging a total of sixty-two seats each). Given this lack of strength at the local level, the NFP could

achieve fairly substantial popularity, but still have great difficulty seriously challenging the LDP.8

Although the NFP's own split up led to the creation of a larger, stronger DPJ, the new DPJ has

suffered from disadvantages much like those of the NFP. Out of the forty-four prefectures that held

local elections in 1999, there was an average of sixty seats per assembly with the DPJ holding 3.9 seats

8 Moreover, assembly members elected for the first time in April 1995 could hardly be considered experienced candidates (although they probably could be called quality candidates) in time for the July 1995 HC election. In short, in terms of its ability to find experienced contestants for the 1995 HC election, the NFP was even weaker than these numbers suggest.

9

per assembly. Considering only those prefectures in which it actually won seats, the DPJ had

representation in forty prefectures, for an average of 4.25 seats per assembly in which it held seats. This

adds up to 6.4 percent of all prefectural assembly seats in these forty-four prefectures, a much smaller

percentage than the DPJ's share of votes and seats in national-level elections.9

When able to run experienced candidates in specific prefectural districts in the HC election in

1995, the NFP was extremely successful in these races and actually won half a million more votes than

the LDP throughout the country in the district level races. Yet, the party was able to find candidates to

run in HC prefectural districts in only thirty-two out of the total forty-seven prefectures, and NFP

candidates won seats in only twenty-two of these. In contrast, the LDP took seats in thirty-one

prefectures (Asahi Shinbun, July 24, 1995). Similarly, in the 1996 HR election, even as its popularity

was on the decline, the NFP was only 4.5 percentage points behind the LDP in terms of PR votes, and

won only ten fewer PR seats than the LDP did. However, with its larger pool of incumbents and quality

new candidates (e.g., pulled from local politics), the LDP crushed the NFP in SMDs. Out of just over

100 new NFP candidates running in Japan’s 300 SMDs, only twenty were victorious in 1996 (McKean

and Scheiner 2000). And these twenty included eleven former local politicians, along with three central

government bureaucrats who had been given field assignments in the prefecture, one former television

newscaster, and one former baseball star from the region.10

More systematic analysis makes the point much more definitively. Analysis of major party (LDP

and DPJ), new candidates—i.e., those who had never previously held HR office—from the June 2000

Lower House election suggests that quality candidacies are absolutely key to party success. To begin

with, simply being a relatively new opposition party is usually a disadvantage, as most candidates of

such parties will be challengers. Out of 242 DPJ candidates, 139 had never held HR office, in contrast

to only fifty-six newcomers out of nearly 300 candidates for the LDP. Indeed, the head of the DPJ’s

electoral strategy bureau explained that the DPJ’s most substantial obstacle to electoral success was the

9 Compiled from various local editions of the Asahi Shinbun the day after the April 11, 1999, unified local election. 10 Presumably, because of the high profile nature of their professions, professional athletes and newscasters, too, might be considered quality candidates.

10

very small number of potentially strong candidates it could run in SMDs.11 This lack of national-level

political experience was a great disadvantage to the DPJ, as new candidates tend to show only limited

success, with twenty-six (18.7 percent) winning for the DPJ and nineteen (33.9 percent) for the LDP in

2000.12

At least as important, though, the LDP was able to run a substantially greater number of

“quality” or experienced new candidates. Similar to Jacobson’s definition, I operationalize quality

candidates in Japanese politics, first and foremost as: (1) former prefectural governors or assembly

members, (2) former city mayor, and (3) a former national Upper House member. In Japanese

politics, there are three other types of candidates whose past experience also make them particularly

strong candidates who are capable of attracting a great deal of support. For this reason, I expand the

definition of quality new candidates to include them. They are: (4) former national level bureaucrats, (5)

former television news reporters, and (6) candidates who “inherited” an SMD from a relative.

National level bureaucrats are typically held in such high esteem so as to attract greater support

merely on the basis of their bureaucratic reputation. In addition, as highly esteemed sons of the

community, national level bureaucrats will be likely to maintain significant connections that will help

facilitate greater electoral success in their chosen district. Television reporters maintain instant credibility

as candidates as a result of their exposure and appearance of expertise. Finally, in Japan the practice of

inheriting districts is quite common, particularly within families of very successful politicians. Given the

district’s strong links to the retired/deceased incumbent, as well as the well-established organization

associated with him, in most cases such inheritance is equivalent to a guarantee of victory. To give the

most obvious recent example, when Prime Minister Keizo Obuchi died in 2000, his twenty-six year old

daughter inherited his district for the 2000 election and won with more than 76 percent of the vote.

It is here that the LDP’s advantage becomes substantially clearer. Using this operational

definition, thirty-two (57.1 percent) LDP new candidates were quality candidates, compared to twenty-

three (16.5 percent) DPJ newcomers. In SMD races in which it fielded a “non-quality” candidate, the

11 Interview, December 18, 1998. 12 Data from the 2000 election were compiled from http://www.yomiuri.co.jp/election2000/main.htm and http://www.asahi.com/senkyo2000/index.html.

11

DPJ won only 13.8 percent (sixteen out of 116), not far off from the 20.8 percent (five out of twenty-

four) won by the LDP’s “non-quality” candidates. In contrast, the LDP and DPJ each won nearly 45

percent of races in which it ran a “quality” new candidate.

My emphasis here, though, is on the importance of candidates with local level experience.

Here, too, the LDP has a major advantage. Only approximately 10 percent of DPJ new candidates had

held office either as a city mayor or prefectural assembly member, in contrast to more than a quarter of

all LDP new candidates.13

Probit analysis of new candidate success in SMDs in the 2000 election demonstrates the

importance of these figures more systematically. My data set is the same as above, all new candidates

for the LDP and DPJ in the 2000 election. The dependent variable is coded 1 if the new candidate won

and 0 if she lost. The principal independent variables in the model relate to the “quality” of the new

candidate. I expect the coefficient on the “quality” variables to be positive and significant, indicating that

quality new candidates are more likely to win than other new candidates.

I also control for a number of other variables. (1) Ran in 96: whether the candidate ran in the

same district in 1996. Technically, a number of these “new” candidates are not entirely new—they may

not have held HR seats in the past, but they may well have run losing campaigns for HR office before. A

number of them ran and lost in 1996, the first election under the new electoral system. I expect that

candidates who ran before (even if they lost) will have created something of a base and will be more

likely to win than those who have not yet run in the district.14 (2) LDP: whether the candidate is an LDP

(coded 1) or DPJ (coded 0) member. (3) Weak Inherit: whether the candidate had a close relative

who was an incumbent in the district going into the 1996 election, but did not win the new SMD seat.

(4) Urban: a three-point scale indicating the district’s level of urbanization. A score of 1 indicates very

rural districts, 3 indicates very urban districts, and 2 indicates something more mixed. Voters in urban

areas tend to be more issue-oriented, and their ties to long-term incumbents are typically looser than for

voters in rural areas. Moreover, displeasure with the status quo appears substantially greater in urban

13 No former prefectural governors were new candidates for the DPJ or LDP in 2000. 14 Also, presumably the party’s willingness to run them again suggests some level of strength on the part of such candidates.

12

areas (Scheiner 1999). Therefore, I expect that new candidates will be more likely to win in urban

areas. (5) Incumbent Opponent: coded 1 if the new candidate faced an incumbent in the district and 0

if not. (6) Major Incumbent Opponent: coded 1 if the new candidate faced an incumbent from either

the LDP or DPJ and 0 if not. New candidates facing incumbents, especially those from major parties,

will be less likely to win.15

Table 1: Probit Model of New Candidate Success in 2000 (LDP and DPJ only) N 195 188 Percent Correctly Predicted 79.5 84.6 Goodman-Kruskal ? (PRE) .111 .333 Chi-sq 35.97 35.39 Prob>chi-sq 0.0000 0.0001 Pseudo R-sq .171 0.187 Log Likelihood -87.356 -76.931 Variables Coefficient Coefficient (SE) (SE) Constant -0.925** -1.332*** (0.405) (0.450) Quality 0.657*** -- (0.236) Local -- 0.647** (0.320) HC -- 1.576** (0.675) Bureaucrat -- 1.090*** (0.414) TV -- 1.655** (0.841) Ran in 96 0.362 0.478* (0.269) (0.285) LDP -0.113 -0.361 (0.259) (0.285) Weak Inherit 0.883 1.071 (0.804) (0.795) Urban 0.229* 0.292** (0.139) (0.151) 15 Some readers will note that I leave out one variable that can conceivably have a substantial impact on candidate success: campaign spending. In particular, many note that the LDP’s sizeable war chest allows it to offer greater funding to its candidates than other parties can. I do not add this variable to the model in large part simply because it is extremely difficult to find reliable estimates of campaign spending in Japan. However, even leaving out campaign spending, I feel quite confident in the reliability of my model for two reasons. First, if, as many argue, the spending advantage is primarily one that helps LDP candidates, then the advantage ought to show up in the LDP variable. Second, there may also be an endogeneity issue here. “Quality” and experienced new candidates in general will attract larger sums of money from both private donors and the party. In this way, “quality” represents not just personal experience, but also the greater spending power that accompanies such experience. (Indeed, introducing campaign spending into the model would be likely to add problems of multi-collinearity as well.)

13

Incumbent Opponent -0.331 -0.029 (0.330) (0.355) Major Incumbent Opponent -0.692*** -0.739*** (0.281) (0.292) *Significant in a one-tail test at .05 level

**Significant in a two-tail test at .05 level

**Significant in a two-tail test at .01 level Data on new candidate past careers available at http://www.yomiuri.co.jp/election2000/main.htm and http://www.asahi.com/senkyo2000/index.html. Data on levels of urbanization generously provided by Kabashima Ikuo. Remaining information from data in Miyagawa (1997, 2000).

The results listed in Table 1 clearly indicate the importance of running quality candidates. In the

first column of results, I group all of the quality categories into a single “quality” variable. Quality is

positive and statistically significant: “quality” candidates were more likely to win office. Utilizing the

coefficients from the composite Quality variable model, I find that if all variables are held at their mean,

the probability of a new DPJ or LDP candidate winning office in 2000 was .19. However, holding all

other variables constant at their means, the probability of a new Quality candidate winning office was

over .32.

In the second column, I disaggregate the quality variable into a series of 0-1 dummy variables

that indicate the candidates’ career history more specifically: (1) Local, coded 1 if the candidate was a

former city mayor or prefectural assembly member. (2) HC, coded 1 if the candidate was a former HC

member. (3) Bureaucrat, coded 1 if the candidate was a former member of a national level

bureaucratic ministry. (4) TV, coded 1 if the candidate was a former television reporter. (5) Inherit,

coded 1 if the candidate inherited the district from a close relative.

As expected, the coefficient on every single category of quality candidates is positive and

statistically significant.16 For the purposes of my analysis here, the significance of the Local variable is

particularly noteworthy.17 Former local office holders were, all else equal, more likely to win their HR

race than non-local office holders. Utilizing the coefficients from the model, I find that mean DPJ or LDP 16 New candidates who directly inherited the district from a parent or sibling (Inherit) are dropped from the analysis because all seven cases of such inheritance in 2000 led to the new candidate winning. In short, Inherit perfectly predicted success. 17 It should be noted that Local made up the largest group of quality candidates. In contrast to the other categories, which each made up between 4 and 16 percent of all quality candidates, former local office holders made up over one third.

14

new candidates had a .19 chance of winning their SMDs, but, holding all else at their means, a former

local office holder had a .38 chance of winning.

The results of the remaining key variables are largely as expected. Facing an incumbent from

one of the two largest parties is a substantial obstacle for new candidates. In Table 2, I use the

coefficients from the disaggregated model in Table 1 (second column of results) to estimate the

probability of victory in 2000 for new candidates, according to whether they were former prefectural

assembly members or city mayors and whether they faced a major party incumbent.18 As Table 2

indicates, new candidates who were former local office holders and who faced no incumbent had a

nearly 60 percent chance of victory, as compared to the 7.6 percent probability of winning for non-

quality candidates who faced major party incumbents.

Table 2: Predicted Probabilities of Victory

Non-Local Local

Major Incumbent Opponent 0.076 0.216

No Incumbent Opponent 0.341 0.594

All non-Local “quality” variables held constant at 0 All other variables held constant at their means

Particularly notable in Table 1, the LDP variable is statistically non-significant, which means that

LDP candidates acquired no inherent advantage from their party affiliation. New candidates from the

LDP were more successful than new candidates from other parties, not because they were from the

LDP, but because they had more experience and were higher “quality” candidates than those who ran

for other parties.

This is a very important and non-obvious point. Many have sought to explain the opposition

failure as due to the strength of the economy, LDP policy flexibility, capacity to run more efficient

nomination strategies, or general party popularity. However, my findings here indicate that much of the

LDP’s advantage in national races appears to stem from its greater ability to recruit experienced

candidates. The problem for the opposition is therefore quite clear. It simply does not have a sufficient 18 In computing the effect of being a former local office holder, I set all the other “quality” variable values to 0. However, maintaining them at their means scarcely affects the results at all.

15

number of “quality” candidates, in particular experienced office holders and successful local politicians,

to make credible challenges in the bulk of the national district races.

Undoubtedly, some of the problem the opposition faces is that in many districts it must run

against powerful incumbents, who scare off additional candidates from running. However, as noted

above, in many cases such as the DPJ in 2000, the leading opposition party ran no one at all. In a

number of these cases, the party might have had available a local office incumbent or two, whom the

party could have run in the hopes of drumming up support for the party in the PR race or start laying the

seeds for success in future campaigns in the district. However, according to Japanese election law, the

moment a sitting office holder submits his papers to run for a different office, he automatically loses his

existing position, even if the local office holder’s seat is not up for reelection for a number of years.

Therefore, as it will immediately cost the party a seat at the local level, it is a major risk for a party with

only a small number of office holders at the local level to run any of them in races for higher office (and

therefore lose control over the lower office position). In contrast, parties with a larger number of seats at

the local level can more safely afford to run some of their local incumbents in national level races (even

against entrenched incumbents).

Clientelism + Fiscal Centralization = Opposition Local Recruiting Woes

What shapes parties’ ability to be successful at the local level? The answer is dependent upon

the type of system in which the party participates. In systems where strong policy-based or ideological

links bind voters and parties, local party success is typically based on the extent to which voters

approve of parties’ messages. Where party-voter linkages are founded on the provision of goods and

services, success tends to be based on voters’ impressions of which party’s candidates can best

provide such goods and services.

Of course, this does not suggest that all parties and voters within a given system are equally

driven by the same motivations. Even in a clientelist system, many voters and even parties may act

according to ideological or programmatic principles, rather than clientelist concerns. There is

programmatic and ideological behavior in clientelist systems, just as there is substantial

16

service/patronage-oriented behavior in programmatic ones. Nevertheless, the categorization is of course

based on ideal types. While differentiation exists even within a given system, what I describe for each

country case below is the general pattern used to describe citizen-politician linkages within it.

Perhaps less obvious, the extent to which parties are financially dependent on the central

government to achieve anything at the local level is also extremely important in shaping local party

success. By financial independence, I am referring to subnational governments being given the power by

the central government to find their own local financial sources to support locally devised programs.

(Examples would be extensive freedom in revenue collection and subnational policy innovation.) In this

way, a financially autonomous subnational government is not solely dependent on the center for financial

transfers to fund local programs. Therefore, it cannot be readily punished by a national government that

is governed by a party different from that governing the locality.19

Where localities depend heavily upon the central government for funding, voters’ and

candidates’ relationship with parties will tend to be shaped extensively by the state of politics at the

national level. In contrast, where localities exert substantial financial autonomy, voters and candidates

may be more inclined to affiliate with and support parties, irrespective of their position in the national

party context. For example, holding constant a variety of other factors, Sellers (1998) indicates how

decentralized policymaking has helped create significant local strongholds for the Greens in Germany,

while centrally-led expansion in France has offered little space for the Greens to develop much after

their early period of success.

Neither of these factors—party-voter linkages and degree of local fiscal autonomy—

independently shapes local party success. Rather, they combine to have a substantial effect on such

success. To better help spell out the mechanisms, I provide a typology of their likely combined impacts.

To simplify the typology, I make each variable dichotomous, thereby creating four distinct categories:

Programmatic/Financially-Decentralized, Clientelist/Financially-Decentralized,

Programmatic/Financially-Centralized, and Clientelist/Financially-Financially-Centralized.

19 My emphasis here on financial decentralization and subnational fiscal autonomy is important. Some systems, such as Austria and Mexico, are formally federal bodies where subnational units have various administrative powers. At the same time, they are heavily dependent on the central government for their funding, thereby intensifying the need for subnational groups to develop ties to politicians controlling the central government’s purse strings.

17

Table 3: Party-Voter Linkages and Level of Governmental Financial Centralization: Typology and Predictions for Local Party Success

Level of Governmental Financial Centralization Financially-Decentralized Financially-Centralized

Programmatic/Partisan PREDICTION · Within a given region, each party will receive roughly equal numbers of votes in both local and national level elections. · Within a given region, small parties may even do better in local level races than in national level ones. CASES · Germany

PREDICTION · A party’s success in local level elections will be highly correlated with its general level of party support within the region at the time. CASES · U.K.

Part

y-V

oter

Lin

kage

s

Clientelist/Distributive PREDICTION · Within a given region, there will tend to be a low correlation between a party’s success in local and national elections. · Local and national level party success within a given region will tend to be somewhat idiosyncratic. CASES · U.S., Canada

PREDICTION · Parties that sit in the government at the national level will dominate local level elections. CASES · Italy, Austria, Mexico, Japan

In Programmatic/Financially-Decentralized systems such as Germany’s, where party-voter

linkages are based on ideology and a truly federal system predominates, voters cast their ballots in local

elections without a need to consider national politics (whether national policies or the strength of specific

parties). In addition, the ideological basis of politics leads voters to focus primarily on local issues in

making local election voting decisions. As a result, parties that are strong at the national level have no

inherent advantage in local level elections. Parties typically receive roughly equal numbers of votes in

18

both national and local elections, largely because each voter tends to support the same party at both

levels for ideological reasons.20

In Programmatic/Financially-Centralized systems such as Britain’s, where politics is centralized

but policy- or ideologically-oriented, issue-based voting predominates. Since localities have little power,

politics tends to be nationalized. Therefore, there is little need to focus on local issues and voters are

inclined to cast their ballots based on their views of the national parties at the time of voting. As a result,

as can be seen in the British case, party success at the local level is not directly related to party strength

in the national parliament, but rather voter support of the national parties (Miller 1988: 167, 236).

In Clientelist/Financially-Decentralized systems such as in the U.S., where politics is less

ideologically-based and decentralized, local politics is to a large degree independent of national partisan

conditions and local government is largely autonomous. Therefore, with voting based to a large degree

on representatives’ capacity to provide goods and services—but with this capacity not dependent upon

relations with the central government—voters are more inclined to cast votes in local level races without

consideration to politics at the national level. As a result, in the American case, there is very little

correlation between parties’ success in local and national elections, even within a single state.21

Ultimately, the heart of my argument focuses on the Clientelist/Financially-Centralized type. In

such systems, the combination of clientelism and local dependence on the center creates very strong

incentives for (1) ambitious local politicians—who rely heavily on the image of being able to pull in

money and projects from the center—to ally with the party controlling the purse strings at the national

level (the ruling party), and (2) voters to cast ballots in local elections for such candidates.

I should note that in my discussion of “ambitious” politicians, I am pointing to ambition in

Schlesinger’s (1966) sense, and am not suggesting that politicians who affiliate with non-ruling parties

are somehow unambitious. I am certainly not ignoring the fact that many candidates will choose non-

ruling parties for other reasons, such as ideology. My main point, however, is that, in general, politicians

who wish to increase their probability of winning office, and in fact win higher office in the future, will

have greater incentive to affiliate with the national ruling party. Similarly, none of this is to suggest that 20 See German electoral results at http://www.statistik-bund.de/wahlen/e/index_e.htm. 21 See U.S. electoral results at http://www.stateside.com/snapshot.html#top.

19

voters cast ballots only for local candidates affiliated with the ruling party. Many voters will have

personal, ideological, or policy based reasons for supporting other parties. However, where voters have

reason to value political connections to the central government, in general they will have greater incentive

to vote for politicians affiliated with the national ruling party.

In this way, as Diaz, Magaloni, and Weingast (2000) argue: “The national incumbent party’s

political control of the central government’s fiscal resources drives a wedge between the local incumbent

party and opposition candidate. In equilibrium, the incumbent party maintains its monopoly, but not

because people prefer it. Rather, the incumbent party’s dominance of the center allows it to use the

fiscal system to affect local elections” (4). In such systems, opposition parties, therefore, find great

difficulty attracting candidates and voters for local office. As I argue in greater detail below, the result of

this is that where parties hold little strength at the local level, they will have difficulty getting their message

out and will find themselves with a dearth of strong candidates to run in national level races. In turn, they

will have much greater difficulty winning national level seats.

In short, in Clientelist/Financially-Centralized systems such as those in Italy, Austria, Mexico,

and Japan, where party-voter linkages are founded on service and goods provision rather than ideology,

and localities depend on the central government for funding, voters will be most likely to cast local

election ballots for candidates of parties that are in the government at the national level. Diaz, Magaloni,

and Weingast’s (2000) discussion of voter incentives in Mexico offers a cogent micro-logic behind the

behavior here: where localities depend on the central government for funding, even voters who prefer

opposition parties will be more likely to cast votes for local candidates of the national ruling party

because of the danger that the ruling party would punish regions that throw their support to the

opposition.

Let me sharpen the analysis further. Local dependence on the center is not sufficient to ensure

national ruling party success at the local level. In systems such as Britain’s, where localities depend on

the central government but politics is more “programmatic,” voters cast ballots according to party ties

and ideology. For this reason, parties controlling the national government in Programmatic/Financially-

Centralized systems will not necessarily own politics at the local level. Moreover, similar to the

20

Clientelist/Financially-Decentralized type, voters in Clientelist/Financially-Centralized systems cast their

ballots for candidates they judge to be most capable of delivering goods and services. However, in the

Financially-Centralized type, local politicians rely heavily on partisan links to the central government in

order to have access to these goods and services. For this reason, candidates will have great incentive

to join parties that control the purse strings at the national level. In short, the mechanism involves both

voters and local elites.

Preliminary comparative analysis confirms the expected lack of success non-national ruling

parties have at the local level in Clientelist/Financially-Centralized systems. In Mexico, an

overwhelmingly clientelist system centralized around the presidency, the opposition recently has made

huge strides in the national legislature and has increased its numbers in local elections. Nevertheless, the

Mexican system (and all the patronage in it) continues to be centered on the Mexican presidency. Not

surprising, therefore, the opposition has tended to control at best only 20 to 25 percent of all

municipalities, but, as expected, controls the bulk of the large, more-autonomous ones (Klesner 1997;

Rodriguez 1997). At the same time, with the new PAN Fox presidency, we should see an increase in

opposition strength at the local level.

Austria too is heavily clientelistic and, despite a formally federal republic, the central government

is much stronger than the Lander (Dachs 1996). Unlike Japan, domination by a single party was

consciously avoided by early postwar leaders to prevent a repeat of the authoritarian era (Lauber

1996). Even though there have been two parties (usually jointly) dominating national politics in Austria

throughout most of the postwar era, each has managed to carve up niches for itself in specific Lander,

whereby regional politics is centered on that single party. Party competition has grown in recent years in

Austria, but this shift occurred first at the national level—local voters and elites did not give up their

distributive gravy train until it was no longer running at the national level (Dachs).

At first blush, Italy appears to run counter to my argument here. Despite a strongly

21

clientelist and centralized system, data on Italy from the 1970s indicates a very high correlation between

party strength—for all parties—at both the national and local levels.22 However, there are two features

of Italian politics that explain this. First, the impact of constant multi-party coalition governments at the

national level has no doubt had a sizeable impact. With a coalition always needed to form a government,

even smaller parties enter governments and gain a portion of the spoils, which they share with their local

party mates. Second, the Communist Party (PCI)—the permanent party of the opposition—actively

lobbied the central government for pork (Tarrow 1977) and appears to have done especially well in

many local elections, in large measure because of the ruling Christian Democrats’ great willingness

during that period to share patronage with the PCI in an effort to defuse greater polarization (Pempel

1998; Tarrow 1990).

Even short of patronage-oriented compromises of the sort seen in Italy, there appears to be one

clear way in which opposition parties can develop at the local level in Clientelist/Financially-Centralized

systems, but it is largely out of the control of the parties themselves. To gain more certain access to

goods from the center, local politicians in Clientelist/Financially-Centralized systems usually develop

very close ties to national level politicians. On certain occasions national level politicians defect from the

ruling party and join alternatives, creating a dilemma for local politicians. In such cases, fear of losing

their national level patron may induce local level politicians linked to such politicians to defect as well. In

this way, opposition parties may develop fairly quickly in specific regions, but such development is very

dependent upon the defection first of national level leaders hailing from those localities. For example, as

Bruhn (1997) explains, in the 1988 Mexican elections, much of the support for opposition presidential

candidate Cardenas came from defectors from the ruling PRI. Where the ex-PRI members of

Cardenas's movements were strong, they were able to take over the local party headquarters and run as

the local candidates. However, where there was no such group of elites, local party development was

more difficult (123).

Evidence from Japan 22 Data on Italian local elections are very hard to come by. Fortunately, Zariski (1984) makes available seat totals for the 1970, 1975, and 1980 regional elections.

22

To what extent does this framework accurately describe Japan? In this section, I demonstrate

the link between clientelism/centralization and opposition failure at the local level in Japan. If my

argument is correct, two patterns should be apparent. First, opposition party strength at the local level

should vary along with localities’ level of dependence on the central government. Second, party

development should tend to be top-down and moves by sitting local politicians to new parties should

occur only in conjunction with changes at the national level.

Each of these patterns holds in Japan. Using a series of direct and indirect measures of local

dependence on the central government, I find that shifts in Japanese localities’ level of autonomy have

also led to shifts in opposition party strength at the local level. I also find that, while party defection from

the LDP in 1993-94 was fairly widespread at the national level, local politicians simply would not leave

the party unless pulled by their national Diet member patron, their link to central governmental largesse.

Background

The Japanese party system is founded on personalistic and clientelistic competition. (For just

two of numerous examples, on Japanese personalism, see Fukui and Fukai (1996) and Reed (1994).)

This is due in part to various historical/path dependent and resource-based features of Japanese politics,

but also to the longstanding SNTV/MMD electoral system that, as Carey and Shugart (1995) argue,

encouraged personalistic political behavior to a far greater degree than any other electoral system

utilized in recent years. Despite the advent of a mixed system that includes a PR component, the new

Japanese electoral system also offers significant incentives for personalistic and clientelistic behavior

(McKean and Scheiner 2000).

In addition, Japanese localities are very dependent financially on the national government, and

local officials see central funding as absolutely critical to local viability. In a recent survey of local

officials in Japan, nearly every respondent cited insufficient governmental financing as, at a minimum,

“quite important” to their subnational governments’ financial considerations (Kobayashi 1999). The

heavy dependence of the localities on the center in Japan is captured by the phrase sanwari jichi, or 30

percent autonomy. While Japanese localities increased their independence somewhat over time, for

23

years local taxes covered only about 30 percent of local governmental revenues, while the remaining

roughly 70 percent comes from other sources, especially the central government. As the Japanese

economy grew in the 1980s, this figure reached as high as 44.3 percent, but with the collapse of the

Japanese economic bubble, it has dropped sharply back down to the low 30s (Ministry of Home

Affairs, various years). All other revenues, though, must come from the central government (Akizuki

1995).

From a comparative perspective, local taxes covering 30 to 40 percent is at or above the

median for major industrialized countries and particularly high for a unitary state (Reed 1986).23

However, what this figure does not capture is the flexibility of revenue sources. As Tarrow (1978)

notes, it is extremely difficult to use accounting of revenues and expenditures at the national and local

levels to demonstrate greater or less local autonomy. Unless we have independent measures of the

centralization of government decision making and the political influence of local elites and the mass

public, knowing the degree to which funding is based in the center or the periphery does not tell us who

is actually controlling how the money is spent (Tarrow 1978: 12).

In reality, Japanese local governments lack the long-term flexibility to be able to raise their own

sources over the long run to pay for large projects or for higher levels of services. In particular, the

regulation of local taxes by the central government is especially restrictive in Japan, even when

compared to other unitary systems—and grants and loans are typically given solely for purposes defined

by the central government (Reed 1986: 27-29). The central government determines how much money

each locality needs according to a complex, but fairly politically blind variety of formulae (Ishihara 1986:

139-141; Yonehara 1986: 161-163). According to an index of fiscal capacity, localities are expected

to make up a certain proportion of their vital needs through local taxes and the central government

makes up the bulk of the remainder of the localities’ primary “needs” through allocation tax transfers: the

central government collects such a tax from all regions, but then redistributes that revenue

disproportionately to the poorer ones. Additional funding in the form of central government

disbursements (subsidies) is typically determined far more according to central governmental discretion 23 For a more detailed sense of how the 60 to 70 percent of revenues covered by the central government compares to other countries, see Scheiner (2001).

24

to cover projects that go beyond those that are so “need-based” (Ishihara 1986; Yonehara 1986). In

this way, the central government is able to push its own priorities at the local level through its power to

provide and deny subsidies.

Subsidies are more likely to be given to poorer localities, but the central government also

appears to use subsidies politically. Igarashi and Ogawa note cases in which the central government

halted the disbursement of subsidies when faced with citizen opposition to planned public works

projects in regions heavily dependent on central funding (1997: 92-3). For this reason, large numbers of

local officials willingly give up much of their local discretion for the sake of gaining greater funding from

the center (Kobayashi 1999). Frequently, the only other revenue-raising option for localities is

borrowing, something that the central government has been known to restrain at times. In short, local

governments are extremely dependent on the central government for funding and, as a result, in the

heavily clientelistic Japanese party system, local legislators are typically evaluated on their ability to

develop a relationship with national governmental leaders and bring home funding for local projects (see

Reed 1986: 29). For this reason, ambitious local politicians have very strong incentives to affiliate with

the LDP if they wish both to perform their function as local representatives and do well electorally.

Akizuki (1995) writes, “In order to obtain…subsidies, local governments have to lobby either directly

or through politicians (usually the ruling Liberal Democratic Party’s Diet members from the local

district)” (354).

A description of politics in Toyama Prefecture offers a nice example of how this process works.

In Toyama, the LDP maintains a branch in every municipality in the prefecture. These party branches

collect citizens’ issues and demands and present them to local governments and legislatures (which of

course tend to be dominated by LDP members). If the most local levels of the party approves, then the

prefectural LDP will consolidate the proposals and present them to prefectural governments and

legislatures. Proposals that receive approval at the party's prefectural level are, in turn, taken to the

appropriate central ministries and agencies, either by the LDP national headquarters or individual LDP

Diet members (Fukui and Fukai 1996: 275). My own discussions with politicians in numerous other

prefectures suggest that the process in Toyama—while a bastion of LDP power—is the norm: the

25

LDP’s role in the process of allocating subsidies is similar in most prefectures. I should also add that

local politicians’ lobbying does not stop at the prefecture’s edge. Delegations of local politicians (from

all prefectures) constantly visit (especially ruling party) representatives from their prefecture to arrange

introductions to and appointments with the bureaucrats who oversee the process of subsidizing the

proposal for which they seek funding (Fukui and Fukai: 278).

Ultimately, the effect of all of this is substantial. The LDP controls the national government and

therefore has discretion in deciding to which localities it will allocate subsidies beyond the most basic

needs. As noted earlier, the government tends to disburse subsidies to those most in need, but, because

the LDP has been known to halt such disbursements for political reasons, the LDP is typically deemed

more likely to give subsidies over to areas that support the party (Igarashi and Ogawa 1997). For

example, according to a local observer, in Takaoka City—a city in Toyama, a very poor prefecture that

is heavily dependent on the central government—the weakened state of the infrastructure was due to the

fact that the city was led by non-LDP parties, and, as a result, was punished by the LDP-led central

government (Fukui and Fukai 1996: 283).

On a similar note, LDP politicians have explained that one of the main things their time out of

power taught them was that governmental ministries tend to give far greater attention to requests made

by politicians in the government than those who are not. Especially striking as evidence that local

politicians certainly have this impression, when the Socialist Party joined the LDP in a coalition

government in 1994, national-level Socialist members suddenly received a marked increase in the

intensity and number of requests for help funding local projects (Ibid.). As Fukui and Fukai write,

“Distribution of resources needed for the development of local and regional economies by pork barrel

politics gets inevitably skewed in favor of localities that are politically better connected” (285). Indeed,

writing about the 1990s, Okuda (2001) offers preliminary statistical evidence that as the number of HR

seats held by ruling parties in a given prefecture increases, so too do the subsidies awarded to that

prefecture.

Moreover, when facing electoral threats, leaders of the LDP are quick to suggest that victories

by non-LDP candidates will be likely to lead to the central government cutting off funds to the locality.

26

For example, two cases of such threats were clearly expressed by LDP leaders in the period just before

the 1999 HC by-election (Asahi Shinbun, October 19, 1999) and the 2000 gubernatorial election

(Yomiuri Shinbun, September 19, 2000) in Nagano prefecture. In the latter case, Yoshida Hiromi,

speaker of the Nagano prefectural assembly, warned the public that the election of an inexperienced,

non-LDP-associated candidate to office could have “severe consequences,” as “the governor is

required to have…connections with the central government” (Ibid.). Similarly, before the 2000 HR

election in the third district in Miyazaki Prefecture, associates close to LDP incumbent Mochinaga

Kazumi made it clear that if construction and related businesses in the district did not make a clear show

of support for Mochinaga, they could expect construction contracts to dry up and bidding opportunities

to disappear in the future (New York Times, June 24, 2000).

Such threats and punishments do not appear to be merely a recent phenomenon. Based on

research from the early 1960s, Kuroda (1974) finds that the central government regularly used pork

barrel to keep local voters in line and was regularly used to ensure that local political leaders would

remain with the LDP (133). Indeed, according to Kuroda, although only 43 percent of the local voting

public in the localities he studied had a genuine preference for the LDP and its principles, 95 percent of

local governments’ “top leadership” chose the LDP. Kuroda explains this finding in a way quite similar

to my argument: by holding local office in a fiscally very centralized system, local officials do have not

enough power to separate themselves from the national party leaders (Ibid.).

Even if governmental threats to cut benefits are simply bluffs, as long as the LDP is

dominant at the center and has discretion over subsidies, it is easier for areas with a large number of

LDP local politicians (compared to areas with fewer of them) to lobby the LDP central government for

subsidies. As a result, even voters who normally support the opposition parties—because of the issues

or principles they represent—have an incentive to vote for LDP local politicians if they, as voters, also

care about procuring club goods for their localities or even private goods for themselves, and ambitious

potential local politicians have an incentive to join the LDP.24

24 Readers may see substantial similarities between Japan’s one-party dominant clientelistic system and that existing under the solidly Democratic South in the U.S. for years. (See, for example, Key 1977.)

27

While I do not have public opinion data on voting in local elections, analysis of vote choice in

the 1996 HR election gives greater insight into the important role played by subsidies in the Japanese

system and the central part the LDP plays in their distribution.25 Respondents were asked to name all

the candidates they could remember from their district. Respondents were then asked to provide

characteristics for each candidate. Among respondents who mentioned an LDP candidate first, one out

of every six described the candidate as able to deliver subsidies to the district. Moreover, out of 309

respondents who mentioned at least one candidate as being capable of delivering such subsidies, 211

cited the LDP candidate from their district. The second most cited party was the NFP, which was

mentioned only fifty-eight times. Finally, multivariate analysis of vote choice in the election indicates that

voters citing an LDP candidate’s ability to deliver subsidies to the district were markedly more likely to

support the LDP than other parties (Scheiner 2001). This result also indicates the obverse: those not

citing subsidy delivery ability were more likely than those citing such delivery capacity to support

Japan’s new opposition parties.

Naturally, not all voters and subnational regions supported the LDP. In many cases, ideological

principles or ties to particular unions (which typically maintained specific ideological principles) drove

the vote for non-LDP parties. Moreover, the anti-LDP vote was particularly prominent in cities. In

addition to ideological or union-based reasons, four explanations in particular help to explain the lower

vote support for the LDP in urban areas. First, the closer community ties in rural areas facilitate the

creation of the personal support organizations that history and large sums of money make particularly

available to the LDP. Such ties are much less well developed in urban areas. Second, and related, the

very well organized farmers groups are among the leading support groups for the LDP, and naturally are

far more powerful in rural areas. Indeed, the LDP used Japan’s leading agricultural organization,

Nôkyô, as a major vote-gathering machine to help maintain its foundation in rural areas throughout much

of the postwar period (Mulgan 2000: Ch. 6).26

25 The data analyzed here are from the 1996 Japan Election and Democracy Study (JEDS96), conducted by Bradley M. Richardson, Mitsuru Uchida, and associates. 26 Naturally, in exchange Nôkyô was able to command substantial influence in shaping LDP spending on its farmers.

28

Third, it is more difficult to use patronage such as public works projects to attract voters in

urban areas. In urban areas, such projects typically influence the lives of countless people—many of

whom are not even voters in the area under construction—and therefore make it more difficult for

parties to target specific groups and claim credit for the benefits of the projects. In contrast, it is far

easier to both target specific, geographically concentrated groups and claim credit for spending in rural

areas, where a larger percentage of beneficiaries of such spending live and vote in the area. Fourth,

most urban areas are usually better able than rural ones to cover a larger proportion of their spending on

their own. As a result, urban areas simply depend less on central governmental support and voters,

therefore, have less incentive to cast ballots for the LDP. Despite these above exceptions, by and large

a majority of voters and localities supported the LDP in subnational elections.27

Given that local politicians are expected to focus on the delivery of pork, it should not be at all

surprising, then, that the LDP is almost wholly dominant in subnational elections, with its most potent

challenge coming from (typically conservative) independent candidates. At the same time, I do not mean

to give the impression that other factors besides pork are not important in generating support for

politicians at the local level. In particular, the personal vote is particularly important in Japan and most

local politicians have well developed bases of support founded on

their personal relationships with voters (see, for example, Reed 1994).

Nevertheless, while opposition politicians at the local level are perfectly capable of developing

such personal relationships, this ought to do little to undercut the advantage the national ruling party

ought to have at the local level. Presumably, politicians have two main reasons for running for local

office. First, many pursue subnational office out of personal ambition and seek to gain sufficient

experience, connections, and a base of support in order to run for higher office in the future. Second,

many have policy interests they wish to pursue. In either case, the national ruling party continues to

maintain an advantage in recruiting candidates. For local candidates who are running for reasons relating

to personal electoral ambition, the national ruling party is clearly the best option. The national ruling

party has a larger number of high positions to which the candidate can aspire. The party itself maintains 27 These explanations also all fit in very well with my larger argument about the impact of clientelism and centralized funding.

29

a larger store of resources to share with candidates. And of course access to the party also means

greater access to national governmental patronage that greatly helps candidates’ electoral chances. At

the same time, candidates pursuing local office for policy reasons also have strong incentives to ally with

the national ruling party. In financially centralized systems such as Japan’s, local governments are

extremely limited in their ability to engage in serious policy making.28 Politicians’ best chance at

policymaking is in the ruling party-controlled central government. For this reason, all else equal,

candidates with policymaking aspirations will seek to work their way up within the ruling party, a path

that typically begins with election to local political office under the party’s banner.29

As is clear from Figures 1 and 2, even when the opposition has done moderately well at the

national level, it is unable to translate this success into seats at the subnational level. Indeed, even while

non-LDP success in national level elections has fluctuated over time, the number of seats held by non-

LDP parties at the local level has stayed very low—at a nearly constant level—over the past 30

years.30 In contrast, the LDP always maintained a level of dominance at the prefectural level, which, in

many prefectures and many years, exceeded that which it held at the national level.

Indeed interviews bear my argument here out. When asked about the reasons for LDP

hegemony at the local level, local opposition leaders from a variety of Japanese prefectures explain that

it was because LDP politicians typically are perceived as more capable of acting as a “pipe” between

the LDP-led central government and the localities.31 If this argument is correct, voters and candidates

will be loathe to affiliate with the opposition, unless they are not greatly dependent on the center. That is,

local opposition strength ought to occur primarily in areas with substantial financial autonomy from the

center.

28 Indeed, as I discuss in greater detail below, local policy innovation only occurred in Japan once local governments gained greater financial autonomy from the center. 29 Moreover, while aspiring politicians with views wildly divergent from the LDP will of course avoid running under its banner, on the whole the LDP has been fairly flexible in its policy positions, and politicians within the party hold a fairly wide array of issue positions. Interestingly, the party’s heavy use of patronage to prop up voter support may actually give the party greater latitude in considering more public goods types of issues. 30 The mean proportion of seats held over this time by the LDP at the national level has been .540 (with a standard deviation of .058); for the LDP at the prefectural level it has been .560 (standard deviation of .047); for the non-LDP parties at the national level it has been .445 (standard deviation of .052); but for the non-LDP at the prefectural level it has been .302 (with the tiny standard deviation of .022). 31 Interviews conducted by the author with local opposition leaders in Japan from October 1998 through August 1999.

30

Measures of Local Autonomy and Local Opposition Success

Although Japan is heavily centralized compared to other countries, not all subnational units in

Japan are equally dependent upon the center. The degree of local fiscal autonomy varies with both

region and time. Certain areas of the country may be more financially autonomous than others, and the

country as a whole might go through periods where localities hold greater independence. Moreover,

where localities are more independent, politics should be less like the Clientelist/Financially-Centralized

system I describe above and more like the Clientelist/Financially-Decentralized type instead. Under

such conditions, voters need worry less about supporting local politicians from the ruling party and can

more freely support politicians of their most preferred party.

In this section, using a number of measures of local autonomy, I demonstrate that in the highly

clientelistic Japanese system, local opposition strength tends only to arise when local autonomy is fairly

high, and such strength declines when autonomy is low.

Local Autonomy Index and Prefectural Assembly Success

While clearly a centralized, unitary system, Japan is split up into a variety of subdivisions. At the

highest of these subdivisions, Japan is divided into forty-seven prefectures, which are typically smaller

geographically than the average for administrative subunits in most countries, but about average in terms

of population. At its most local level, the country is divided into over 3,000 cities, towns, and villages

(each being a legally distinct category, with all of Japan contained in either a city or a town or a village).

The twelve largest cities are specially designated as metropolitan zones, and the largest of them—

Tokyo—is coterminous with a prefecture.

While candidates for office at the city, town and village level usually run without official party

affiliation, most candidates for prefectural-level assembly office carry the affiliation of one of the major

national parties.32 If local dependence on the center is truly important to local electoral politics,

opposition parties should do poorly in highly dependent prefectures and be most successful in

32 For a number of years, there has been a significant rise in the number of unaffiliated candidates at the prefectural level, but most prefectural assembly members continue to affiliate with a specific national level party.

31

autonomous ones. To investigate this, we need some way of measuring variety among Japanese local

governments in their degree of fiscal independence.

A commonly used index of local fiscal strength (zaiseiryokushisû) in Japan is computed by

dividing the localities’ own revenues (e.g., local taxes) by a measure (computed by the central

government according to a complex formula) of the localities’ expenditure needs. Using this index for

each prefecture in the fiscal year directly before the seven local elections held 1967 to 1991, there is a

very substantial .57 correlation between local fiscal strength and the proportion of the prefectural

assembly seats held by opposition parties. Figure 3 illustrates this relationship. Only in the most

autonomous prefectures does the opposition win a large proportion of assembly seats.

Through OLS regression analysis I conduct a more systematic analysis of the relationship

between prefectural autonomy and opposition success in the prefectural assembly. The dependent

variable in the model is PrefOppit, the proportion of the prefectural assembly seats won in prefecture i

in election t.33 The data come from seven local elections held 1967 to 1991. I use four independent

variables. (1) Autonomy is the local fiscal strength score for the prefecture in the year leading up to the

election. Greater autonomy should lead to opposition success, so the

Figure 3: Opposition Prefectural Assembly Success by Level of Autonomy (1967-1991)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

Local Autonomy Index

Pro

po

rtio

n o

f P

refe

ctu

ral

Ass

emb

ly S

eats

Hel

d b

y O

pp

osi

tio

n

Par

ties

Source: Tôkei de miru ken no sugata.

33 Before running the regression, I use a logit transformation in order to make this dependent variable unbounded.

32

coefficient for Autonomy ought to have a positive sign.34

(2) HROpp is the proportion of the prefecture's national HR seats that are held by opposition

parties. Even controlling for other factors, we have to expect that voters in certain prefectures will—

whether due to ideology or simply socialization—be more likely to support either the LDP or the

opposition in all elections. In short, this variable offers a way to control for partisan predispositions in

national-level elections. In addition, it also controls for non-LDP connections that opposition parties

might have to national politics. The coefficient on HROpp should be positive: we should expect to see a

higher proportion of prefectural assembly seats being won in prefectures where the opposition holds a

higher proportion of national seats as well.

(3) GDP is the percentage by which Japan's GDP grew in the year leading up to the election.

Because higher growth rates ought to lead to greater support for the ruling party, the coefficient on GDP

ought to be negative. (4) 1967-71 is a dichotomous dummy variable, coded 1 for the 1967 and 1971

elections and 0 for all the other years. Over the late 60s and early 70s, opposition parties’ popularity

increased dramatically because of their stances—and the LDP’s unresponsiveness—on, in particular, a

devastating pollution crisis. Scholars have frequently argued that much of the opposition’s success in

those years was due to the opposition’s willingness to take stronger stands on issues that the LDP

appeared incapable of responding to during those years (Steiner, Krauss, and Flanagan 1980). I expect

that, even while controlling for other factors, the opposition ought to have been more successful during

those two elections, as voters sought to reward the opposition for its policy stands during this time.

It is important to note that because I examine the same set of prefectures over a series of

election periods, this data set involves repeated observations on the same set of cross-section units.

Because the number of cross-section units (prefectures) is larger than the number of elections, I use an

34 In an earlier draft of this analysis, I argued that the causal arrow may have gone both ways between Autonomy and PrefOpp and therefore used a two-stage least squares (2SLS) model. (In the model, I used prefectural income as an instrumental variable and, in the results, the adjusted Autonomy variable’s coefficient was positive and statistically significant.) My reasoning was that prefectural legislatures with larger proportions of LDP legislators might have been more likely to receive more support from the central government. However, in reality, the Autonomy variable is a measure of localities’ ability to find their own sources of revenue and not the likelihood of receiving funding from the center. Since there is no reason to suspect that PrefOpp will influence Autonomy, I run an OLS model instead of 2SLS.

33

estimation method that correctly treats the data as panel data (Johnston and DiNardo 1997: Chapter

12).35

As Table 4 indicates, each variable is statistically significant and in the expected direction. Most

important, the positive coefficient for Autonomy tells us that as prefectures’ ability to fend for themselves

increases, opposition parties win a larger share of their assembly seats. Autonomy remains significant

and positive, irrespective of the inclusion of other independent variables.36Comparing public opinion

data from urban and rural areas offers support for this as well. As noted above, rural areas tend to be

less autonomous than urban ones. Indeed,

analysis of the 1996 JEDS public opinion survey indicates that while nearly 30 percent of rural voters

noted “ability to bring in subsidies” as a significant characteristic of at least one HR

Table 4: Factors Shaping Opposition Party Success in Prefectural Assembly Elections (1967-1991)? Panel Data Estimation37 Variables Coefficient (SE) Constant -1.916***

(0.116)

Autonomy 0.653***

(0.173)

HROpp 1.615***

(0.274)

GDP -0.022***

(0.008)

1967-71 0.231*** (0.076) N 326 F (4, 46) 23.36 Prob > F 0.0000 R-Sq .4466 Number of clusters (prefectures): 47

*Significant in a one-tail test at .05 level 35 I utilize Stata’s “cluster” function, which produces OLS coefficients with panel-adjusted standard errors. I also confirm these results using fixed and random effects models. 36 Moreover, I would like to emphasize the importance of local fiscal autonomy, as opposed to other, more time specific factors such as the policy appeals of the late 60s and early 70s: while eliminating Autonomy from the model leads to a dramatic drop in the R-squared, dropping the 1967-71 dummy variable had scarcely any effect whatsoever. 37 Using fixed and random effects models (and a Hausman specification test to distinguish between the fixed and random effects models’ results), I confirm the statistical significance of all the variables.

34

**Significant in a two-tail test at .05 level

**Significant in a two-tail test at .01 level Compiled from data available in Japan Statistical Yearbook (various years) and Tôkei de miru ken no sugata (various years).

candidate in their district, little more than 10 percent of urban voters did. In short, generating financial

support from the central government for localities is a more important task of politicians representing

rural areas than for those representing cities. Given this and the fact that opposition parties win roughly

half as large a proportion of assembly seats in rural as compared to urban areas (Japan Statistical

Yearbook, various years), it seems that much of opposition parties’ weakness in rural areas stems from

voters’ perception that they are less able to deliver subsidies.38

Slack Resources and the Opposition’s Ability to Gain Subnational Executive Office

Reed provides a very useful insight into the relationship between resources and policy freedom:

“Local governments with resources will have greater capacity to innovate and to avoid slavishly

following the fiscal incentives of central grant programs…Central policies that affect local budgeting at

the margin will be more likely to result in central influence and more likely to inhibit local innovation”

(1986: 159). What this suggests is that subnational governments that have larger amounts of slack

resources—that is, resources that the local governments can spend and use at their own discretion—will

hold greater policy freedom and will be less beholden to the central government. As they are less

dependent on the center, we ought to expect that localities that maintain large amounts of slack

resources will also be more likely to elect opposition party candidates to subnational office.

At the same time, it can be difficult to measure slack resources. In particular, it is difficult to

determine under what conditions particular resources are slack, involving discretion on the part of

subnational governments. However, we can determine proxy measures that can provide a signal to

when local governments have such resources. I expect to find that during times when and places where

38 For more on the urban-rural split and the greater importance of subsidies in rural areas, see Scheiner (2001).

35

such proxy measures indicate greater amounts of slack resources available to subnational governments,

opposition parties will be more likely to win larger numbers of subnational office seats.

One such proxy is economic growth. During periods of high economic growth, local

governments ought to gain greater slack resources, leading to less reliance on the center. After resting at

a fairly low level in the 1950s, over the course of the 1960s and first half of the 1970s local government

tax revenues picked up dramatically simply because Japan grew richer. Japanese local tax revenues

increased by 1,200 percent over the 1959–1974 fiscal years before dropping in 1975 (Ministry of

Home Affairs, various years).39 This growth in slack resources gave local governments the capacity to

pursue new programs, and, as might be expected, local programmatic innovation and the number of

opposition local governmental executives both grew dramatically during this time. Not surprising, during

this period the mean value of the Autonomy index reached its highest point in the entire postwar era.40

Perhaps the most important “innovation” by local governments was their response to the

pollution crisis of the time. Various forms of pollution in the 1950s and 60s in Japan were quite literally

disfiguring and killing Japanese citizens, and, while the central government was very slow to respond to

the crisis, local governments introduced a series of policies to combat the problem. (See, especially,

McKean 1977, 1981.) Successful innovation in response to pollution appeared to encourage local

governments to pursue other problems as well, in particular social welfare programs. Governors such as

Minobe in Tokyo had started much of the trend toward progressive local government by pushing

spending programs such as free health care for the elderly in late 1969 (Campbell 1979). Once such

programs began in one region, they developed a type of snowball effect, carrying over to other regions

as well. That is, as Reed (1986) explains, once Tokyo enacted policies of this kind, a very large number

of local governments also put into place similar programs. More specifically, Campbell writes,

“Minobe’s action received extensive national publicity, and the idea of free medical care for the elderly

39 After a one-year decline, local tax revenues continued to increase until 1991, but not at the same dramatic rate as over the 1960s and early 70s. 40 There are two contradictory forces, which were parsed out in the results in Table 4. Economic growth leads to LDP support. However, the increased autonomy that grew out of economic growth is expressed in the Autonomy variable and leads to opposition success.

36

diffused among other local governments with astonishing speed,” so that within two and a half years

forty-four of Japan’s forty-seven prefectures were offering elderly health care support (1979: 333).

My analysis of course suggests that much of this innovation was due to greater local fiscal

autonomy. It might be asked, though, if budget autonomy was the issue, why did the opposition

politicians win on public goods issues such as the environment during these years? Reed (1986) offers a

very simple answer, which I implied above: “Economic growth provided local governments with slack

resources that could be allocated to new programs…It makes eminent sense to argue that economic

growth and increased revenues were necessary conditions for the spurt of innovative local policymaking,

providing local governments with the wherewithal to respond creatively to new needs and demands”

(58-59).

As it was seen that local governments could engage in greater policy innovation, even

conservative local governments innovated as well.41 Moreover, as more local governments saw that

they could spend more on such programs and exert their own individual plans and as progressive parties

and candidates actively pushed such measures, without serious punishment by the central government,

progressive parties and candidates became more likely to run and voters became more likely to support

them. Opposition progressive parties42 made great strides in local executive elections, going from

eighty-four mayoral seats in 1963 to ninety-two in 1967 to 114 in 1971 and 122 in 1975 (Zenkoku

Kakushin Shichôkai 1990).43

During times of greater budgetary restrictiveness, deficit spending is another possible proxy

measure for slack resources and autonomy. For Japanese localities, deficit spending represents money

borrowed to pay for spending that goes beyond what they are able to raise on their own through various

local taxes and what they receive from the central government through transfers and subsidies. If deficit

spending were a perfect proxy, localities with higher deficit spending would somehow be more

41 Reed notes the example of the conservative Kumamoto Prefecture’s ordinances relating to medium-sized retail stores in the mid-70s (1986: 53). 42 I define a progressive executive as any mayor holding the endorsement of the Socialist or Communist parties, but not that of the LDP. This leaves out purely centrist candidates, but there were so few of these during this period that it scarcely affects the statistics on the total number non-LDP/party endorsed candidates during these years. I explain my exclusion of centrist candidates in greater detail below. 43 The significant and positive sign on 1967-71 in Table 4 further documents local opposition success in late 60s and early 70s.

37

autonomous than those with less. This is of course silly for a number of reasons. In particular, the most

autonomous governments are those with sufficient funding of their own so that they do not need to

operate at a deficit. Nevertheless, there is reason to think that deficit spending might work as a proxy. In

particular, insofar as deficit spending might work as a way around some of the constraints placed on

local governments by the central government, deficit spending might represent the assertion of discretion

over spending by local governments. That is, when local governments are denied subsidies by the

central government, often times their best bet for funding is to borrow and run a deficit.

During the 1960s, the high-growth era, there was no clear relationship between the number of

localities operating at a deficit and the number of progressive mayors in Japan (Ministry of Home

Affairs, various years). Early in the 60s, many localities were spending more than they were taking in,

but, as their tax revenue increased because of the country’s unprecedented economic growth, by the

late 60s and early 70s far fewer were falling into the red.44 As noted above, thanks to growing local

innovation, the number of progressive mayors elected in cities during this time increased monotonically.

However, with the further increase in local programs in the early 70s and the first oil shock, the number

of localities operating at a deficit increased over 1972 to 1975.

Beginning in the latter half of the 1970s, anxious because of the era of lower growth rates that

clearly lay ahead, and consciously seeking to roll back the local programmatic “excesses” of the early

70s, the Ministry of Home Affairs and Ministry of Finance initiated a campaign to curb local social

welfare programs and to push many localities away from their previous programmatic promises. In

particular, these efforts rested on controlling local finances (Reed 1986: 55-56; Yomiuri Shinbun, May

25 and August 6, 1982). The debates surrounding such moves were partly ideological left–right battles

over the substance of policy, but in reality they were also very much focused on the issue of whether the

central government had the right, even the obligation, to maintain control over spending by local

governments (Reed 1986).

The outcome of the battle is quite clear: the central government won and far fewer localities

were permitted to operate at a deficit. Indeed, by 1979, the number of municipalities operating at a 44 In the first half of the decade, roughly 20-25 percent of all cities were operating at a deficit, while from the late 60s to early 70s, only about 10-15 percent were (Jichishô, various years).

38

deficit was at its lowest in over two decades, down to just five percent of Japan's municipalities. As

Figure 4 indicates, this decline in slack resources available to localities was also accompanied by a

decline in the number of progressive party government executives.45 As further evidence that the decline

in the number of municipalities operating at a deficit was due to central governmental constraints (and

not local governmental choice), in the post-1960 period,

the mean local autonomy index for all localities was at its lowest in the latter half of the 1970s.

Figure 4: Proportion of Progessive Mayors and Proportion of Cities Operating at a Deficit

0

0.05

0.1

0.15

0.2

0.25

1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987

Year

Pro

po

rtio

n

Progressive Mayors

Deficit

Compiled from information in Japan Statistical Yearbook (various years), Ministry of Home Affairs (various years), Richardson (1997: Table 4.2, p. 87), Zenkoku Kakushin Shichôkai (1990: Appendix, pp. 549-558).

It might be argued that I have reversed the causal arrow and that the decline in the number of

progressive mayors actually led to the decline in the number of cities in the red. However, there are two

reasons to dismiss this argument. First, despite only a moderately small drop in the number of

progressive mayors between 1975 and 1979, there was a very substantial decline in the number of cities

in the red. This decline is particularly striking given the additional spending needs the localities must have

had in the late 70s due to the second oil crisis.

45 As noted in a footnote above, I do not include purely centrist mayors in the lists of progressive mayors. There were very few purely centrist mayors in the late 60s and early 70s, therefore, scarcely affecting the total number of non-LDP mayors during those years. However, purely centrist mayors increased substantially in the 80s (Richardson 1997: Table 4.2, p. 87), as they developed closer ties to the LDP. I therefore exclude them from the proportion of progressive mayors in Figure 4 because they would provide an inaccurate picture of anti-LDP strength at the local level in the 1980s.

39

Second, the most serious drop in the number of progressive mayors did not occur until after

there was a bottoming out of the number of localities working at a deficit. Most likely, voters and

candidates were used to the fluctuations in their local governments’ capacity to run deficits and were

unprepared for the long-term nature of the constraints placed on their budgetary autonomy. That is,

voters may only have begun to appreciate that budgetary austerity was going to last quite a while after

the second oil crisis arrived in 1979, and only then did they begin to seek alternatives with tighter links

to the center. Ultimately, having been elected partly on the basis of their promises of welfare spending,

but having less money to use because both local sources and central spending dropped in size, many

progressive local officials were seen as going back on their word (Curtis 1988: 72-74).

Using the Autonomy measure utilized above in the analysis of prefectural assembly seat holders

reinforces the discussion from above. Figure 5 plots the mean Autonomy measure by year and the

proportion of local executive office holders who were progressives. I standardize the measures in order

to make their patterns more visually recognizable.46 Even before standardizing the variables, the

correlation between them is an extremely high .8. That is, as Figure 5 indicates, as Autonomy increases,

so do the proportion of local executive seats won by progressive candidates. And as Autonomy

decreases, so does the proportion of local, progressive-held executive seats.

46 I standardize the variables by dividing the value of each variable by the largest value obtained in that variable for any year. However, again, this is merely for the sake of visual representation. The correlation between the two variables is very high even before standardization.

40

Figure 5: Proportion of Local Executives Who Are Progressive and Mean Levels of Autonomy

0.8

0.85

0.9

0.95

1

1.05

1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979

Year

Sta

ndar

dize

d V

alue

s

Autonomy

Progressive Executives

Sources: The Local Fiscal Autonomy data were compiled from Tôkei de miru ken no sugata. The progressive executive data were compiled from information in Japan Statistical Yearbook (various years), Ministry of Home Affairs (various years), Richardson (1997: Table 4.2, p. 87), Zenkoku Kakushin Shichôkai (1990: Appendix, p. 549-558).

It is important to point out that the decline in progressive/non-LDP mayors was not entirely due

to such mayors actually losing office. In many cases, progressive and centrist mayors actually linked up

with the LDP, so that they were supported by not just the opposition camp, but by the LDP as well. As

Figure 6 shows, the number of joint LDP-opposition party mayors increased dramatically from 1976,

while the number of opposition-only mayors plummeted to around 10 percent of all mayors. 47

47 Note that this table begins with 1976, the first year listed in Richardson’s (1997) table. Also note that in some ways, Figure 5 in fact over-estimates the strength of the opposition, by including purely centrist mayors—who, as I noted above, were seen to have moderately close ties to the LDP during the 1980s—in the Opposition-only group.

41

Figure 6: Proportion of LDP-Affiliated and Opposition-only Mayorships

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

0.50

0.55

1976 1979 1983 1987

Year

Pro

po

rtio

n

LDP-Opposition Mayors

Opposition-only Mayors

Data from Japan Statistical Yearbook (various years) and Richardson (1997: Table 4.2, p. 87).

The rise in joint LDP-opposition local executives also provides evidence against two potential

counter-arguments to my own here. First, some might claim that I am overstating the importance of ties

to parties in the national government in Japan. Rather, some might argue that, for the sake of

constituency service for one’s supporters, it is just as important, if not more so, to be affiliated with the

ruling coalition at the prefectural level, rather than the national level. There is no doubt that any

politician would, all else equal, prefer to be a member of the ruling party at both the national and

prefectural level. However, the fact that the leaders of prefectural ruling coalitions, the governors

themselves, have tended to prefer to affiliate with the national ruling party indicates how important it is to

hold close ties to central governmental funding. Indeed, based on an analysis of prefectural governors

prior to 1993, Weiner (2001) finds that roughly half of all governors elected as progressive, non-LDP

members in their first election linked up with the LDP for their second election. These progressive

executives were delighted to defeat the LDP in their first election, but, having done so, did not want to

be detached from the national government as their major source of funding (Ibid.).48

48 Although, to be fair, Weiner (personal communication) also notes that part of the reason for the pan-party support for governors relates to the great power of governors at the prefectural level, as LDP prefectural assembly members

42

In addition, in the years of confusion following the LDP’s split in 1993, many governors were

particularly eager to hedge their bets and engage in pan-party coalitions because of their inability to

predict which party would control the government at the national level.49 Moreover, unlike countries

such as Brazil, which are noted for their “reverse coattails effects” (Ames 1994), national legislators in

Japan do not demonstrate a concern for linking up with whatever party the prefectural governor belongs

to (Weiner 2001). In sum, the importance of ties to the central government powerfully shaped the

decision of subnational office holders to create links to specific political parties, those governing at the

center.

As a second criticism of the model I have presented here, some might argue that in fact being a

member of the ruling coalition is not necessary for a politician to provide constituent service. For

example, the Japan Communist Party (JCP) has found success in urban areas by acting as “gadflies,”

who lobby local bureaucrats for local services until they get want they want. However, it is noteworthy

that the JCP typically finds such success in urban areas, which typically receive relatively little money

from the central government and tend to be more autonomous. They, therefore, are more like the

Clientelist/Financially-Decentralized type I discussed above, where partisanship tends to be less

important.50 Also, the JCP’s ability to act as service providers is clearly an exception in Japan. Indeed,

Weiner notes that, throughout the bulk of the pre-1993 period, in order to gain greater access to pork,

all of Japan’s non-Communist parties have sought to link up with the LDP, behind a single candidate in

prefectural governors’ races throughout much of the country (Ibid.).

Finally, it is very noteworthy that the shift away from opposition-only to LDP-linked mayorships

occurred in the late 1970s when the central government tightened the screws on local governments’

ability to operate at a deficit and when local governments’ ability to generate their own revenues had

declined. That is, we should not be surprised that opposition party local executives developed stronger

ties to the LDP at exactly the time when links to the central government appeared more critical to

funding local programs. It is likely that as localities received fewer slack resources to pursue their own were loathe to be detached from the governor and therefore pushed hard to get the governor to accept their endorsement. 49 Personal communication with Robert Weiner. 50 Some urban areas may even be more like the Programmatic/Financially-Decentralized type I discussed.

43

funding, they linked themselves up with the LDP in order to better extract resources from the central

government.

Top-Down Party Development51

In the previous section, I indicated how the Clientelist/Financially-Centralized system shaped the

opposition’s ability to win office at the subnational level. In particular, the opposition tended to win only

local office in regions that were less dependent on the central government. In this section, I consider the

role of Japan’s Clientelist/Financially-Centralized system in shaping the likelihood of sitting subnational

LDP politicians defecting from the ruling party to join alternatives. In particular, I find that Japan’s

Clientelist/Financially-Centralized system gave sitting local politicians little incentive to join new parties,

unless doing so was their principal means to maintaining their primary connection to the central

government.

That is, if my argument is correct, the importance of the “pipe” function—which connects

localities to the central government through a series of LDP politicians—will push party formation

toward the top-down variety. In short, even if national political winds seem favorable to the formation

and development of new parties, local candidates—who rely so thoroughly on the assistance of

politicians at the national level—will be unlikely to join such parties unless doing so improves their ability

to maintain a relationship with those at the center. If so, given that joining non-LDP parties usually

weakens a politician’s ability to bring in pork, this creates another obstacle to opposition parties

developing a base of local politicians.

New party development in postwar Japan has been almost wholly from the top-down. With the

exception of Kômeitô—a party founded to represent an already established constituency, a Buddhist

lay organization—every major party born in the postwar period began at the national level. Most parties

have grown out of splits from existing parties, but even the Japan New Party (JNP), a new party made

51 I frequently refer in this section to interviews and correspondence with “X prefecture political reporter.” In a number of cases, in interviewing members of the media, I was asked not to cite their name or company, as their sources might not approve of their sharing inside information about the prefecture and its politics. So as to not jeopardize the position of any member of the media with whom I spoke, I keep all reporters’ names and affiliations confidential.

44

up of inexperienced, non-incumbent politicians in 1992-93, began at the national level. With a few minor

exceptions, the party had almost no local governmental presence, as discussions to form the party did

not begin in earnest until a number of months after the 1991 local elections, and the party largely

became subsumed by a new, larger party (the NFP) in 1994, months before the 1995 local elections

(Nagai 1997). Most likely, local politicians feel little incentive to affiliate with a party that has no ties to

central power, unless their locality has independent sources of revenue.

Similar to the Daley machine in Chicago, Japan has yûryokusha or political bosses running the

local political show (Curtis 1971). However, unlike the Chicago model where the mayor is the boss of

the local machine, in Japan the Diet member is usually the leader and local politicians follow (Reed

1986). Tsurutani (1980) and Reed (1984) offer evidence that demonstrates just how effective these

vertical linkages are. In the late 1970s, the LDP’s presidential selection process was shifted from the

backrooms to a primary election. In the party’s first presidential primary election in 1978, voting very

closely mirrored the factional alignment at the national level: “The preferences of Dietmen determined

the votes of rank-and-file voters through a vertical factional structure” (Reed 1986).

As noted above, because most localities are very dependent on the center for financing and

because local politicians are evaluated on their ability to bring in pork from the center, ambitious

politicians looking to run for local office have an incentive to run under the LDP’s banner. Within this

system, there is what is called keiretsu or patron-client relationships, linking national LDP politicians to

LDP politicians at the local level (Inoue 1992).52 Fukui and Fukai (1996) nicely spell out the nature of

these relationships:

A politician with a good reputation as an effective ‘pipeline’ into the national

government attracts local administrators and politicians and becomes leader of an

electoral keiretsu (line). This is an electoral coalition formed for, and on the basis of,

mutual benefits between a Diet member and local politicians. From the Diet member’s

52 “Keiretsu” here has a rather different meaning than that which is usually used in discussions of Japanese political economy. The term typically refers to a particular type of economic partnership between closely linked firms in Japan. In contrast, Inoue (1992), Fukui and Fukai (1996), and I are all referring to a political relationship between national and subnational level politicians.

45

point of view, it is primarily a vote-mobilization machine through which affiliated local

politicians get their constituents to vote for the keiretsu leader in a Diet election and, if

possible, get other voters to do the same…From the local politicians’ point of

view, the coalition connects them to a ‘pipeline’ to the national government,

helps them bring pork to the local constituency, and thus enhances their

reputation as somebody with clout and their chances of reelection…[Since] it is

formed primarily for electoral purposes, there is also an element of opportunism

in [keiretsu] members’ behavior (280-1, emphasis mine).

As Kataoka (1997) explains, “under LDP rule, its [prefectural and local] assembly members

could expect their demands upon the national government to be registered through a connection with

their patron LDP Diet member” (208). Therefore, in order to have access to national pork, many local

politicians become part of national members’ keiretsu. As evidence implicating Clientelist/Financially-

Centralized system in shaping keiretsu, it is important to note that, in general, members of traditional

opposition parties in Japan do not have keiretsu.53

Keiretsu tend to be tied only to the pork center of Japan, the LDP-controlled central government.

If this relationship is important to local politicians’ ability to get pork, then, if keiretsu leaders do

not move from one party to another at the national level, then we should see no established local

politicians making such a move to either established or new parties, as leaving the LDP would cut off

their primary pipeline to the center. However, in the 1990s, the defection of many members from the

LDP at the national Diet level left local politicians with a difficult choice to make. Particularly in 1993-94

when the new parties were part of the national government, local politicians who were in the keiretsu of

LDP defectors had an incentive to join the new party. Switching maintained their pipeline to the new

central government, and, even if local politicians were (correctly) skeptical of the non-LDP

53 This statement is based, in particular, on interviews with opposition party politicians and staff who usually explained that they—unlike the LDP—did not have tightly linked vertical linkages of this kind. In some cases, the word keiretsu even confused opposition party staff members. This is in sharp contrast to LDP politicians and staff who always understood the term and all its implications (although many found the term distasteful, as it implied something of a dirty arrangement).

46

government’s ability to stay in power, because their keiretsu leader provided their main tie to the central

government, it was in their interest at least to consider shifting with him.

These keiretsu relationships clearly played a huge role in defection at the local level. Over

1993-94, a substantial number of members of Japan’s national Diet split from the LDP to form new

parties. However, whereas more than 15 percent of LDP Diet members joined the new Shinsei

(Revival) party, only roughly 5 percent of LDP prefectural assembly members did. The primary

scholarly work on the 1993-94 Diet level defections from the LDP finds that (1) reformers, unhappy

with the LDP’s ineffectual efforts to pass reform, (2) more electorally insecure members, seeking a new

electoral edge to gain support with a reform-minded public, and (3) junior members, with less of an

attachment to the party and less well-developed bases of support, were the most likely LDP members

to defect from the party (Cox and Rosenbluth 1995; Kato 1998; Reed and Scheiner, forthcoming).

While data on support for reform among local level politicians are not available, I am able to test

the extent to which level of seniority and electoral strength correlate with defection at the local level.

Utilizing data on all LDP members from the four prefectures with the highest rate of defection—Aomori,

Iwate, Nagano, and Shizuoka—I tested a probit regression model of defection from the LDP and entry

into Shinsei, the main destination for LDP defectors. In these prefectures, out of a sample of

approximately 150 LDP local legislators, just over fifty split from the LDP and joined Shinsei.54

Nevertheless, in a model containing merely the number of terms served and measures of electoral

strength (measured both as margin of electoral victory, and as margin of victory that also controls for

district magnitude and total number of voters in the district), as well as another model that includes an

interaction between them, none of these variables came up even close to significant. Neither junior

standing nor electoral marginality appeared to have led prefectural assembly members to defect and join

the new party.

54 While determining party membership at the prefectural assembly level can be difficult to do on a year-to-year basis, information on prefectural assembly kaiha, or parliamentary groupings, is readily available in the Yomiuri Shinbun’s yearly Bun’yabetsu Jinmeiroku, essentially a “Who’s Who” of Japanese public and private sector leaders. Great thanks to Kataoka Masaaki for introducing this source to me. While kaiha do not always match up perfectly along party lines, in most cases available here, it is quite clear with which party each assembly member is associated based on the kaiha information available. (For a more detailed explanation of kaiha at the Diet level, see Curtis 1999: 173-179.)

47

Yet, there is a feature common to all four of these prefectures, one that probably accounts for

the fact that defection rates among prefectural assembly members were the highest in these four

prefectures. In each of the four prefectures, the major political figure at the national level was also a

leading LDP Diet member who joined Shinsei, and was a leading member of the new party.55 In fact,

wherever this was true in Japan, we see similar trends in rates of defection. In interviews, four 1993-94

Diet level defectors from the LDP—Ishiba Shigeru (Tottori prefecture),56 Kano Michihiko

(Yamagata),57 Kumagai Hiroshi (Shizuoka),58 and Watanabe Kozo (Fukushima)59—each explained

that all the members of their keiretsu left the LDP with them and that in their respective prefectures the

only prefectural assembly members to defect from the LDP were those in the keiretsu of a Diet

member who defected from the LDP as well.

Similarly, Terayama Toshio, a DPJ member of the Tokyo assembly, noted that the defection from the

LDP in that assembly also tended to follow keiretsu lines.60 Detailed written mappings provided to me

by political reporters in three prefectures—Aomori, Shizuoka, and Mie—indicate clearly which

prefectural assembly members left the LDP and which joined Shinsei. In all three prefectures, the only

prefectural assembly members from the LDP to join Shinsei were in keiretsu of Diet-level LDP

defectors and the only assembly members of defectors’ keiretsu who did not join Shinsei were two in

Aomori, who, while nominally still part of defector Tanabu Masami’s keiretsu, had for years been

moving significantly closer to a different LDP Diet member, Ôshima Tadamori, who did not leave the

ruling party.61

More systematically, as Table 5 shows, there was a very high correlation between defection

from the LDP at the national level and defection at the local level. Out of forty-seven total prefectures in

Japan, by 1994, twenty-eight had national level Diet members who split from the LDP to join Shinsei,

55Three of the Diet members were in fact probably the three leading members of the new party: Ozawa Ichiro in Iwate, Hata Tsutomu in Nagano, and Kumagai Hiroshi in Shizuoka. 56Interview, July 23, 1999. 57Interview, May 21, 1999. 58Interview, May 27, 1999. 59Interview, May 21, 1999. 60Interview, May 27, 1999. 61 Written correspondence with prefectural political reporters in Aomori, (May 5 and 14, 1999), Shizuoka (May 12 and 19, 1999), and Mie (June 4, 1999).

48

and in twenty out of those twenty-eight prefectures, prefectural assembly members left the LDP to join

Shinsei as well (Yes-Diet-Defection/Yes-Prefectural Assembly Defection or YD/YP). In addition, in

only three of the nineteen prefectures where no national Diet members left the LDP to join Shinsei did

any assembly members defect (No-Diet-Defection/Yes-Prefectural Assembly Defection or ND/YP).62

Clearly, though, the correlation between joining Shinsei at the national and local levels is

not perfect. However, the exceptions are useful for what they can tell us about the reasons for defection.

As noted above, there were three prefectures in which no sitting LDP Diet incumbents left the party and

eventually joined Shinsei, but assembly members did (ND/YP): Aichi, Miyazaki, and Ehime. In Aichi,

longtime LDP Diet member Esaki Masumi retired from the Diet before the 1993 election and his son

Esaki Tetsuma ran in his stead. Thus, technically the sitting incumbent did not defect, though his anointed

replacement, the younger Esaki, did run under the Shinsei banner and, accordingly, the members of his

father's keiretsu did in fact join Shinsei with

Table 5 – Number of Prefectures with LDP Defection to Shinsei

Diet member defection?

Yes (YD)

No (ND)

Yes (YP) 20 3

Prefectural Assembly member

defection? No (NP) 8 16

Compiled from information in Yomiuri Shinbun’s Bun’yabetsu Jinmeiroku (various years) and interviews and correspondence with local political reporters and national level Diet members. him. In Miyazaki, two LDP prefectural assembly members had a dispute with the LDP leadership over

whom to support in the Miyazaki City mayoral election, and split from the party in anger, joining Shinsei

62Actually, thirty-eight prefectures had a Diet member in one of the LDP splinter parties. However, in five prefectures, the only Diet members who did defect from the LDP, either joined the small party Sakigake, or a different, even smaller new party. In addition, in two prefectures, there were LDP defectors who joined Shinsei, but these were candidates who had been out of office from 1990 to 1993 and had lost their assembly member base. Finally, in three prefectures, the only Shinsei Diet member was a candidate elected for the first time in 1993, therefore giving them insufficient time to build up a base of local assembly members. (Sources: Asahi Shinbun, December 11, 1994: 7; Kokkai binran 1998; Kokusei Jôhô Sentâ 1998; Miyagawa 1997.)

49

without following any national-level keiretsu leader.63 Finally, in Ehime, seventh-term prefectural

assembly member Doi Kazutoyo was on the verge of being kicked out of the party for violating party

rules when he switched to Shinsei.64 This was clearly a desperate tactic adopted to salvage his

individual political career, not requiring the stimulus of any senior keiretsu leader.

Recall that in eight of the twenty-eight prefectures that held LDP Diet members who joined

Shinsei, there were no LDP prefectural assembly members who followed this lead and also switched to

the new party (YD/NP). These exceptions suggest all the more powerfully the importance of the pipe

connection in shaping local defection. As might be expected from the discussion in the previous section,

because of cities’ greater wealth and, hence, fiscal autonomy—which in turn reduces the need for local

politicians to develop relationships with members of the central government—keiretsu ought to develop

more rarely and with far less strength in very urban areas. Indeed, LDP staff and political reporters

confirm this hypothesis.65 Not surprising, in three of the YD/NP prefectures—Osaka, Hyogo, and

Hiroshima—there were Diet defectors from very metropolitan districts who simply had no established

keiretsu who could have followed them. On the other hand, members of prefectural-level keiretsu

attached to national-level defectors from less urban districts in Hyogo, Hiroshima, Fukuoka, and Oita

decided not to join Shinsei out of concern over what such a move would do to their ability to get

funding for the prefecture from the national government.66

However, if in 1993-94 Shinsei was in the national government, why would prefectural

assembly members be wary of defecting to it? The main reason has to do with some politicians’ sense of

stability at the time regarding the staying power of the non-LDP government. While defection at the

national level did indeed put Shinsei in the anti-LDP government, the government’s tenure was quite

short. Indeed, from the time that the anti-LDP government was formed to the time that it lost its Diet

majority, only nine months elapsed. Given the brevity of Shinsei’s time in government, insecurity about

63 Written correspondence with Miyazaki prefectural political reporter, May 8, 1999. 64 Written correspondence with Ehime prefectural political reporter, May 15, 1999. 65 Written correspondence with secretary of an LDP Diet member from Osaka (May 27, 1999) and telephone interview with a Hiroshima prefectural political reporter (August 1999). In addition, in a survey I conducted of national LDP legislators, in contrast to rural politicians, Diet members from urban areas were much more likely to claim that they had no keiretsu . 66 This information is based on correspondence and interviews with local political leaders and media from these prefectures, May–August, 1999.

50

the stability of the new party system, and local skepticism about the new party’s staying power in

government,67 it is reasonable to think that many local politicians did not see Shinsei as a reliable long-

term pipe to central governmental resources.

In Tottori, a YD/NP prefecture, HR member Ishiba Shigeru split from the LDP and eventually

joined Shinsei. However, Ishiba’s underlying reason for leaving the LDP was his displeasure with the

LDP’s military policy. Upon leaving the LDP, he became an independent, was unconcerned with

creating a new, stronger party, and merely asked his supporters to leave the LDP, not join Shinsei.

Therefore, Ishiba’s keiretsu members became independents, rather than stay LDP or become Shinsei

members.68 In heavily conservative Okayama prefecture, LDP assembly members affiliated with LDP-

to-Shinsei defector Katô Mutsuki were very concerned about their electoral chances if they were not

members of the LDP and also expressed concern about their ability to be involved in the prefectural

legislative process if they were no longer members of the LDP.69 Only the politicians from one

prefecture, Saga, had substantive policy reasons for prefectural assembly members affiliated with the

LDP choosing not to switch to Shinsei. Saga assembly members were nervous about the potential

electoral costs of losing the LDP label, but, as members of an area that is highly dependent on its

agriculture industry, had possibly even greater misgivings given the agricultural liberalization being

pushed by the Diet coalition government of which Shinsei was a major part.70

To sum up, in most cases, local LDP politicians defected when their national level keiretsu

leader did, and local LDP politicians did not defect when their keiretsu leader stayed in the LDP. As

for the exceptions: out of the three cases of ND/YP, one really does turn out to have followed the rule

rather than the exception (because the younger Esaki and his father's keiretsu did all go to Shinsei in

Aichi), and the other two cases were clearly more about individual behind-the-scenes squabbling than

party realignment. In the cases of the local-level LDP politicians who did not follow their national-level

67 Interviews I conducted over May–August 1999 with Diet members from three different parties (the DPJ, Kômei, and the LDP), with the DPJ prefectural organization in Hyogo (June 24, 1999), and correspondence over the same period with prefectural reporters in Hiroshima and Saga confirmed the uneasiness held by a number of subnational politicians with regard to the stability of the party system and the anti-LDP government. 68 Interview with Ishiba, July 23, 1999. 69 Telephone interview with Okayama prefectural political reporter, August 10, 1999. Two of Katô’s twelve keiretsu members did leave the LDP (Fukui and Fukai 1996), but neither joined Shinsei. 70 Written correspondence with Saga prefecture political reporter, August 12, 1999.

51

politicians into the Shinsei (the YD/NP category), fear of not being able to get (or not being able to get

credit for) assistance from a future national level LDP government was clearly the most common and

influential force stopping local defection.

Had no local politicians defected, or had they defected without any regard to the moves made

by their keiretsu leaders, we would probably conclude that there simply were not very close links

between national and local politicians from the same region. Had all local politicians in the keiretsu of

LDP Diet defectors also defected, we would not have been able to tell if they were defecting out of

personal loyalty or as an attempt to maintain their primary pipe to the center. Instead, the following

occurred: (1) On the whole, local politicians did not defect unless their national keiretsu leader did. (2)

Most local politicians defected only when their keiretsu leader did, but some did not, out of fear of

losing their funding from the center. This suggests that there is a strong link between local and national

level politicians and that this link is founded to a large degree on the ability of national level politicians to

act as a pipe for the local level politicians’ distributive needs.

The case of Nagano Prefecture provides further evidence that defection at the prefectural

assembly level was not due solely to loyalty and personal ties (although they too are important). As

mentioned above, Hata Tsutomu was one of the leaders of the group that defected from the LDP in

1993 and the new party movement in Japan. Hata briefly served as Prime Minister in a minority, non-

LDP government in 1994 and he is the most powerful and well-known politician in Nagano Prefecture.

After Hata defected from the LDP, his keiretsu in Nagano defected as well. However, in 1995, the

trend in Nagano became somewhat different from other prefectures: the anti-LDP (Hata) and LDP

parliamentary groups within the Nagano assembly dissolved altogether and joined together to form a

new group called Shinkenseikai (The New Prefectural Political Group). With very few exceptions, as

of 2000, the LDP and DPJ simply did not exist at the prefectural assembly level in Nagano. On one

hand, loyalty and personal ties to Hata drew prefectural assembly members to keep their ties to the

LDP only at the very informal level. On the other hand, their need to maintain a pipe to the central

governmental funding helped keep prefectural assembly members from joining Hata’s party the DPJ.71

71 Phone interview with staff member of the DPJ’s Nagano local organization, August 1999.

52

Moreover, other evidence even suggests ways that the keiretsu relationships had more to do

with electoral considerations and desire for patronage than personal ties and loyalty. Remember that in

1994 Japan changed its Lower House electoral system. Among other things, the changes in district lines

were sure to complicate keiretsu relationships, as some local and national LDP keiretsu lines, with Diet

and subnational politicians who had each represented identical areas in the past, were now divided into

different areas. That is, in some cases, the Diet member and subnational office holder no longer shared

portions of a district. Kataoka (1997) indicates that, in a number of such cases, subnational politicians

were willing to look for new patrons, whose shared sense of region made them better suited for

electoral purposes. In addition, it should be highlighted that in 1994 the NFP had been gradually

establishing more and more of a presence in a number of prefectural assemblies by drawing additional

LDP defectors, until the LDP abruptly returned to power (Kataoka 1997: 210). In short, local

politicians were particularly willing to defect to the party that was in power at the national level, and

were largely unwilling to defect from the LDP once it was in power again.

The main point here is: if Japan’s clientelistic and fiscally centralized system prevented

defection for even those whose links to national politicians might have given them an incentive

to defect, it is no wonder that opposition parties have been able to attract so few potential local

politicians, even in years when such parties had fairly substantial popularity. In case the importance of

financially centralized, clientelistic politics in shaping opposition party strength at the local level needs

further hammering home, one ought to consider the following. In the post World War II Occupation

period, concerned about the role centralized power had played in Japan’s prewar imperialism, the

Occupation authorities took steps to increase local autonomy. However, once the Occupation period

ended, the central government gradually decreased the amount of autonomy it was willing to grant

subnational units. As the central government increasingly took away autonomy from these units during

the late 1950s and early 60s, several progressive party mayors switched to the LDP in order to improve

their chances of obtaining public works projects (Reed 1986: 45).

Alternative Explanations

53

Are there alternative models for the failure of Japan’s opposition to succeed at the local level?

Various explanations have been offered for one-party dominance in Japan, but none fully explains the

failure of opposition parties to mount a serious challenge. Moreover, they face even greater difficulty in

explaining why opposition parties have such difficulty gaining office at the local level, or found greater

success in the late 60s and early 70s, only to fizzle out afterwards.

Cultural arguments emphasize that the Japanese political culture has led to weak partisan

ties throughout much of the electorate. In large measure for this reason, voters have tended to base their

voting decisions on cues from local notables and appeals to traditional Japanese values. Some scholars

point to the essentially conservative, change-resistant culture of the Japanese electorate as a factor in

maintaining the LDP’s hegemony (Hrebenar 1986). Others argue that with its appeals to traditional

symbols and its promotion of clientelist policies, the LDP has managed to create in the voters a sense of

indebtedness and obligation (Richardson and Flanagan 1984). Another explanation for one-party

dominance and non-responsive opposition focuses on the Japanese HR’s now-defunct SNTV/MMD

electoral system, which helped fragment the opposition and protected the ruling party remarkably well.

This school of thought, led by “institutionalist” scholars such as Cox (e.g., 1997), emphasizes the

fragmentation effect of SNTV/MMD, which made it extremely difficult for the opposition, in particular,

to carry out efficient nomination strategies.

Ultimately, neither theories based on culture nor those based on electoral rules can explain the

failure of opposition parties in local elections. Given what is in fact very thin support for the LDP (and

considerable hostility), there does not seem to be a cultural predilection to serve the party. As for

electoral rules, one might posit a local-level version of this argument, saying that the failure of the

opposition at the local level is due to the fact that prefectural assembly elections in Japan continue to use

SNTV. However, even when SNTV was used at both the local and national level, the opposition did

significantly worse at the local level. This is particularly noteworthy because prefectural assembly races

also frequently utilize districts of significantly greater magnitude than those at the national level. Partly

because the largest of these prefectural-level districts are in relatively urban areas—where voters are

less tied into conservative politicians’ networks and which are less dependent on financial support from

54

the center— but also partly because they reduce many of the problems surrounding the coordination of

candidates, opposition parties are typically markedly more successful in such large districts than they

are in smaller national-level SNTV/MMD districts or the municipal-level SNTV/MMD assembly

districts. That is, because the large magnitude districts (by definition) maintain so many seats, even the

smallest groups can win seats, so there is very little need to coordinate on jointly endorsed candidates.

As a result, in such districts, opposition parties ought to do better than they do in the smaller district

magnitudes that exist at the national level.

Perhaps a more subtle electoral system effect might be at work here. That is, more autonomous

prefectures tend to be particularly urban and urban prefectures have more large magnitude districts. In

this way, perhaps the correlation between autonomy and opposition success is a spurious one and

opposition parties are actually winning more seats in autonomous prefectures because such prefectures

have more districts with large magnitudes. However, there are two reasons for rejecting this argument.

First, in more financially autonomous prefectures, opposition parties win a substantial number of seats,

even in small magnitude districts. It is in the more financially dependent prefectures that the opposition

does particularly poorly in the districts with low magnitudes. Second, much of the opposition’s failure in

the smaller local assembly districts comes not from the coordination difficulties that are usually attributed

to SNTV, but rather because no opposition party even puts up a candidate in the first place. This point

is particularly consistent with my argument about the effects of clientelism combined with financial

dependence: not only are voters more leery about casting ballots in such prefectures for opposition

parties, but ambitious candidates are also less comfortable running under their banners.

My argument in the previous paragraph might imply that my finding is perhaps based on a

different spurious correlation: autonomy and urbanization are highly correlated and voters in urban areas

might have reasons beyond independence from the center for supporting opposition parties. For

example, because LDP politicians rely so strongly on well-organized networks of voters, such politicians

are less able to find success in urban areas, where voters are less tightly knit. In short, opposition party

success at the prefectural level is dependent on level of urbanization and not level of autonomy.

Arguments of this kind certainly have some merit and I would never deny the greater impact of

55

networks in rural areas. However, there are two very good reasons to believe that the combination of

clientelism and fiscal centralization also plays a

critical role in shaping opposition failure at the subnational level.

First, without including the impact of clientelism and dependence on the center, focusing on

other urban-rural differences cannot offer a sufficient explanation for why the opposition does so much

worse in subnational level elections. In contrast, my argument about the impact of the combination of

clientelism and financial centralization posits that voters and candidates will be especially likely to affiliate

with the national ruling party in contests for subnational level office.72 Second, as I discussed in the

subsection above on slack resources and the opposition’s success in subnational executive elections,

subnational opposition success was not simply a factor of urbanization, but also increased and

decreased over time. As I argued above, changing levels of opposition success were very much a result

of, first, greater overall local governmental financial independence, and then a return to greater

dependence on the central government for funding.

Conclusion

It appears that opposition parties face tremendous obstacles to gaining power at the local level

in Clientelist/Financially-Centralized systems. In cases such as Japan’s, where strong candidacies are

needed to do well in national-level races, these obstacles prevent opposition parties from developing a

pool of local level officeholders, and thereby harm opposition parties substantially not only at the local

level, but at the national level as well. Independent of moves made by politicians at the national level,

opposition parties face long odds of developing in such systems.

This creates two additional problems for opposition parties. First, Park (1998) writes about the

advantages local politicians bestow upon a party:

72 Indeed, if the argument counter to mine were correct, opposition parties would do roughly as poorly at the national and local levels. This suggests that HROpp—the proportion of HR seats held by opposition parties in a given prefecture—would correlate very highly with the proportion of seats held by opposition parties at the prefectural level, to the extent that it might wash out any effect of Autonomy. However, in the regression results listed in Table 4, even while controlling for HROpp, Autonomy remains significant.

56

Securing the cooperation of local politicians is critical to the success of the [national

level] electoral campaign. Local politicians are specialists in campaigning; they are not

only knowledgeable but also willing to engage in campaigning. These politicians can

identify the political loyalties of nearly every household in their home territories. And

as they are engaged in politics as a career, they can volunteer their time for the

campaign…In addition, as local politicians have their own personal supporters, Diet

candidates can use their channels to secure easy access to ordinary voters (76).

Without such infantrymen, opposition parties in Japan are severely disadvantaged.

Second, Ames (1995) demonstrates that in Brazil, former city mayors running for national office

tend to have geographically concentrated bases of support. Preliminary evidence suggests that the same

is true in Japan. Japanese campaign laws, such as those restricting television and radio time, make it

extremely difficult for candidates and parties to campaign on broad-based, mass appeals. Therefore,

Japanese opposition parties are doubly disadvantaged. They can neither make broad appeals nor,

because of their lack of local assembly members, take advantage of local, targetable, and concentrated

bases of support.

Japan’s opposition faces a Catch-22. It needs strong local candidates to generate quality

challengers for national elections. But, without strength at the national level, it has great difficulty finding

such candidates. To develop local level strength, and break this vicious circle, it must rely upon the

outcome of events at the national level. That is, the opposition must hope for further LDP defections at

the national level in order to develop at the local level. This is highlighted by the words of a staff member

in the DPJ’s local party organization bureau, who explained that for the DPJ to succeed, “The LDP

would have to split again.”73 That is, a large group of LDP politicians, compatible with those in the DPJ,

would need to split off from the ruling party and link up with the Democrats.

73 Interview, February 1, 1999.

57

References

Akizuki Kengo. 1995. Institutionalizing the local system: Ministry of home affairs and intergovernmental relations in Japan. In The Japanese civil service and economic development: Catalysts of change, eds. Hyung-Ki Kim, Michio Muramatsu, T.J. Pempel, and Kozo Yamamura, 337-66. Oxford: Clarendon Press.

Ames, Barry. 1995. Electoral rules, constituency pressures, and pork barrel: Bases of voting in the Brazilian Congress. The Journal of Politics 57:324-43.

Asahi Shinbun. Various Issues. Bruhn, Kathleen. 1997. Taking on Goliath: The emergence of a new left party and the struggle for

democracy in Mexico. University Park, PA: Pennsylvania State University Press. Carey, John and Matthew S. Shugart. 1995. Incentives to cultivate a personal vote: A rank ordering of

electoral formulas. Electoral Studies 14:417-39. Cox, Gary W. 1997. Making votes count: Strategic coordination in the world’s electoral systems.

New York: Cambridge University Press. Cox, Gary W., and Michael F. Thies. 2000. How much does money matter? ‘Buying’ votes in Japan,

1967-1990. Comparative Political Studies 33:37-57. Curtis, Gerald. 1971. Election campaigning Japanese style. New York and London: Columbia

University Press. Curtis, Gerald. 1988. The Japanese way of politics. New York: Columbia University Press. Dachs, Herbert. 1996. The politics of regional subdivision. In Contemporary Austrian politics, ed.

Volkmar Lauber. Boulder and Oxford: Westview Press. Diaz-Cayeros, Alberto, Beatriz Magaloni, and Barry Weingast. 2000. Federalism and democratization

in Mexico. Paper prepared for delivery at Annual Meeting of the American Political Science Association, August 31-September 3, Washington, DC.

Fukui, Haruhiro and Shigeko N. Fukai. 1996. Pork barrel politics, networks, and local economic development in contemporary Japan. Asian Survey 36:268-286.

Hrebenar, Ronald. 1986. The changing postwar party system. In The Japanese party system: From one-party rule to coalition government, ed. Ronald J. Hrebenar. Boulder: Westview.

Igarashi, Takayoshi and Ogawa, Akio. 1997. Kôkyô jigyô wo dô suru ka [What to do about public works]. Tokyo: Iwanami Shinsho.

Inoue, Yoshihiko. 1992. Kokkai giin to chihô giin no sogô izonryokugaku: Daigishi keiretsu no jisshô kenkyû [Interdependence of the Diet members and local politicians: An empirical study of the “Keiretsu”]. Leviathan 10:133-55.

Ishihara, Nobuo. 1986. The local public finance system” In Public finance in Japan, ed. Tokue Shibata. Tokyo: University of Tokyo Press.

Jacobson, Gary C. 1990. The origins of divided government: Competition in the U.S. House elections, 1946-88. Boulder, CO: Westview.

Japan statistical yearbook. Various Years. Kabashima Ikuo. 1994. Shintô no tôjô to Jimintô ittô yûi taisei no hôkai [New parties merger and LDP

one-party rule collapses]. Leviathan 15:7-31. Kataoka, Masaaki. 1997. Changing local politics: party realignment and growing competition. In

Japanese politics today: Beyond Karaoke democracy, eds. Purnendra Jain and Takashi Inoguchi. New York: St. Martin’s Press.

Key, V.O. 1977. Southern politics in state and nation. Knoxville: University of Tennessee Press. Klesner, Joseph L. 1997. "Democratic Transition? The 1997 Mexican Elections." PS: Political

Science and Politics 30:703-711. Kobayashi, Yoshiaki. 1999. Enhancing local fiscal autonomy: A study of the Japanese case with

comparative reference to South Korea and the United States. In The challenge to new governance in the twenty-first century, ed. National Institute for Research Advancement (NIRA). Tokyo: NIRA.

Kokkai binran. 1998. August edition. Tokyo: Nihon Seikei Shinbunsha Hakkô. Kokusei Jôhô Sentâ. 1998. Shitteokitai kokkai giin no ugoki [The party movements of members of

the Diet]. Tokyo: Kokusei Jôhô Sentâ.

58

Lauber, Volkmar. 1996. Conclusion and outlook. In Contemporary Austrian politics, ed. Volkmar Lauber. Boulder and Oxford: Westview Press.

McKean, Margaret A., and Ethan Scheiner. 2000. Japan’s new electoral system: La plus ça change…. Electoral Studies 19:447-477.

Miller, William L. 1988. Irrelevant election? The quality of local democracy in Britain. Oxford: Clarendon Press.

Ministry of Home Affairs. Various Years. Chihô zaisei hakusho [Local fiscal white paper]. Tokyo: Okurashô Insatsukyoku.

Ministry of Home Affairs. Various Years. Chihô zaisei tôkei nenkan [Local fiscal statistical yearbook]. Tokyo: Okurashô Insatsukyoku.

Miyagawa, Takayoshi. Various years. Seiji handobukku (Political handbook). Tokyo: Seiji Kôhô Sentâ.

Mulgan, Aurelia George. 2000. The politics of agriculture in Japan. New York: Routledge. Park, Cheol Hee. 1998. The enduring campaign networks of Tokyo’s Shitamachi district.” In How

electoral reform boomeranged: Continuity in Japanese campaigning style, ed. Otake Hideo. Tokyo and New York: Japan Center for International Exchange.

Pempel, T.J. 1998. Regime shift: Comparative dynamics of the Japanese political economy. Ithaca: Cornell University Press.

Reed, Steven R. 1986. Japanese prefectures and policymaking. Pittsburgh: University of Pittsburgh Press.

Reed, Steven R. 1994. Democracy and the personal vote: A cautionary tale from Japan. Electoral Studies 13:17-28.

Reed, Steven R. 1997. Providing clear cues: Voter response to the reform issue in the 1993 Japanese general election. Party Politics 3(2):265-77.

Reed, Steven R., and Ethan Scheiner. Forthcoming. Electoral incentives and policy preferences: Mixed motives behind party defections in Japan. British Journal of Political Science.

Richardson, Bradley. 1997. Japanese democracy: Power, coordination, and performance. New Haven: Yale University Press.

Richardson, Bradley, and Scott Flanagan. 1984. Politics in Japan. Boston: Little Brown. Rodriguez, Victoria E. 1997. Decentralization in Mexico: From reforma municipal to solidaridad

to nuevo federalismo. Boulder: Westview Press. Scheiner, Ethan. 1999. Urban outfitters: City-based strategies and success in post-war Japanese

politics. Electoral Studies 18:179-98. Scheiner, Ethan. 2001. Democracy without competition: Opposition failure in one-party dominant

Japan. Ph.D. diss., Duke University. Sellers, Jefferey M. 1998. Place, post-industrial change and the new left. European Journal of

Political Research 33:187-217. Tarrow, Sidney. 1977. Between center and periphery: Grassroots politicians in Italy and France.

New Haven: Yale University Press. Tarrow, Sidney. 1978. Introduction. In Territorial politics in industrial nations, eds. Sidney Tarrow,

Peter J. Katzenstein, and Luigi Graziano. New York: Praeger Publishers. Tarrow, Sidney. 1990. Maintaining hegemony in Italy: “The softer they rise, the slower they

fall!” In Uncommon democracies: The one-party dominant regimes, ed. T.J. Pempel. Ithaca: Cornell University Press.

Tôkei de miru ken no sugata. Various Years. Tokyo: Nihon Tôkei Kyôkai. Weiner, Robert. 2001. Anti-competition in “competitive” party systems. Ph.D. diss., University of

California at Berkeley. Yomiuri Shinbun. Various Years. Yomiuri Shinbun. Various Years. Bun’yabetsu Jinmeiroku (Key names in various fields). Yonehara, Junshichirô. 1986. Financial relations between national and local governments. In Public

finance in Japan, ed. Tokue Shibata. Tokyo: University of Tokyo Press. Zenkoku Kakushin Shichôkai. 1990. Shiryo Kakushin Jichitai [Materials on progressives at the local

level]. Nihon Hyoronsha.

59

Zariski, Raphael. 1984. Coalition formation in the Italian regions: Some preliminary findings and their significance for coalition theory. Comparative Politics 16:403-420.