This server contains information on the British Columbia Political Studies Association (BCPSA) and its activities. The current BCPSA president is Alexander Netherton ( of the Department of Political Studies at Vancouver Island University in Nanaimo, BC.

Do Japanese Voters Vote Strategically

by Masaru Kohno (UBC)
for the British Columbia Political Studies Association Conference Simon Fraser University Burnaby, British Columbia May 5-6, 1995


Do Japanese voters vote strategically? Despite the massive volume of
studies accumulated over  the last four decades on Japanese elections
and political participation the treatment of this  question in the
existing literature is incomplete at best Researchers specializing in
Japanese  voting behavior are so heavily influenced by the
sociological/survey-analysis tradition of the  Michigan school that
they have ignored the issue of strategic voting entirely 1 

The two recent  exceptions to this oversight, Reed (1990) and Cox
(1994), address strategic voting in Japan, but  they do so in the
context of inquiring into the structural effects of the Japanese
electoral system  of multi-member districts and a single non-
transferable vote, which existed between 1947 and  1994 for the
House of Representatives In their analyses, the possibility of
strategic voting is  assumed, rather than treated as a matter of
empirical dispute Both Reed and Cox do, however,  provide some
evidence to demonstrate that strategic voting takes place in Japan By
extending  what is known as "Duverger's Law," they argue and show
that, under the above Japanese  electoral system, strategic voting See
for example Flanagan et al (1991), which claims to be the "major
comprehensive study of the  attitudes and voting choices of the
Japanese electorate" (p 45) with no reference to the concept  of
strategic voting See also a similar omission in Miyake (1989), a
textbook on voting behavior  written by a leading Japanese scholar in
the field   produced competition among M+1 "serious" candidates in
M-seat districts. 2 In this paper, I  take these analyses one step
further and explore the determinants of Japanese strategic voting,
other than the structure of the electoral system. In particular, I seek
to highlight the spatial  constraints on strategic voting and their
implications. While strategic voters will desert their  most preferred
candidate, if he/she has poor chance of winning a seat, they will not
necessarily  vote for any candidate who has a better electoral chance.
Their voting choice is constrained by  the policy/ideology positions of
available alternative candidates in that they are likely to vote  for a
candidate who is close to their own policy/ideology preference. While
Cox's model assumes  this spatial] effect explicitly (1994, p. 610), as
does Reed implicitly (1990, pp. 338-40), neither of  these earlier
studies directly reveals such an effect. 

My primary concern here is to examine the  pattern of strategic
voting by the supporters of the Japan Communist Party (JCP) who are
assumed to be located on the far left of the Japanese policy/ideology
space. 3 I focus on this  particular (and relatively small) subset of
Japanese voters for several reasons. 

First, according  to conventional wisdom, the JCP has a more solid
organization than its rival parties and its  supporters are believed to
be among the s strongest party loyalists. In this sense, the case of the
JCP supporters provides a "tough case," and a successful
demonstration of strategic voting by this  subset of voters would
constitute strong evidence that such voting takes place in Japan more

Second, although most of its candidates face poor electoral prospects,
the JCP usually  runs candidates in all electoral districts in general
elections, providing a member of observations  large enough to
impose various analytical controls. 2 Throughout the period from
1947 to 1993, a majority of Japanese electoral districts for the House
of Representatives, the more important lower house, consisted of 3-5
member districts, although there have been some exceptional single
member, 2-member and 6-member districts. 3 I thus assume that
Japanese party competition takes places in a unidimensional
policy/ideology space. For empirical support for this assumption, see
Kabashima (1986) .  

 Third, especially for the purpose of demonstrating the spatial effect
of strategic voting, a focus  on the JCP supporters makes the analysis
straightforward because, being located at the far end  of the
policy/ideology space, their choice of alternative candidates is
limited to one direction.  

Finally, a successful demonstration of strategic voting by the JCP
supporters would also have  broad implications, highlighting the
inter-party competition on the left, namely between the JCP and the
Japan Socialist Party (JSP), as an  important factor in understanding
the evolution of the Japanese party system. 

In the following section, I build on the insights from the previous
research and advance a set of  hypotheses regarding the
determinants of strategic voting in Japan. In the third section, I
present my model, analyze the data, and evaluate the results. The
final section discuss the broad  implication-s of this study and
concludes with suggestions for future research. 

Hypotheses In the literature on contemporary Japanese politics, Reed
(1990) and Cox ~1994) are the only  two studies that address the
issue of strategic voting in Japan. 4 These two studies, however, are
driven primarily by a concerns with the structural effect of the
peculiar Japanese electoral  system and, in their analyses, the
possibility of strategic voting is assumed, rather than treated  as a
matter of empirical dispute. An alternative, and no less important,
research design would  treat the pattern of Japanese strategic voting
as a behavioral outcome and explore its various  determinants. 

Under this approach, the Japanese electoral system may still be
found to be an  important factor, 4 Compared to Cox's formal model
of "strategic voting equilibria," Reed (1990) downplays the
importance of strategic voting, relative to elite-level "mechanical
effect," in producing the M+1 results.   but it may not be the only
source of Japanese voters' decisions to vote strategically. In fact, the
discussion and evidence presented by each Reed and Cox point to the
inappropriateness of making a categorical assumption about the
possibility of strategic voting in  Japan. While advocating the "M+ 1
Rule," these studies acknowledge that it took a long time for  the
number of competing candidates to reach equilibrium (Reed 1990)
and that the optimization  has been imperfect (Cox 1994). 

Faced with these observations, both authors develop informal  and
post hoc explanations for the apparent gap between their models'
predictions and the real  data. Reed claims that the time required for
optimization u as a product of "learning," as opposed  to instrumental
rationality, that guides the decision-making of strategic voters. Cox,
on the other  hand, relies on the existence of non-instrumental
voters, lack of information and other kinds of  "noise" prevalent in
the real world, to account for somewhat ambiguous empirical

The purpose of the following analysis is to explore more
systematically the set of conditions  under which Japanese voters
will vote strategically. Thus, instead of assuming a priori that every
Japanese voter is always strategic, I seek to identify those factors
which will motivate voters to  engage in strategic voting. I consider
three factors in particular, and hypothesize as to their  effects as

First, I retain the original insights of the two preceding studies and
examine the structural effect  of the Japanese electoral system on the
pattern of strategic voting. Generally, other things being  equal. the
more likely their votes will be wasted (by voting for the most
preferred candidate),  the more likely voters will vote strategically.
Meanwhile, the likelihood of their votes being  wasted varies
according to the different institutional context in which voters cast
their ballots.  As mentioned earlier, Japan for a long time (1947-93)
had an electoral system in which voters  cast a single non-
transferable vote to elect 2-6 representatives in each district. The
varying size  of electoral districts, i.e., the varying number of
assigned seats, would alter the proportion   of the vote required to
guarantee a seat, known as the Droop quota, thus affecting the
likelihood that voters would waste their votes. 

Hence, specifically: Hypothesis 1: 

Voters in larger-sized districts will be less likely to vote strategically
than those in  the smaller-sized districts. 

Second, aside from the size of electoral districts, voters' decisions to
vote strategically are  influenced more generally by their
expectations about the electoral prospects of the most  preferred
candidate. If voters are led to believe that he/she has a good
(although, of course, not  guaranteed) chance of getting elected.
voters have no reason to vote for an alternative candidate.  5 

Thus, as specifically discussed by Cox (1994), information about the
candidate's expected vote  share in the upcoming election is likely to
influence voters' decisions. What, then, constitutes  such information?
This is a difficult question to answer not on}y because some kinds of
information may be available (asymmetrically) only to some voters,
but also because each voter  may have a different way of assessing
the reliability of the given information. Since it would be  impossible
to create an exhaustive and balanced list of informational sources, I
focus here on one  kind of information about the candidate, which is
among the most widely (and thus  symmetrically) available to voters
in Japan: his/her incumbency vs. non-incumbency status. 

If  the most preferred candidate is an incumbent seeking re-election,
voters will have higher 5 The strategy of the supporters of small
parties, like the JCP, is straightforward, because these parties will not
endorse more than one candidate in an electoral district. However,
the strategy of the supporters of large parties, like the Liberal
Democratic Party (LDP), will be more complicated. In addition to
deserting candidates who are too weak, strategic voters may also
abandon candidates who are too strong and thus likely to obtain far
more votes than the Droop quota. Such excessive votes would also
constitute "wasted votes" in the sense that these votes, if distributed
correctly, might have otherwise contributed to electing one more
candidate from the same party. Because this second aspect of
strategic voting is irrelevant to the case of the JCP supporters, its
possibility and implications will be ignored.   6 expectations of
his/her electoral success. 

Thus: Hypothesis 2: Voters whose most preferred  candidate is an
incumbent will be less likely to vote strategically than those whose
most preferred candidate is not the likeliest. 

Finally, I consider the spatial constraints on strategic voting. While
strategic voters will desert  their most preferred candidate, if he/she
has poor chance of winning a seat, they will not  necessarily vote for
any candidate who has a better electoral chance. As assumed in any
formal  model of strategic voting, voters' choices are constrained by
the policy/ideology positions of  available alternative candidates in
that they are only likely to vote for a candidate who is  considered
close enough to their own policy/ideology preference. 

This, of course, begs the question: how close is close enough, or, how
far, in terms of  policy/ideology distance, are voters willing to
compromise in their search for an alternative  candidate? If voters
are liberal in trading off their policy/ideology preference, they may
continue their search along the policy/ideology space until they find
such a candidate. If voters,  on the other hand are reluctant to
compromise their preference, they may abstain, or may vote  for
their most prefered candidate even though they know that their
votes are likely to be  wasted. Thus, in the absence of an
immediately obvious second-best candidate, voters face  multiple
options. For he case of Japan, the lack of relevant survey data makes
it impossib to inquire  systematically into voters' choices among
these options. One can, however, still test the degree to  which voters
are liberal/reluctant about trading off their ideology/po}icy
preference. Practically.  voters evaluate each candidate's ideological
position based on his/her partisan affiliation. 

Thus,  to ask the question "how close is close enough?" is to ask
whether voters will be willing to vote  on}y for a candidate from the
second most prefered party, or will they also be willing to vote for
candidates from   third, fourth ... nth parties. Accordingly, I make
two competing hypotheses regarding the spatial  determinant of
strategic voting. 

The restricted version is: Hypothesis 3A: Voters' decisions to vote
strategically will be influenced by the availability of a  candidate
from their second most preferred party. 

The expanded version is: Hypothesis 3B: Voters' decisions to vote
strategically will be influenced by the availability of  candidates
from other parties. 

In sum, there are at least three important sources of voters' decisions
to vote strategically in  Japan. The varying size of electoral districts
may provide an institutionally-structured incentive for the
voters not to waste their votes. The most prefered  candidate's
incumbency vs. non-incumbency status may have an informational
effect on the  voters' electoral expectations. Voters may also be
concerned with the spatial implications of  strategic voting in
choosing an alternative candidate from a pool of available candidates.
Are all  of these determinants empirically significant? 

Data Analysis 

This section develops a set of regression models to test the
hypotheses outlined above. I analyze  primarily the voting pattern of
JCP supporters as an important subset of Japanese voters, using  the
data from the lower-house election in 1993. 

(1) Models and Results 

Table 1 summarizes the specifications of the two competing models
(one corresponding to above  Hypothesis 3A and the other to 3B) and
their estimation results.   Both models include the Communists'
relative vote share (of the total vote cast) in 129 districts  as the
dependent variable. 6 Model A, the restricted model. includes three
independent  variables: 1) the size of electoral district, which ranges
from 2 to 6; 2) a dummy variable, a value  of 1 assigned to the cases
where the JCP candidate was an incumbent candidate seeking for
reelection, and 0 to the other cases;7 and 3) a dummy variable, a
value of 1 assigned to the cases  where a candidate from the Japan
Socialist Party (JSP) was not compelling against the JCP  candidate,
and 0 for the other cases. Model B, the expanded model, includes four
additional  dummy variables to explore whether voters' decisions are
influenced by the availability of  candidates running from parties
other than the JSP. namely the Clean Government Party (CGP),  the
Democratic Socialist Party (DSP), the Japan New Party (JNP), and the
Social Democratic  Federation (SDF) .86 

This choice of the variable has a disadvantage of not taking into
account abstention. An alternative approach would be to use the
actual votes received by the JCP as the dependent variable but, even
with this, it would be impossible to identify "who abstain" without
appropriate survey data. The usage of the actual votes would also
cause some bias because of the relatively substantial
malapportionments of seats across electoral districts in Japan. 	7	

The Japanese mass media usually uses three categories in describing
candidates: 1) "Gen", or incumbents, 2) "Shin" or new faces, and 3)
"Moto," for former-incumbents. The last category refers to those who
had been elected before but lost their seats in the immediately
previous election. For the present analysis, I merged 2) and 3) to
form a dichotomous category, incumbents vs. non-incumbents. 8 The
CGP, which began their political activities in the mid 1950s as a
political arm of a Japanese Buddhism organization (Soka Gakkai),
entered the lower house electoral race in 1967. The DSP was
established by right-wing members of the JSP who had left the party
in 1960. The SDF was created in 1978 by another	group of former
JSP members. These three parties constituted the traditional centrist
camp in Japan, located ideologically between the JSP and the LDP.
The JNP, on the other hand, was formed as a new party (with no
incumbent politicians in the lower house) only a year before the
election. In addition to the JNP, two other new parties were
competing in the 1993 election, Shinsei-to and SakEgake. However,
because those two parties, unlike the JNP, were both established by
former members of the LDP, I have excluded them from the model.

TABLE 1 About Here 

AS is clear from the table, all the coefficient estimates for the three
independent variables for Model A are statistically  significant. The
signs of these coefficients are also each in the expected direction. On
the other hand, some of the coefficient  estimates in Model B are
insignificant (and some of the signs are not in the expected
direction). Especially noteworthy is that,  by including the additional
four dummy variables, the otherwise significant coefficient for the
district size variable becomes  insignificant (and its sign reversed).
While the inconsistency about this particular variable deserves
careful evaluations (see  discussion below), it seems appropriate to
conclude, at this point, that the three original variables in Model A
are all  important determinants for voters' decision to vote
strategically. (2) 


What do these results imply at a more concrete level? First, on the
one hand, the results confirm the theoretical intuition  underlying
Reed's and Cox's previous studies that the Japanese voters' decisions
to vote strategically are structured by the  existing electoral system.
While replications of similar analyses for the past elections are
warranted, the finding that the  voting decisions of even the JCP
supporters, presumably the strongest Party loyalists, are affected by
varying district size  illustrates the   In prominence of the
institutional effect on the pattern of strategic voting in Japan. 9 

One can  further argue, based on these results, that the 1994
electoral reform, especially the introduction  of a single-member
plurality system, will have a considerable impact on the JCP's vote
share in  future elections. 1O According to Model A, the JCP will lose
approximately .8 percent of its vote  share per each unit decrease in
the number of assigned seats; this should provide a rough  estimate
of the expected Communist vote in each electoral district under the
new electoral  system. l l 

Second, on the other hand, the above analysis also reveals the
limitations of the  previous studies, by indicating that strategic
voting is not simply a product of institutionally- structured
incentives of the voters not to waste their votes. As is evident from
Table 1, at least  two more factors, other than the structure of the
electoral system, are important determinants  of Japanese strategic
voting. In fact, according to the relative size of coefficient estimates,
the  effects of the JCP candidate being an 9	If the grouping of
different-sized districts was correlated with other social factors, such
as different levels of urbanization or varying occupational
distributions, the correlation found here would be spurious. For
example, if larger districts (5 or more member districts) were
concentrated in Tokyo, Osaka and other metropolitan prefectures
where, for such social reasons, the JCP might attract more votes than
in rural areas, then the reported results would be open to other
interpretations. It would be difficult, however, to claim that the
grouping of different sized districts coincided with the urban/rural
distinction or with industrial demographics. Densely populated
prefectures, such as Tokyo and Osaka, were subdivided into several
districts, some of which were three and four member districts,
whereas some rural prefectures, such as Shimane and Yamanashi,
constituted single five-member districts precisely because of their
small populations. In fact, the average district size for Tokyo (3.91)
and for Osake (4.00) is almost identical to the average district size for
the entire Japan (3.96).10

The new electoral law enacted in 1994 stipulates, in essence, a
hybrid of the Anglo-American plurality and European proportional
representation systems. Out of the total  500 lower house seats, 300
members are to be elected from single member districts and 200
from 11 PR regional districts.	11	A more accurate prediction,
of course, would also have to take into account the effect of re-
districting and various other changes brought about concurrently by
the electoral  reform, including the greater restriction on political
fundraisings. For the details of the reform,  see Christensen (1994).
ll incumbent and of an JSP candidate being absent as an alternative
choice are both more  substantial than the impact of district size. 

To elaborate further on the relative salience of the three
determinants, the finding that the  incumbency variable has the
largest impact is a particularly noteworthy result. In the
contemporary literature on the personal vote in the United States
and elsewhere, the  incumbency advantage is usually associated with
the distributional "pork-barrelling" aspect of  the candidate's
constituency services. which forms the basis for his/her name
recognition and  reputation (Cain, Ferejohn and Fiorina 1987).
Likewise, in the literature on Japanese politics,  observers
accustomed to the Liberal Democratic Party's (LDP's) long-lasting
regime have  emphasized the clientalistic and often corrupt
relationships between politicians and the  electorate with phrases
such as "machine politics" (Johnson 1986) and "interests-channelling
politics" (Inoguchi and Iwai 1987).12 

In the context of these studies, the formidable advantage  of being an
incumbent Communist candidate in Japan is surprising and
somewhat  counterintuitive. One would expect that the JCP
candidates' incumbency advantage would be  minimal because the
JCP has never participated in forming a government and, being the
party at  the extreme left, it has been excluded from the various
legislative bargaining mechanisms that  have been institutionalized
since the early 1970s (Krauss 1984, Ito 1990). The above result
nevertheless suggests that, even for a JCP candidate, being an
incumbent is a critical asset for  improving his/her electoral
prospects. One implication to draw from this may be that one should
not underestimate the purely informational aspects of the
incumbency advantage, not contingent  upon the material,
constituency-oriented services. Finally, perhaps the most important
is the finding about the significant impact of spatial  constraints on
strategic voting in Japan. 

The above analysis shows that at 	12	See also Calder (1988) and
Curtis and Ishikawa (1984).   least some Japanese voters, namely the
leftist voters analyzed here, are committed to certain
policy/ideology positions and their voting decisions are not
necessarily purely instrumental. As  mentioned earlier, both Reed
(1990) and Cox (1994) discuss the possibility of non-instrumental
voting, but they do so informally and in a post hoc manner only to
account for the shortcomings of  their models' predictions. The
empirical results reported above depict a more precise pattern  with
which such voting takes place, suggesting that two related aspects of
non-instrumentality  seem to be at work. First, voters are evidently
selective in choosing an alternative candidate for whom to vote
strategically. The finding that the absence of a JSP candidate matters
in predicting the vote share  of the JCP candidates indicates that
these voters are willing to cross the ideological boundary  between
the JSP and the JCP. They are not, however, willing to vote for
candidates located to the  right of the JSP. With regard to candidates
from the DSP, the JNP and the SDF, their  presence/absence have no
statistically significant effect on the vote share of the JCP candidates.
With regard to the CGP candidate variable, its negative (statistically
significant) coefficient  estimate can be interpreted as indicating, on
the surface, that the leftist voters are not merely  unwilling to vote
for, but rather hostile to, the CGP candidates. 

Second, other things being equal,  the absence of a JSP candidate
increases the vote share of the JCP candidates by approximately six
percent. This means that the leftist voters are not reluctant to waste
their votes when there is no  immediately obvious second best
candidate, i.e. a JSP candidate. C}early, then, one must conclude  that,
for ideological reasons, at least some Japanese voters choose not to
dessert their most  prefered candidate, even though he/she has poor
chance of winning a seat. (3) 


Some comment is needed on the apparent inconsistency between the
two models regarding the  estimates of the district size variable.
From a methodological standpoint,   1  the result that the statistical
significance of this variable disappears m Model B points to the
problem that it is partially correlated with the four dummy variables
included in the expanded  model. This leads one to explore further
the underlying causal relations behind the apparent  correlations.
Since those dummy variables measure the presence/absence of
candidates from the four centrist  parties, one can speculate that
they reflect a particular pattern of partisan strategies in  nominating
candidates for general elections. 

The next set of tables (Tables 2, 3, and 4) illustrates  more precisely
the nature of this pattern. Clearly, (and as expected from the
reported  correlations), the candidates of these centrist parties are
concentrated in relatively large-sized  districts.13 

TABLES 2, 3, and 4 About Here 

What do these findings imply? The partisan candidate nomination
pattern reflects an elite-level  "mechanical effect" of avoiding
electoral districts where the candidates would have poor electoral
chance. For small parties like the CGP, the DSP, the JNP, and the SDF,
the strategy to concentrate  their candidates in larger-sized districts
is reasonable, given the smaller Droop quota in these  districts. While
this line of reasoning clarifies the apparent inconsistency between
the two  models, it, in turn, raises the question: why does the JCP,
another small party, choose not to  follow a similar strategy and
usually nominate its candidates in all electoral districts? One source
of this assymetry between the JCP and other small parties may
simply be the  difference in their financial resources. Thus, the
resourceful JCP can afford to run many  candidates even while
knowing that many of them have poor electoral chances. Another
explanation may be that the JCP has a different agenda than 13 The
pattern of the SDF candidate distribution is not presented, since there
were only four candidates.   14 the other parties in nominating their
candidates. For example, it is possible that the JCP, whose party
financing relies largely  on the circulation of its oScal newspaper, has
distinctive incentives to maintain and cultivate memberships
through vigorous  campaigning in each electoral district. In light of
the results reported above, however, one can speculate as to one
other  possibility that takes into account the non-instrumentality of
the leftist voters in Japan. That is, the JCP's persistance in
nominating many candidates may reflect a partisan response to the
leftist voters' willingness to waste their votes for the sake  of
ideology, while the centrist parties' strategy may be indicative of a
lack of such willingness on the part of non-leftist  voters. To confirm
this, of course, would require a thorough comparative research on
the nature of strategic voting across  various subsets of Japanese
voters, which is beyond the scope of the present study. 


This study has presented empirical evidence which reveals the
pattern of strategic voting in Japan. Rather than assuming
categorically that every voter is always strategic, my purpose has
been to identify the set of factors that affect voters'  decisions about
whether or not to vote strategically. The data analysis conducted on
the leftist voters suggests that these  voters do vote strategically
because of the institutionally-structured incentive not to waste their
votes and that they do so  based on their expectation of the
candidates' electoral prospects. The analysis, however, also points to
the significance of  spatial constraints on Japanese strategic voting.
The voters are not always purely instrumental: they are selective in
choosing  an alternative candidate and they are willing to waste their
votes for ideological reasons. 

Because the analysis is based on the aggregate data of votes received
by the JCP   15 candidates, the generalizability of the above findings
remain somewhat ambiguous. On the one  hand, given that JCP
supporters are said to be the strongest party loyalists, the finding of
strategic voting by this subset of voters constitutes strong evidence
that such voting takes place  more generally in Japan. On the other
hand, it is possible that the apparent effect of spatial  constraints
may be typical only of the leftist voters, while those voters who
normally vote for  the centrist or conservative parties may not be as
concerned with the policy/ideology positions  of alternative
candidates in deciding to vote strategically. As suggested earlier,
further cross- sectional studies based on large-scale survey are
warranted to determine the varying degree to  which spatial
constraints are significant across different subsets of Japanese voters.
Although the generalizability may thus be limited, I emphasize the
significance of the finding  that the JCP supporters' decisions to vote
strategically are constrained by spatial considerations.  As indicated
earlier, such constraints have not been fully addressed in the
previous studies on  Japanese strategic voting. But, more importantly,
this finding has broader implications for  understanding the
evolution of the Japanese party system since the end of World War
II. The widespread consensus among the students of postwar
Japanese politics is that it was the  failure of the leading opposition
party, the JSP, to become a viable alternative to the LDP that
promoted the fragmentation of the opposition camp and thus
prolonged the Lap's single-party  rule. l4 Conventionally, the JSP's
failure has been explained either as a product of the JSP's  internal
problems, especially its organizational dependence on the
ideologically rigid labor union  Sohyo (see Flanagan 1984, p. 162;
Curtis 1988, pp. 117- 18), or in terms of its nervousness about  the
"reactionary" element in the LDP (Otake 1986, 1991). 

The above finding suggests an 14 The LDP failed to maintain its
legislative majority and retreated to the opposition camp after the
1993 general election, although, less than a year later, the LDP
managed to come back to power by forming a coalition with the JSP
and Sakigake.   16 alternative spatial explanation for the JSP's
failure: it was the JSP's competition vis-a-vis the JCP  over the pool of
leftist voters that prevented, for a long time, the JSP from adopting
more  moderate social-democratic policy positions. I realize that the
evidence presented in the  previous section does not directly point to
a precise ideological position at which the Socialists  had to be located
to remain sufficiently attractive to the JCP supporters. It is possible
that, as  long as the JSP was to the left of other compelling parties,
the Socialist candidates might still have  monopolized the privilege of
being the only alternative choice for leftist strategic voters, 

Two  pieces of further circumstantial evidence, however, indicate
that the range of policy/ideology  space in which Socialists had to be
located, was extremely narrow, adding support to the spatial
interpretation for the JSP's failure. First, the JSP's bent toward left
originally began after the  general election in 1949 when the
Socialists suffered a major defeat and Communists gained a
significant increase in both popular votes and parliamentary seats
from the previous election. 15  This election was held after the
collapse of two consecutive center-left coalition governments, in
which the so-called "Right Wing" Socialists took a major part. In this
sense, the 1949 election w as  critical in demonstrating the leftist
voters' intolerance for the Right-Wing's position and, indeed,  the JSP
subsequently came to be dominated by the Left-Wing.16 The left
dominance continued  until at least the mid 1980s when the JSP only
reluctantly abandoned its Marxist official  platform, but, even then,
with regard to defense and security policy, the JSP clang to its
idealistic position of unarmed 	15	In the previous election in
1947, the Socialists received 26 percent of the total vote cast and
obtained 143 seats in the lower house, whereas the Communists only
received 3.7  percent and obtained 4 seats. In 1949, the Socialists
received 13.5 percent of the popular vote  and obtained 48 seats.
whereas the Communists received 9.8 Percent of vote and 35 seats.
	16	The two Wings engaged in separate political activities
during the period from 1951 to 1955, as a result of their fierce
quarrel over the issues of postwar settlement and  Japan's security
arrangement with the United States, although both officially called
themselves  Nihon Shakai-to (Japan Socialist Party).   17 neutrality. 

The second piece of evidence in support of the above spatial
interpretation is the fact the JSP  finally abandoned the ultimate
symbol of its postwar progressiveness, the unarmed neutrality
policy, soon after the completion of electoral reform in 1994. As
mentioned earlier, this reform  introduced, albeit only partially, an
Anglo-American type of plurality electoral system with  single
member districts. 

The timing of the JSP's decision leads one to suspect that it was born
out  of a new expectation that, under the new electoral law, the
leftist voters would dessert most of  the JCP candidates in single-
member districts. However, if the leftist voters in Japan are not
entirely instrumental, as suggested in this paper, they may continue
to vote for the JCP under  the new electoral arrangement, even
knowing that they would waste their votes. What  determines the
fate of the competition between the JSP and the JCP, then, is the
balance of two  inherently-offsetting forces that influence the voters'
decisions: the institutionally-structured  incentive not to waste their
votes facilitating strategic voting, and the strength of their
policy/ideological preference standing it its way.   

TARI F I Effects of n Distriet Size Incumbency IndeDendent Variables
Crsnstsnt Absence of JSP Candidate Absence of CC,P Cnndidate
Absence of DSP Candidate Absence of JNP Candidate Absence of
Model A 2.52 (1.54) .84* [.384 8.85** [1.08] 6.34*# { 1.62} Model B
11.05** (2.89) -.47 (.40) 7.23** (.99) 5.79** (1.47) -4.16** (.81) -1.25
{.77) 99 (.70) R6 	SDF Candidate	f 1.761 Adjusted R2 F-rstin
Fsignificant at .05 level **significant at .01 level Source: Asahi
Shimbun, July 20, 1993 .44 ns .57 43.55 1R  

R	ELEt-,TION Candidate District Size 5 6 	_ 3	4 	Yes 0
(0'1/o) 6 (15.4(1/o)	13 (38.2%)	33 (71.7%)	2 (100%) 	No 8
(100%) 33 (84.6%)	21 (61.8%)	13 (28.3%)	0 (0%) 	Total 8 39
	34	46	2 x2 = 36.8 (significant at .01 level) Source:
AsahiShimbun,July5, 1993 lq   

ELErwTIoN Candidale District Size 5 6 	2 3	_ 	Yes 0 (O0/0) 4
(10.3%)	8 (23.5%)	16 (34.8%)	0 (0%) 	No 8 (100%) 35
(89.7%)	26 (76.5%)	30 (65.2%)	2 (100%) 	Tnl s1 8 39	34
	46	2 x2 = 10.5 (significant at .05 level) Source: Asahi
Shirnbun, July 5, 1993 2n   Candidate 

District Size 4 		~6 	2	3 	Yes	1 (12.50/,,)	10
(25.6%)	19 (55.9%)	23 (50%) 21 2 (100%) 	No 7 (87.5n-/o) 29
(74.6%) 15 (44.1%)	23 (50%)	0 (0%) 	T(ltsl 8 39 35	46
	2 x2 = 13.7 (significant at .01 level) Source: Asahi Shimbun,
July 5, 1993   


Cain, Bruce, John FereJohn, and Morris Florina.1987.
 The PersonaL Vote, Cambridge: Harvard University Press.

 Calder. Kent E.1988.
 Crtsis and Compensation.
 Public Policy and PoliticaÀ Stability in Japan, 1949-1986.
 Princeton:  Princeton University Press.

 Christensen, Rayrnond V.1994."Electoral Reform In Japan.
" Asian Surv~ey 34: 589-605.

 Cox, Gary W.1994."Strateglc Voting Equilibria under the Single
Nontransferable Vote" American Political Science Reuiew  88: 608-

 Curtis, Gerald L.1988.
 TheJapanese Way of Politics.
 New York: Columbia University Press.

Curtis, Gerald L.
 and Masumi Ishikawa. 1984.
 Doken Kokka Nippon lConstruction State - Jap‡nl Tokyo: Kobun-sha.

 Flanagan, Scott C.1984.
 "Electoral Change in Japan: A Study of Secular Realignment.
" In Electoral Change in Advarlced  IndustrLaÀ Democracies, ed.
 R J. Dalton, S. C.Flanagan and P. A. Beck.
 Princeton: Princeton University Press.

 Flanagan, Scott C.  Shinsau Kohei, Ichiro Mlyake, Bradley M.
 Richardson, and Joji Watanuki. 1991.
 TheJapanese Voter. New  Haven: Yale University Press.

 Inoguchi, Takashi, and Tomoaki Iwai. 1987.
 "Zoku Giin" no Kenkyu fA StucLy of "Policy TEes"1.
 Tnkvn- sshnn KPr7>s  

9himhlln ehn Ito, Tatsumi. 1990.
 "Kensho 'Kokutai-Seigi' no Kozai [Examining the Merits and Demerits
of 'Politics of the Diet  Counterrneasure Committee'l.
" Chuo Koron, December: 130-146.

 Johnson, Chalmers. 1986.
 'Tanaka Kakuei, Structural Corruption. and the Advent of Machine
Politics in Japan " Journal  oSJapanese Studies 12: 1-28.

 Kabashima, Ikuo. 1986.
 'Yukensha no Ideorogi [Ideologies of the Electorate].
" In Nihonyin no Senkyo Kodo IThe Japanese  ELectoral Behavior], ed.
 J. Watanuki, I. Mlyake, T. Inoguchi and I.
 Kabashima. Tokyo: Tokyo Daigaku Shuppan-kai.

 Krauss, Ellis S. 1984.
 "Conflict in the Diet: Toward Conflict Management m Parliarnentary
Politics " In Conf1fct in Japan, ed,  E.
 S. Krauss, T. P. Rohlen and P.G.Steinhoff.
 Honolulu: University of Hawaii Press.

Miyake, Ichiro.1989.
 Tohyo Kodo SVoting Behaviorl.Tokyo: Tokyo Daigaku Shuppan-kai.

 Otake, Hideo. 1986. Adenauar to YoshÀda ShÀgeru fAdenauer and
ShÀ~eru YoshÀda].Tokyo: Chuo Koron-sha.

   Otake, Hideo.1991."Defense Controversies and One-Party
Dominance: The Opposiuon in Japan  and West Germany." In
Uncommon DemocracLes: The One-Party Dominant Regimes, ed, T.J.
Pempel.Ithaca: Cornell University Press.

 Reed, Steven R 1990."Structure and Behavior: Extending Duverger's
Law to the Japanese Case."  BrLttshJournal of Political Sclence 20: