Democratic Representation in Japanese Defense Spending: Does Public Sentiment Really Matter?

Japan’s level of defense spending relative to the size of its economy has been anomalously low among the major industrialized countries. Much of the literature on Japan’s national security suggests that antimilitarist public sentiment constraints Japanese defense budgeting. A substantial amount of research is methodologically descriptive and the policy impact of Japanese public opinion is largely presumed but not closely tested. This study tested the relationship between Japanese public opinion and defense spending. In the process of testing, the existing measurement theory, which has accounted for cases in Western Europe and the Unites States, has been found inadequate to explain the Japanese causal link. With the introduction of alternative measurements, this study found that Japanese public opinion does uniquely constrain defense budget decisions.

For decades, scholars have been puzzled over Japan's reluctance to become a military power commensurate to its economic capacity. Constructivist scholars emphasize that Japan's post-World War II antimilitarist norms have constrained Japanese national security policy (Beger, 1993;Chai, 1997;Katzenstein & Okawara, 1993). Realist scholars argue that the structural and material forces will outweigh antimilitarist norms (Layne, 1993;Waltz 1993) or utility-driven Japan is riding on the back of the United States for military security (Lind, 2004). Is there any evidence that the structural and material forces have always underlain Japan's antimilitarist norms or have these norms trumped the immediate utility of material forces?
This article contributes to the debate by examining the budgetary effects of public opinion, which can be seen as an indicator of antimilitarist norms. It has found that both material calculations and normative considerations significantly affect defense budget decisions, implying that the two modes of influence are not mutually exclusive. It can be assumed that decision-makers routinely reason about both material forces and normative choices. A significant breakthrough for the debate would be to offer evidence as to the ways in which antimilitarist norms interact with material calculations.
The study in this article argues that the domestic variable of Gross Domestic Product (GDP)-tied spending (which reflects a widely shared expectation of antimilitarism) and its compliance mechanisms (which serve as the one percent of GDP ceiling) permit a reliable prediction of changes in representative defense budgeting. In its defense spending, the Japanese state is thus weakly autonomous from society. As constructivists suggest, domestic norms and other social structures may be considered as capable of constraining Japan's defense budgeting. This article claims that norms do not have the independent causal effect on defense spending but rather require institutional motives and material factors for policy makers to comply with norms. Greater attention to institutional and historical variables helped to disentangle the norm compliance mechanism of GDP-tied spending.
To explain the relationships between antimilitarist sentiments among the Japanese public and the comparatively low level of defense spending, this article has examined (1) public responsiveness -whether and how public preferences respond to government expenditures and (2) policy representation -whether and how budgetary decisions reflects public preferences for defense spending. The literature demonstrates that public opinion in both the United States and Western Europe responds to the external environment and foreign policy change in a continuously coherent, systematic way (Eichenberg, 1989;Holsti, 1996;Nincic, 1988;Peffley & Hurwitz 1992;Shapiro & Page 1988) (Note 1). With defense spending a more salient issue to the public than foreign policy in general, the public in these countries has adjusted its preference for more or less defense spending in response to defense than most of its allies, while Britain has shouldered its share of the defense burden larger than any other European NATO partner. Due to their larger share of defense spending in the government budget and GDP, public responsiveness has been remarkably robust in these two advanced democracies (Sokora & Wlezien, 2002;Wlezien, 1996). This strong relationship between greater defense spending and public responsiveness seems to be arrayed against Japan. Its minimal defense spending has a smaller share of the total government budget. As Figure 1 illustrates, it has been less variable with an average of less than one percent of GDP over the past forty years. Therefore, Japanese public opinion would appear to have a weak effect on defense spending.
Mass publics in general do not have preferences over details at a greater level of specificity to which they cannot realistically be expected to pay attention. As most survey organizations suggest, aggregate preferences over time are observed as changes in standing predispositions toward the role of government, which can be expressed simply as "do more," "do less," or "just about right." As the thermostat metaphor illustrates, public reaction reflects a preference for moderation in defense spending. In general terms, once policy moves outside the zone of public acceptance, public preferences will react by wishing for a return to a zone of acceptability. It is not surprising to find that this general notion of acceptability and reaction is also applicable to Japanese public reaction. But since Japan's spending "temperature" changes least among advanced industrial states, it would seem that the Japanese public would also be weakly responsive to defense spending. In the United States and Britain, where societal pressure is competitively divergent, the policy mood for defense spending strongly fluctuates as the spending "temperature" changes. It is in a plurality favoring either an increase or decrease in defense spending that the dynamics of public preferences resemble a thermostat. In contrast, Japan's social structure is usually described as less divided on government policies. In terms of local currency and GDP, Japanese defense spending varies least of all advanced industrial states (Bobrow, 1989;. Public opinion does not change because spending does not vary. In other words, Japan's median preference should behave least like a thermostat as the spending temperature does not significantly differ from that preferred. The measurement theory commonly used for proving the thermostat model led researchers to discover another pattern of public responsiveness to defense spending: the negative reaction to any trade-off in which defense spending grows faster than social spending (Note 2). In post-World War II Japan, the Japanese government displayed a pattern of the "developmental state" by holding back social spending as far as possible (Johnson, 1982, pp. 3-34;Eccleston, 1989, pp. 88-122). In the early 1970s, social spending began to expand dramatically with the introduction of upgraded social benefits (Campbell, 1979). Since then, the ratio of social spending to GDP continues to increase with no sign of a downturn, while that of defense spending is kept under the restraint of a spending ceiling. Over the past four decades , social spending grew much faster than defense spending, except for fiscal years 1971, 1983-85, and 1989-90 when the annual growth rate of defense spending marginally exceeded that of social spending (IPSS, 2008). In Japan, the negative reaction to changes in the relative growth of defense and social spending is thus less likely to take place. The Japanese public is predicted to be less attentive to changes in the relative growth of defense and social spending as the government has continually placed much higher priority on social spending.
The public is sensitive to shifts in external threats and accordingly adjusts its preferences for the level of defense spending, although the effects of external threat vary with each country (Bartels, 1994;. As in other nations, the Japanese public has equally been aware of shifts in its external security threats (Note 3), but its external threat-driven reaction to defense spending ought to be comparatively weak. First, there have been no direct Japanese involvements in inter-state wars since the end of WWII (Note 4), so the Japanese public must be less sensitive than those involved. Second, as a sense of deep dependence on the U.S. security guarantee remains strong (Note 5), the Japanese public is expected to fear less about a possible danger to national security. Third, any perception of external threats should not effectively influence public support for defense spending as the majority of Japanese people do not see the use of force by Japan as legitimate or useful in external relations (Note 6). Therefore, the Japanese public would appear to be less supportive of defense spending to meet Japan's external security threats.

Policy Representation
If the public is attentive to defense spending, then the important question is whether budgetary decisions reflect public preferences for that. In general terms, how does policy making respond to societal demand? If public preferences follow similar patterns, then where do differences in policy output between various countries come from? The literature suggests that we need to examine domestic structures which show the institutional arrangements linking state and society (Eichenberg, 1989, pp. 235-41;Gourevitch, 1986;Katzenstein, 1978;Risse-Kappen, 1991, pp. 484-86). In the area of defense spending, domestic structures provide a locus for coalition-building needed to build the support basis for budgeting. The benefits, which will be generated by defense spending, tend to motivate particular issue groups, such as defense industry and veterans' groups, to raise the budget issue. Coalition-building processes thus seek to link public and private sectors for making budget requests.
In postwar Japan, the peak organizations of the governing party, the national bureaucracy and big business sought to build consensus among the elite groups in various issue areas (Pempel & Tsunekawa, 1979). Japan also seems to be one of the most homogeneous societies in the world displaying relatively few ethnic, religious and regional cleavages. The unitary system of Japanese government should be able to easily control society and to prevent domestic resistance. Japan's state-led corporatist policy networks with big business seem to provide public opinion with less opportunity to affect policy outcomes (Note 7). In the area of defense spending, Japanese veterans' pensions are not included in defense-related expenditures, but handled as part of social spending (Otake, 1984, p. 24). The Nippon Izokukai (Association for Family Members of Japan's War Dead), which has over one million member households, thus has not been a lobbying group for greater budgets for defense expenditure. Only Japan's defense industry has consistently, yet unsuccessfully, lobbied for a larger defense budget commensurate to its economic status and technological capacity (Green, 1995;Samuels, 1994, pp. 154-269). In Japan, the coalition-building process of defense spending takes place in an underdeveloped interest-group environment; the level of spending should be easily coordinated at the top of monopolistic association between the elite groups.
Apart from the effects of public preferences, defense budgeting is seen as part of fiscal policy which seeks to control the level of the total expenditure for macro-economic purposes. Like other industrial nations, Japan's elite Ministry of Finance has a monopoly of information on the macro-budget and its expertise, and it is in a strong position to impose spending limits on the Ministry of Defense (formerly Defense Agency) (Campbell, 1977). This is doubly so, since Japan's defense administration lacks institutional autonomy (Katzenstein & Okawara 1993, p. 95). Macro-budgeting considerations must have a strong independent effect on Japanese defense spending.
The effect of the external environment on national defense expenditures seems to be obvious from the viewpoint of realists. National defense has been traditionally regarded as a policy area where decision-making takes place in isolation from the controversies of public reaction and is left to expert judgment. But how do policy makers decide on an appropriate level of defense spending by estimating the level of threat to a nation's security? Some researchers argue that policy makers are most likely to use changes in adversaries' defense spending for deciding their own appropriate level of threat (Hartley & Russett, 1992;Marra & Ostrom, 1992). Others assume that, as the conflict involvement of the nation rises, its policy makers are most likely to see it as a sign of increasing threats to the public and decide defense spending accordingly . Perhaps inter-state war involvements for the nation would have a clear effect on defense spending. However, since its defeat in World War II, Japan has neither gotten involved in interstate wars nor actively initiated MID actions. The literature found that there are two distinct factors necessary for considering the impact of external threat and alliance dynamics on Japanese defense spending. The first one is US security guarantees: the US Occupation of Japan (1945-52) initially demilitarized Japan, which in turn became dependent on militarized US for Japan's security. The second is US pressures: the United States eventually urged Tokyo to make greater security contributions. On the one hand, Japanese policy makers weakly respond to the external environment to the extent that they depend upon the United States for shifting the cost of fighting onto the ally (Dower, 1979;Otaka, 1988;Pyle, 1992). Some argue that this is the primary reason why Japan is reluctant to use its economic capacity for defense spending (Waltz, 2000, p. 34;Lind, 2004). On the other hand, due to the burden of US-Japan alliance borne by the United States in a disproportional way, Japanese policy makers are more prone to U.S. pressures for Japan to increase its defense spending (Keddell, 1993).

Public Responsiveness -Net Support
Public responsiveness presupposes that the public is attentive to what policy makers do. The focus of public responsiveness for defense spending is not so much on varied personal preferences, but rather on the estimated preferences of the median voter for whom policy makers have much incentive to be accountable (Sokora & Wlezien, 2004, p. 534;Wlezien, 1996, p. 84). Public opinion is the dependent variable to be explained for public responsiveness. In the literature, the dependent variable (public support for defense spending) is operationalized as a measure of "net support." In this process of operationalization, the response category of "the same as it is" or "about right" is ignored on the assumption that they do not have a clear stand on defense spending, thus leading to the measurement of nonattitudes. It is further presumed that, whether a respondent even actively supports the existing level of defense spending or not, the effect on defense budgeting has no differences with maintaining the status quo.
This research has shown that the measurement of net support alone does not account for all the information in the opinion surveys, to which policy makers are potentially responsive. A vast majority of Japanese prefers the status quo of defense spending. In the past decade, almost six in ten responded that they wished it to neither increase nor decrease. One suspects that the status quo responses predominantly reflect the active support of the existing level and contain only a small number of nonattitudes. Japan's Cabinet Office surveys asked the question: "Do you have interest in security issues and the SDF (Self-Defense Forces)?" On average between 1975 and 2006, 50-65 percent of the public indicated their interest ("very interested" plus "interested") in security issues and the SDF (JCO, 2004, p. 7;JCO, 2006). During the same period, although 35-50 percent showed no interest ("not really interested" plus "no interest at all"), only a third of those respondents answered that they had no interest "because I do not understand the SDF or security issues very well" (JCO, 2004, p. 7;JCO, 2006). The remaining two-thirds had clear opinions regarding their indifference: no pressing military threat, no necessity of the SDF, no relation to their personal life, and others. At least 80 % of the respondents actually had a stand on the issue and these respondents were not "nonattitudes" toward the issue.
By the late 1970s, specific forms of anti-militarism were accepted widely by the Japanese public. The Japanese people had no desire to return to the country's militarist past yet they are very reluctant to change the status quo of the postwar order. It is safe to say that a majority saw no alternative to the US-Japan security relationship and accepted the existing level of SDF capabilities. This passive antimilitarism demonstrated their opposition to a stance of unarmed neutrality by abrogating the US-Japan Security Treaty and preferred the continuation of the SDF to either the reduction or abolition of the SDF. Japan's Cabinet Office surveys show that those who answered, "(We) protect Japan by means of the Japan-US security regime and the SDF as arranged currently" became the majority (54 percent) in the 1975 survey and remained so continually thereafter (more recently 71 percent in 2000 and 72 percent in 2003) (JCO, 2006). The Japanese perception of national security thus supports the claim that a vast majority of the status quo or "neutral" responses to defense spending has a crystallized opinion and should not be ignored. The status quo responses reflect the continuation of the passive anti-militarism in Japan.
Equally important, the effect of net support may vary with the percentage of those favoring the existing level: the higher the percentage of the status quo, the lower the effect of a given net support, simply because the higher percentage of the status quo response persuades policy makers to maintain the status quo (Higgs & Kilduff, 1993, pp. 227-38). Therefore, the response category of "the same as it is" may be essential to a better understanding of public preferences for defense spending.
Perhaps even more importantly, net support well captures the information of opinion surveys only in a plurality of public opinion divided on defense spending. In big-defense-spending nations such as the United States and Britain, as a percentage of GDP, net support has generally indicated wider cyclic swings of the median preference, which can be seen as a plurality favoring either an increase or decrease. As Figures 3 and 4 show, none of the response categories in Britain and the United States dominate public preferences over time for defense spending. Compared with the swings in those countries, Figure 2 indicates that Japanese public opinion has been more stable and consistent over time. It may be presumed that in the United States and Britain, the contours of median preference lie at the competitive environment of societal demands (net support) while in Japan, those exist in the collective expectation of social consensus (status quo). The active stand of Japanese public opinion on the existing level suggests that the measure of status quo, not of net support, is a plausible representation of citizen opinion on defense spending.

Public Responsiveness -Salience
Japan has a comparatively low level of defense spending as a share of GDP and of the total budget. Does it really become less salient to the public? The concept of issue salience refers to the "importance" that the public ascribes to a given policy issue in relation to other issues (Sokora, 2003, pp. 28-29;Wlezien, 2005, pp. 556-61). The level of spending can be seen as only one of the contributory factors, albeit one of the most influential ones, enhancing the importance of policy issues. When defense spending is less than others, the general public does not necessarily see the issue as less important than others. Although the importance of issues does change over time, the Japanese public has quite consistently ascribed the importance of defense spending to their country-specific historical past (Berger, 1993). In post-WWII Japan, the profound distrust of its own military has been continuously reflected in public opinion. The collective memories of war and militarism shaped Japanese public's strong opposition to full-scale rearmament (Berger, 1993). This presumably high level of salience of anti-militarism, of which a constrained defense budget is one component, expects a greater effect on public responsiveness to defense spending. The Japanese public thus reacts to the slightest change at the comparatively low level of defense spending. Japan has a strong citizen responsiveness to defense spending yet its public preferences are strongly influenced by measurement indexes closely associated with the constant and collective expectation.

Public Responsiveness -Institutionalization
In postwar Japan, public responsiveness to defense spending was institutionalized by the constant and collective expectation of anti-militarism. This made Japanese public preferences more distinct from others. There is substantial evidence that public preferences in the United States and Western Europe are influenced by changes in the constant price of defense spending itself Wlezien, 1995). In contrast, the Japanese public is more attentive to defense spending as a share of GDP. Its defense spending fell below one percent of GDP as early as 1967. The one percent of GDP ceiling on defense spending became a tacit target in the late 1960s within the Defense Agency. By 1974, it was regarded by lawmakers as the firm upper limit and in November 1976, was enunciated by Prime Minister Miki Takeo. This research suggests that the one percent ceiling had been institutionalized to win public support for defense spending (Boei o Kangaeru Jimukyoku., 1975, p. 36;Calder, 1988, p. 437;Chai, 1997, pp. 400-402;Keddell, 1993, p. 55). Governing party politicians anticipated that defense spending would continue to steadily increase in absolute terms as an economic growth would automatically create a larger base of one percent of GDP. The one percent of GDP ceiling was politically feasible for public acceptance while attempting to meet Japan's strategic security needs. Therefore, the Japanese public is more responsive to defense spending as a share of GDP than aggregate defense spending itself.

Policy Representation -Norm Compliance
One would expect that the centralization of the Japanese decision-making apparatus should be even greater in national security policy that generally requires secrecy and isolation from the controversies of domestic politics. Japanese national security policy, however, is a particular area in which the Japanese state is far less autonomous from society than one would expect. In this policy area, the policy networks linking state and society are heavily institutionalized within the collective expectation of anti-militarism. In other words, public preferences have been institutionalized in regularized ways which have become embedded in policy-makers' expectation. Among the formally regularized rules, the most prominent is Article 9 of the Japanese constitution. Although the article officially renounces the use of force as means of settling international disputes, its interpretation has changed over time. The interpretation of Article 9 has been introduced into a set of informally regularized rules or "semiconstitutional rules" (Katzenstein & Okawara, 1993, p. 129). Such institutionalized constraints, which draw the political limits of national security policy, consist of the following: (1) the ban on SDF overseas dispatch, which the upper house adopted as a resolution in June 1954 (conditionally lifted in recent years); (2) the ban on arms exports, which was first announced by Prime Minister Eisaku Sato in April 1967 (allowed in 2004 to export missile defense-related technology to the United States); (3) the three non-nuclear principles, to which Prime Minister Sato referred in December 1967 to prohibit Japan from manufacturing, possessing or allowing entry of nuclear weapons into Japan; (4) the ban on collective self-defense, which became the tacit government understanding of the Japanese constitution in the early 1970s; and (5) the one percent of GDP ceiling on defense spending. The defense budget is an issue area where policy makers receive more public scrutiny than any other national security area. Mass public opinion cannot be institutionally ignored by policy makers. Therefore, the institutionalization of public preferences (based on anti-militarism) with regularized rules effectively constrains the course of budgeting and spending.

Policy Representation -Macro-budgeting
In general, budgeting is a critical component of government management for the successful implementation of fiscal policy. Defense spending is expected to be consistent with economic conditions. To this end, finance ministers need to enforce defense spending macro-budgetary constraints, which come from the limited growth in revenue, the large size of budget deficit, and economic recessions. Nonetheless, Japan's defense budget has been largely regarded as "seiiki" (untouchable spending item) by lawmakers. Despite Japan's fiscal crisis in the early 1980s, the one percent of GDP ceiling allowed defense spending to escape the Finance Ministry's pressure to balance government budgets (JHC, 1982;JHR, 1982). There is no clear fiscal rationale for the rigid one percent of GDP ceiling on defense spending because it is not quantitatively accountable for macro-budgeting purposes such as estimated revenues, expected deficits, and economic forecasts. In fact, the powerful Ministry of Finance (MOF) demands that the defense budget must stay within the GDP framework. Setting the GDP-tied limit on the defense budget provides the MOF with a device to manage defense-spending-specific constraints, which are not necessarily consistent with economic conditions. Therefore, the defense budget is affected, neither by the growth in government revenues nor by the pressures from government deficits, but rather directly by minute fluctuations in defense spending as a share of GDP.

Policy Representation -Systematic Strategy Rationale
A strategic rationale, based on the realist's emphasis on realists' emphasis on structural and material forces, indicates a "back-passing" approach. But even in this view, Japan would need to balance against a threat to the extent that it cannot rely solely on the efforts of the United States. If Japan consistently applied back-passer's defensive strategies to its alliance with the United States, then the cost-minimizer would have significantly increased its independent military capabilities under the following conditions: (1) significant reductions in larger member's strategic contributions to its allies (Lind, 2004, p. 106), (2) immediate security threats to the buck-passer (Snyder, 2002, p. 162), and (3) the larger member's unaccountability and uncertainty to the buck-passer (Cha, 1999). Under each of those conditions, Japan should have acted to preserve an existing distribution of power (balancing) rather than taking no action. These conditions are the limits of Japan's defensive strategy that requires balancing rather than buck-passing. These strategies are supposed to lower US credibility and increase the level of Japan's defense spending. Despite this prediction, Japan clearly did not show an overall trend of non-incremental increases in defense spending during two key phases of significant reduction in US strategic commitments to East Asia (one in the early 1970s and the other in the early 1990s) and during the period of Soviet Far East military build-up in the late 1970s. In contrast, as data, provided by the International Institute for Strategic Studies in the various issues of The Military Balance, demonstrate, South Korea's defense spending dramatically increased from 4.2 percent of GDP in 1980 to 7.6 percent in 1982 as SS-20 IRBM and BAKFIRE bombers in the Russian Far East were deployed within the effective range. Taiwan and the Philippines had already begun their upward trends respectively from 6.9 percent in 1975 to 9.3 in 1976 and from 1.6 percent in 1973 to 3.3 percent in 1977 as once the US Congress passed the Case-Church Amendment. Instead of developing its reliable capabilities in response to the Nixon Doctrine on Asia, Japan's antinuclear sentiment had become stronger than ever before over the debate on Okinawa's reversion. To facilitate the reversion, the party in power adopted a Diet resolution in November 1971 to apply the Three Non-Nuclear Principles to Okinawa. Instead of meeting an international request for a greater role in the post Cold War environment, Japan's constitutional issue over SDF overseas dispatch and use of force resurfaced in Japanese politics. Public opinion established clear outer limits beyond which no party would go if it wished to gain or retain public office. A majority of the Japanese supported the need to revise the ban on overseas dispatch to allow the SDF to participate but only in noncombatant missions related to UN peacekeeping. Meanwhile, the United States continued to demand a more balanced alliance from Japan, while Japan's defense spending does not respond systematically to its changing security environment. This suggests that it is worth knowing if US pressures on Japan, along with the domestic factors, are an important determinant of Japan's national security policy and defense spending. If so, changes in the level of defense spending do not reflect the systematic evaluation of strategic rationale, but rather reactions to external pressures in an ad hoc fashion.

Japanese Public Opinion Does Have a Strong Effect
Measurements employed in this study are specified below for both the responsiveness of public support for defense spending and representation in the defense budgeting process. The results for public responsiveness and policy representation are reported in Tables 1-2 and Table 3 respectively. The findings are then discussed.

Measuring the Responsiveness of Public Support for Defense Spending
In this study, public responsiveness is tested by regression estimates. To test public responsiveness, this modeling employs two different measures for the dependent variable: net support (Note 8) and the percentage of those favoring "the same as it is" (alternative to net support). Japan's Cabinet Office (JCO) has conducted a triennial survey since 1969 allowing the respondents the alternatives of "increases," "reductions," or "the same as it is." The JCO's survey information is adopted as the core series in this study (Note 9). The other information specified in Appendix is also used only if the question was identical in substance with the one posed by JCO's surveys and asked in JCO's non-survey years (Note 10). Regression estimates of change in each of the two dependent variables are made in two sets of equations respectively. Following the existing measurement theory, the first potential independent variable is past levels of defense spending: the annual rate of growth in real and local currency terms (Note 11). The public is presumed not only to be responsive to the annual growth registered in fiscal year t (spending change from fiscal year t-1 to fiscal year t), but also to that in fiscal year t-1 (spending change from fiscal year t-2 to fiscal year t-1) with a time-lag. To test the claim in this study, defense spending as a percentage of GDP is also added as the second potential independent variable for past levels of spending. As the measurement theory commonly used for proving the thermostat model further suggests, apart from past levels of spending, changes in budgetary priorities are included as potential independent variables: a trade-off variable of defense for social spending. The trade-off is computed as percentage differences by subtracting the annual rate of social spending growth in real terms from that of defense spending growth in real terms. The simultaneous and lagged effects of the four independent variables are tested by their values in fiscal year t and fiscal year t-1.
A second group of potential independent variables is related to the effects of external environment. Following the measurement adopted by Richard Eichenberg and Richard Stoll (2003, p. 407), the modeling operationalizes potential threats from major adversaries and neighbors as two potential independent variables: the number of "militarized interstate disputes" (MIDs) in which the Soviet Union/Russia was engaged and the number of MIDs any part of which took place in Asia (Note 12). Following the lagged versions in the literature, it tests the effects by the values of the two threat variables in year t-1, t-2, and t-3.
Given an initial set of the explanatory candidate variables, all possible regressions are explored. Since there are 12 variables, there are 4,095 regression summaries (2 k -1). With help of statistical packages indicated in Appendix, the best model of each size (sorted out by the number of variables) is identified (Note 13).

Measuring Representation and Change in Defense Spending
This modeling estimates variants of regression equations in which the dependent variable is the annual rate of defense spending growth in real and local currency terms. Following the literature, a potential independent variable of public opinion, net support, (which compresses the two response categories, "increases" and "reductions," into one) is constructed. A single response category, "the same as it is," is used as another public opinion variable. In Japan, the process of budgeting starts shortly after the beginning of each fiscal year on April 1, when national ministries begin formulating their budget requests for the following fiscal year. We would expect public opinion during the year t-1 of budget negotiations to affect the approved amount of defense expenditure, and thus evaluate the impact of public preferences on defense spending with a time-lag.
By summer, the Ministry of Finance derives an estimate of tax revenues from the expected state of the economy in the coming year and decides the overall size of the budget for the following fiscal year. Macro-budgetary pressures are thus presumed to affect the size of approved defense expenditure. The measuement uses two types of operationalized macro-budgetary pressures: the annual rate of GDP growth in real terms and debt services ratio (as a percentage of general account expenditure). Since GDP or debt services are estimated for the following year, there should be no time-lag to see the impact of these candidate variables. To test the fiscal constraints of the one percent of GDP ceiling, it also estimates the impact of defense spending as a percentage of GDP in the prior year (t-1) on the change in defense spending for year t.
National security environments are candidate variables peculiar to defense spending. Along with the two threat variables specified in the modeling of public responsiveness, two new external variables are introduced into the equations. The first one is the number of MIDs in which the United States was engaged. This variable is intended to see whether Japan, as a key ally of the United States, acts to preserve an existing distribution of power in support of the US involvement or largely ignores the US involvement by enjoying a cheap ride in defense spending. The modeling further specifies the impact of alliance politics, that is, U.S. pressures for Japan to increase its defense efforts. These pressures are used as the second one, which is measured by "net defense burden": US military spending as a percentage of GDP minus Japan's defense spending as a percentage of GDP. The measures of external conflict/threat and US pressures, which have been debated and legislated, may have a lagged effect on spending change. The regression estimates evaluate them accordingly (the lagged versions t-1 and t-2).
In the regression estimates of policy representation, this study will look at every possible combination of 12 explanatory candidate variables, and produce the best model of each size whose output is sorted by the number of variables (Note 14).

Findings on Public Responsiveness
The findings appear to be mixed and inconsistent with those from the existing studies. However, these apparently mixed results can be interpreted consistently by the alternative measurements utilized in this article. The results are reported in Tables 1 and 2 respectively-whether net support (Table 1) and "the same as it is" (status quo support - Table 2) react to past levels of defense spending and budgetary priorities over different policy areas, and levels of external threats. The first preference, net support, is Japan's marginal calculated from minority opinions favoring either an increase or decrease, while the second preference, status quo support, is a majority opinion revealing the median. Two findings stand out: neither net support nor status quo support reacts significantly to changes in aggregate defense spending while status quo support does closely follow changes in the defense spending/GDP ratio (Note 15). This is in stark contrast to the predominance of the "thermostat" in the United States and Britain where public opinions react negatively and significantly to changes in defense spending (in either the current or previous year). As Table 2 indicates, defense spending as a percentage of GDP is the most consistent, significant predictor of changes in status quo support. The public adjusts its preferences for more (or less) status quo in response to "more" (or "less") spending as a percentage of GDP for fiscal year t-1. This study is interested in this closer, more reliable link between status quo support and the GDP ratio. This statistically significant link is a strong indication of public responsiveness to the one percent of GDP ceiling on defense spending. The median voter would seem to be extremely sensitive to the GDP ceiling rather than to changes in defense outlays themselves.
As seen in Tables 1 and 2, a trade-off of defense for social spending is not significantly related to public preferences, with the exception that the best model of eight explanatory variables for status quo support is slightly significant at the 0.1 level for the trade-off variable at the current year and several equations slightly miss conventional levels (p = 0.1) of significance for the trade-off variable in the current year. On balance, a trade-off effect of defense for social spending is not reliably pronounced on the level of either net support or status quo support. As far as the trade-off effect is concerned, this is consistent with the initial prediction of weak public responsiveness. The public was thus hardly attentive to changes in the relative growth of defense and social spending. This implies that the Japanese public was preoccupied with the early stages of welfare resource allocation, with the ever-increasing expenditures for social security that was less relevant to the level of constrained defense spending in the public eye.
Variables in the external environment have a significant effect on public preference for defense spending. This is somewhat surprising since three candidate factors (no direct war involvements by Japan, Japan's dependence on the U.S. security guarantee, and the illegitimate use of military force by Japan) are predicted to constrain the effects of events in the external environment on public opinion. Public opinion, while revealing no clear signs of the constrained effects, reacts significantly to some measure of the external conflict environment: the conflict involvement of the former Soviet Union/Russia that has a strongest impact on net support (one year lagged) and a negative and reliably significant impact on status quo support (two years lagged). As Table 2 shows, the conflict environment in Asia is also significantly related to status quo support but works its way differently. Both one-year-lagged and two-year-lagged effects of the conflict environment in Asia are significant yet the former increases the level of status quo support and the latter decreases it. This should be an indication of the initial status quo support for the conflict environment in Asia, followed by the growing disenchantment among the Japanese people.
Like people in the United States and Western Europe, the Japanese are equally sensitive to external security threats. The public perception of external threat in Japan showed cyclical swings as declining toward the end of the Vietnam War, bouncing back during the Soviet military build-up in the Far East of the mid-1970s to the early 1980s, dropping again to the 1988-1997 low plateau of the post-Cold War environment, and peaking in 2003 after the 1998 test-firing by Pyongyang of a suspected Taepodong-1 missile (Note 16). Nonetheless, the Japanese public adjusts its preferences for defense spending in its own way. The conflict involvement of the Soviet Union tends to evoke an upward change in public support for defense spending; it is positively related to the minority opinions of net support while corresponding negatively to the majority opinions of status quo support. The conflict environment in Asia has no significant effects on net support but it is a reliable predictor for fluctuations of status quo support.

Findings on Policy Representation
The results displayed in Table 3 show that public opinion does indeed influence defense spending, yet differently from the prediction of the existing literature. The most significant aspect of the results is that defense spending as a share of GDP exerts the strongest effect on defense spending. Its one-year-lagged effect on defense spending indicates a negative, significant coefficient; this fits perfectly with the claims made in this article. Since Japanese policy elite and the public clearly differ in their beliefs of defense policy, Japan's social consensus over defense spending by itself is not necessarily the independent source of influence. The representation of social consensus requires the institutionalization of public preferences or the exercise of material leverage in motivating Japanese policy elite to adopt these collective expectations into political action. Defense spending as a share of GDP, which reflects a compliance mechanism embedded within the regularized rule of the GDP ceiling to win public support, is a primary constraint on policy elite. Net support is not even a weak significant predictor of changes in defense spending. This suggests that Japanese policy makers are indirectly far more sensitive and responsive to status quo support that closely follows changes in the defense spending/GDP ratio ( Table 2).
The results show that macro-budgetary consideration, GDP annual growth and government debt services, has no significant effects on changes in defense spending. The more government spends on debt services, the less government has freedom in providing other services (including defense). In contrast, the higher the GDP is, the larger the tax base it creates. The resulting tax revenue provides more scope for fiscal flexibility. But these fiscal constraints do not act upon Japan's defense spending. Japanese defense budgeting is not significantly accountable for macro-budgeting purposes, but rather responsive to the politically viable GDP-tied-ceiling at the domestic level.
The striking feature of the external-variables-effects is the predominant fact that Japanese defense budgeting is clearly sensitive to US pressures but not to the external conflict environment (though the two-year-lagged conflict involvement of the Soviet Union/Russia is weakly significant in two equations). US pressure is a reliable predictor for measures of change in defense spending. The candidate variable, net defense burden, which is entered with lag lengths of one to two years, is positively and significantly related to changes in defense spending. As the gap between US and Japanese defense burdens (borne by the United States in a disproportional way) widens (narrows), Japanese defense budgeting increases (decreases) the level of spending. As some methodologically descriptive studies suggest, burden-sharing has been a continuous policy issue and the subject of various negotiations between the United States and Japan (Keddell, 1993, pp. 78-124). The lag lengths thus seem to allow debates and discussions in the United States and put pressure on Japan, influencing changes in Japanese defense spending.
One interesting finding concerns the impact of US military involvement abroad on Japanese defense spending.
As Table 3 shows, the number of MIDs in which the United States was engaged is a weakly significant predictor of changes in Japanese defense spending. The coefficient is negative and indicates the 0.05 level of significance in two equations (US MIDs t-1) and one equation (US MIDs t-2). The more the United States gets involved in MIDs, the less Japan spends on defense. There is thus a hint of policy representation whereby Japanese policy makers tend to not take the US military involvement as an external threat factor (where Japan may become involved) but to deepen their dependence on the US security guarantee (which may imply a sign of free riding).

Discussion and Conclusion
The thermostat model suggests that there is an equilibrating relationship between public opinion and government behavior over time. The findings of the Japanese case in this study are consistent with this continuous equilibration of public preferences and defense budgeting. However, the study shows that the theory of measurements commonly used for proving the thermostat model is not suited for detecting the double-sided democratic equilibration between attentive publics and responsive policy-makers in the case of Japanese defense spending. It suggests that the context of domestic structures and coalition-building process needs to be reflected on measurement indices. Our results have found that with the introduction of alternative measurements indices, Japanese defense spending is in fact strongly representative.
The representation of public preferences in policy making is the key to democracy-building. Defense spending has been considered a special area to which the public is more attentive than any other national security area. In the case of Japan's policy representation, defense spending is indeed one of the most salient policy areas. The regression estimates in this article present compelling data that Japanese defense spending is in fact strongly representative. However, there is little evidence that the dynamics of representation in the case of Japan's defense spending is consistent with the results found in the literature. In the United States and Britain, where constant building and rebuilding of coalitions takes place in the absence of societal consensus, their society-dominated policy network provides the shifts in public opinion with opportunities to influence policies. There is a striking contrast with Japanese public opinion on defense spending, which has been far more stable and consistent than any other industrialized democracy. Japan is normally seen as a strong state, but its national security policy is one issue area where the state can neither persuade nor resist societal demands. In particular, the homogeneous public opinion on defense spending not only permeates the state institution, but also helps to shape the link between society and the political system. In fact, Japanese defense spending is significantly constrained by the institutionalization of popular pressure, such as the one percent of GDP ceiling, in the elite-public policy network.
The observation of the results suggests that we need to note conditions under which the case of Japanese defense spending does not duplicate the thermostat phenomenon although the generalizations of the thermostat process are still consistent to the extent that public opinion does not change because spending does not vary. Obviously, it is in a plurality favoring either an increase or decrease in defense spending that the dynamics of public preferences in the United States and Britain uniformly resemble a thermostat. Japanese society is probably the most homogeneous of all industrialized democracies and public responsiveness sticks to the status quo in the current defense budget. While this is certainly true, democratic representation requires Japanese policy makers to respond to the institutionalized constraining forces of societal consensus. Further measurement and interpretation of defense spending, along with the improved availability of a reliable time series of public preferences, would do well to take into account this more complex view of defense budget policymaking.

Public Responsiveness
Issue salience evidently plays a key role in influencing the degree of public responsiveness. In comparative perspective, defense budgeting appears to be most salient when a state continues to bear a greater defense burden ratio (defense spending as a percentage of GDP/GNP or as a percentage of government spending). Another comparative difference is the relative responsiveness of public opinion to fluctuations in defense spending. Defense spending is more salient in those countries where it varies greatly and the public is thus more attentive to it. Given the larger size of, and greater fluctuations in, defense spending in the United States and Britain, public responsiveness has been most significant in these two advanced democracies. In that Japan's defense budgeting represents one of the lowest defense burden ratios and varies least among advanced democracies the Japanese public should be weakly responsive to it. However, the findings suggest that Japan's defense budgeting, according to different measurements of salience, is actually highly salient; citizen opinion in Japan is remarkably attentive to defense spending.
Perhaps budget size and volatility are one way of measuring salience, with other measurements yet to be explored. The importance of policy issue for voters (salience) may not only be measured by mathematical/physical values but also by non-material values. Specifically, the current literature needs to take into account the nature and patterns of political culture (such as Sweden's popular support for armed neutrality), which specific historical paths help to shape. The greater gap between public preferences and policy itself will not necessarily be of more importance to voters. Even if no gap exists, voters may still care about and have crystallized meaningful opinions on the issue. Japan's defense spending is a case in point. It varies least, and while the Japanese public wishes to maintain the status quo, the majority does not simply represent their neutral stance, but actually take a stand. The findings show that the majority of Japanese voters are remarkably sensitive to the slightest changes in the ratio of defense spending to GDP. The Japanese public ascribes the importance of defense spending to the country-specific attributes of the domestic political setting: Japan's postwar culture of anti-militarism.
On balance, citizens in the United States and Britain tend to react to changes in defense spending in a thermostat way, even strongly so when defense spending grows faster than non-defense spending. The reaction is measured by net support that is claimed to reflect the median preference. However, it is highly unlikely that net support effectively follows the contours of the median preference in post-WWII Japan. Japanese public preferences for keeping the same level of defense spending and for military security have accounted for the greater degree of public consensus as well as the lesser degree of societal fragmentation. The status quo support continues to be by far the dominant public view, reaching nearly half of the adult population in 1975 and remaining a three-fifths majority for the past decade. The reliable time series of public opinion demonstrates that we cannot justifiably ignore the predominant category of status quo as "non-attitudes" since at least 80 percent of the respondents continue to have had a crystallized stand on national security. The vast majority of Japanese people have no wish to return to the militarist past yet they have no desire to change the status quo of the postwar order: the US-Japan security relationship and the existing level of SDF capabilities. It is safe to say that the Japanese public has been strongly attentive to defense spending.
In contrast to the existing literature, the findings indicate that changes in the spending gap between Japanese defense and social policy do not produce any effective trade-off in public preferences. The trade-off effect of defense for social spending is not significant, probably due to the fundamental reason being as a Japan's heritage of developmental state. Japan's social policy began to dramatically expand in the early 1970s and thus redirected its attention and resources to social needs. In other words, Japan experienced the first stage of welfare resource allocation. Since then, the most important single change in spending patterns has been the ever-increasing expenditure for social policy. Japan's social spending has grown much faster than its defense spending in a disproportional way while in more liberal welfare states such as the United States and Britain (where the state emphasizes the market by passively providing modest social benefits), in the public eye the growth patterns of social and defense spending are competitively comparable. Unlike the US and British public, the Japanese public thus seems to be less responsive to the relative growth, with no necessity for capturing the comparable spending gap.
Not surprisingly, as in the United States and Western Europe, the Japanese public has equally been quite sensitive to external threats or real-world developments. Furthermore, the Japanese public reliably adjusts its preferences accordingly. This result is somewhat unexpected since public mistrust of the state ability to use armed forces remains intact. Overwhelming majorities of Japanese firmly oppose the SDF becoming involved in overseas combat, even within a UN peacekeeping framework. It is conceivable that the study findings would indicate that the Japanese public does not support the use of armed forces but does support military readiness.

Policy Representation
The relatively centralized Japanese political system is less open to public pressures. This may suggest that Japan would have weakly representative policy-making, but the state institutions do not directly determine its specific scope and nature. To explain specific policy outcomes, the mechanisms and processes of interest representation need to be examined. In post-WWII Japan, national security policy has remained the one policy area where national leaders and the general public clearly differ in their beliefs. We need to examine the link between public preferences and elite decisions in defense spending, how the institutional arrangements between state and society channel societal demands into budgetary decisions. In this process, despite its relatively centralized political system, Japan might not always be able to conduct highly efficient policies from the viewpoint of national leaders. Part of the policy networks in Japan would come close to being society-dominated, depending on the dynamics of societal demands from one issue area to another. This study has found that there is a compelling suggestion that the Japanese state is far less autonomous from society in defense spending.
The literature suggests that the coalition-building processes for budgeting take place on the basis of familiar ground rules, and thus Japan's coalition-building of defense budgeting is carried out within the familiar territory of the quasi-corporatists' framework. The Japanese policy network has been characterized by state-led corporatism in which the state is able to block or capture some societal demands including labor union ones. This article has argued that the area of Japan's defense spending remains strongly responsive to the institutionalized social consensus on defense expenditure. In the processes of Japanese defense budgeting, political and societal actors would not be strongly engaged in continuous bargaining processes at each stage of decision making. Rather, the institutionalization of public preferences with regularized rules effectively constrains and structures the course of budgeting and spending. As a result, budgetary decisions on Japanese defense tend to reflect a somewhat middle-of-the-road-policy, representing the common denominator of public preferences derived from the post-WWII culture of anti-militarism and anti-defense spending bias.
Overall, Japan's defense budgeting seems to have been made with relatively weak regard for potential external security threats but heavily influenced by alliance dynamics -US pressure. Japan's lack of response to potential external security threats would thus seem to provide a clue to understanding the anomalous nature of Japan's reluctance to convert its economic resources into military capabilities. According to a variation of defensive realism, this apparent lack of response implies that Japan has still been actively responsive in a strategically defensive way. In its view, Japan has strategically chosen a defensive strategy of "buck-passing" at minimal cost; to hold back by shifting the cost of fighting onto the United States. To be consistent with the logic of defensive strategy, however, Japan should have balanced against a threat while transferring as much of balancing costs as possible to the United States. As discussed previously, patterns of Japanese defense spending indicate otherwise. Japan's incremental spending did not recognize the need to balance against a threat when it is under the conditions where balancing is preferable to buck-passing.
In short, though driven by self-interested, utility-based objectives, the cost minimizer of Japanese national security is compelled to operate within a normative rationality, based on shared, historically contingent standards of precedents. Defense spending, which is one component of Japanese national security, is heavily influenced by the ways in which antimilitarist norms interact with material calculations. The institutionalized constraint of the one percent ceiling, while coming from normative considerations, is intended to win public support for political gains. The complex ways of balancing norms and material calculations significantly account for public responsiveness and policy representation in the issue area of Japanese defense spending.
Shinbun (Y). The core source is a time series of public preferences for Japanese defense spending conducted by Japan' Cabinet Office (JCO) every three years since 1969 (1969,1972,1975,1978,1981,1984,1988,1991,1994,1997,2000,2003,2006). The survey question reads: "Do you prefer increases, the same as it is, or reductions?" The other polling organizations provide an occasional series of data on public opinion regarding defense spending. This article compiled a total of 99 national opinion surveys taken on defense spending by these organizations from 1968 to 2006. All the polling organizations have used similar methods; the survey information adopted by the organizations is based on personal interview surveys with a sample size of 3,000 (except for JCO's survey in 2000 with a sample size of 5,000). This article used the data provided by JCO as the main source and also the data by other than JCO only if the question was (1) identical in substance with the one posed by JCO's surveys and asked in JCO's non-survey years; and (2) was not about whether Japan's Self-Defense Force (SDF) should be enlarged but about spending itself. The JCO information was supplemented accordingly by an occasional series of public preferences for Japanese defense spending conducted by six polling organizations: M (1968, 1971), NHK (1974NHK ( ), K (1980NHK ( ), Y (1982NHK ( ), J (1987NHK ( , 1989NHK ( , 1990, and A (2004).

The Militarized Interstate Dispute (MID) Participant Data
The external threats were measured by the number of MIDs involving the adversary and the number of MIDs taking place in the region (inferred from the locations of the states that are involved in the MID). The MID data (Version 3.1), 1816-2001, provided by the Correlates of War Project, were used to count such disputes. This data set records all events of when one state threatens, displays, or uses armed force, short of war, against another. This article also relied on Keesing's Record of World Events for the missing information, 2002-2006.