Testing the four hypotheses outlined in Chapter 2 required a means of measuring my quantity of interest, which is candidate electoral strategy. In an ideal world, we might want to run the following experiment. First, we would randomly assign candidates to multimember or single-member districts and to ruling and opposition parties, respectively, and inform them that the constitutional structure under which they would be expected to deliver on promises made during election campaigns was British-style parliamentary. Second, we would announce an election. Third, we would place each candidate in a room on their own, give them a piece of paper of identical size, and instruct them to write anything they wanted on this piece of paper, with the proviso that it would be the only information voters would receive about them before casting their vote. Finally, we would collect up all these pieces of paper and use them to compare the extent to which candidates running in both kinds of districts and from both kinds of parties emphasized pork, policy, and within the latter category, national security policy. While this experiment is impossible in the real world, a combination of near-draconian campaign restrictions in Japan, the practice of requiring candidates for office to produce candidate election manifestos, which local electoral commissions are required to distribute to all registered voters in the district before the election, and advances in methods for quantitative text analysis enabled me to construct indicators of candidate electoral strategy that approximate this ideal for the 7,497 serious candidates running in the eight House of Representatives (HOR) elections held between 1986 and 2009.

This chapter proceeds as follows. First, it explains why existing approaches to measuring candidate electoral strategy do not capture my quantity of interest. Second, it explains why the persistence of unusual, near-draconian campaign restrictions across time, space, and technology in Japan have the effect of making a medium of communication that is available to all candidates, the candidate election manifesto (senkyo koho in Japanese), reliable material from which to measure this quantity of interest. Third, it addresses potential threats to the reliability of the manifesto, including the possibility that it is unimportant and therefore not taken seriously by candidates and the possibility that the party leader is disciplining what conservative candidates write in their manifestos after electoral reform.

Chapter 1 described the puzzling turnaround in attention to national security by conservative Japanese politicians in 1997. Chapter 2 argued that this turnaround is best explained by a shift in their electoral strategies from pork for groups of voters in their respective districts to broad policy issues such as national security, brought about by electoral reform to the House of Representatives (HOR) in 1994. Chapter 3 explained how candidate election manifestos and quantitative text analysis were used to obtain measures of the degree to which each of the 7,497 serious candidates competing in the eight HOR elections held between 1986 and 2009 were relying on pork, policy, and within the policy category, national security policy. Chapter 4 used those measures to test Hypothesis 1, which contended that conservative candidates adopted electoral strategies dominated by pork prior to electoral reform, Hypothesis 2, which contended that conservative candidates facing higher levels of intraparty competition adopted electoral strategies of more pork than conservative candidates facing lower levels of intraparty competition, and Hypothesis 3, which contended that conservative candidates adopted electoral strategies dominated by broad policy issues like national security after electoral reform.

This chapter focuses exclusively on the role of national security in their electoral strategies. It proceeds as follows. First, it presents a descriptive account of National Security Policy (Topic 6) and explains why the level of discussion of this topic in the manifestos of conservative candidates contesting these eight elections provides a reliable estimate of the degree to which these candidates focused on national security in these elections. Second, it demonstrates that these candidates adopted electoral strategies comprised of no national security under the old electoral system and switched to adopting electoral strategies comprised of national security under the new. It also provides evidence that as expected, national security expertise was not something to flaunt under the old electoral system, but something to hide.

While the evidence presented in this and the previous chapter supports my claim that conservative politicians started paying attention to national security after 1997 because their electoral strategies changed, both the shift in strategies and turnaround in attention to national security are consistent with a number of alternative hypotheses, none of which have anything to do with electoral reform.

In October 2012, the Japanese media covered a debate within the Japan Restoration Party. In a forum to unveil the new party's policies to voters, party leader and Osaka Mayor Toru Hashimoto offered his thoughts on the dispute between Japan and South Korea over sovereignty of the islands known as Takeshima in Japanese and Dokdo in Korean. Acknowledging the impossibility of overturning South Korea's occupation of the islands with force, Hashimoto proposed that Japan change its policy to a more practical one: joint management of the islands with its neighbor (Yomiuri Shimbun 2012). No sooner had he spoken than his words met with a barrage of criticism from fellow party members. Former Liberal Democratic Member of the House of Representatives Kenta Matsunami asked him to “leave decisions about national-level policies like foreign and security policy up to the Diet Members in the party,” which prompted Hashimoto to clarify that he was not suggesting Japan rescind its claims to sovereignty of the islands, only that it work with South Korea to establish rules for the joint utilization of the area and its resources. Another former Liberal Democratic Member of the House of Representatives, Kenzo Yoneda, counseled Hashimoto against further efforts to resolve the dispute, warning that “national security and territorial disputes are the lifeblood of politicians these days” (Sankei Shimbun 2012).

Yoneda's remarks are telling because they overturn a conventional wisdom about Japan, which is that conservative politicians do not pay much attention to national security and are not very interested in making security policy. Scholars of international relations and observers of recent episodes of tension between Japan and China over islands in the East China Sea might find this difficult to believe, but for decades, it was true. The conservative politicians who governed Japan as members of the Liberal Democratic Party (LDP) acted as if they were oblivious to the security threats Japan faced and indifferent to the opportunities arising from the extraordinary expansion of Japan's economic power and standing in the world. In 1997, this changed. All of a sudden, these same conservative politicians were rushing to create Diet Member leagues to tackle matters of national security, clamoring to make statements about security issues in newspapers and on television, and devising ways to make them part of their election campaigns.

Chapter 1 described the puzzling turnaround in attention to national security by conservative Japanese politicians in 1997. Chapter 2 argued that this turnaround is best explained by a shift in their electoral strategies from pork for groups of voters in their respective districts to broad policy issues such as national security, brought about by electoral reform to the House of Representatives (HOR) in 1994. Chapter 3 explained how candidate election manifestos and quantitative textual analysis were used to obtain measures of the degree to which each of the 7,497 serious candidates competing in the eight HOR elections held between 1986 and 2009 were relying on pork, policy, and within the policy category, national security policy. This chapter uses those measures, merged with the Japan MMD Data Set and Japan SMD Data Set (Reed and Smith 2007, 2009), to test Hypothesis 1, which contended that conservative candidates adopted electoral strategies dominated by pork prior to electoral reform, Hypothesis 2, which contended that conservative candidates facing higher levels of intraparty competition adopted electoral strategies of more pork than conservative candidates facing lower levels of intraparty competition, and Hypothesis 3, which contended that conservative candidates adopted electoral strategies dominated by broad policy issues like national security after electoral reform. Subsequent chapters address the role of national security in their strategies and the strategies of candidates affiliated with the opposition parties.

Chapter 1 described the puzzling turnaround in attention to national security by conservative Japanese politicians in 1997. Chapter 2 argued that this is best explained by a shift in their electoral strategies from pork for groups of voters in their respective districts to broad policy issues like national security, brought about by electoral reform to the House of Representatives (HOR) in 1994. Chapter 3 explained how candidate election manifestos and quantitative text analysis were used to obtain measures of the degree to which each of the 7,497 serious candidates competing in the eight HOR elections held between 1986 and 2009 were relying on pork, policy, and within the policy category, national security policy. Chapter 4 used those measures to test Hypothesis 1, which contended that conservative candidates adopted electoral strategies dominated by pork prior to electoral reform, Hypothesis 2, which contended that conservative candidates facing higher levels of intraparty competition adopted electoral strategies of more pork than conservative candidates facing lower levels of intraparty competition, and Hypothesis 3, which contended that conservative candidates adopted electoral strategies dominated by broad policy issues after electoral reform. Chapter 5 demonstrated that conservative candidates adopted electoral strategies comprised of no national security prior to electoral reform and strategies comprised of national security after electoral reform. It also presented evidence that the shift in strategies and turnaround in attention to national security are not explained by seven categories of alternative explanations.

This chapter focuses on the electoral strategies of candidates running from the seventeen major opposition parties that fielded candidates in these eight elections. The theory outlined in Chapter 2 explained why the absence or extremely low levels of intraparty competition faced by these candidates would have given them incentives to adopt electoral strategies dominated by broad policy issues under both electoral systems. The theory also explained why the need to capture a relatively small proportion of the vote to win a seat under the old system, Single-Nontransferable-Vote in Multimember Districts (SNTV-MMD), would have inclined them toward positions on these issues that were relatively ideologically extreme (e.g. Cox 1990; Kohno 1997; Pekkanen and Krauss 2005; Maeda 2012).

A recent article by two prominent international relations theorists suggests that hypothesis-testing is “bad” for international relations (Mearsheimer and Walt 2013). These scholars suggest that “the creation and refinement of theory” is the most important pursuit international relations scholars ought to be engaged in. This book took an empirical puzzle about postwar Japan, namely, why the politicians who governed Japan for more than fifty years paid so little attention to national security before 1997 and so much attention after 1997, used theory from comparative politics to generate testable hypotheses that could explain this puzzle, and collected a wealth of new data to evaluate these hypotheses. It also used other theories to generate eight categories of alternative hypotheses and used other new data to evaluate these. Its conclusion, that the electoral strategies these politicians were forced to adopt under Japan's old electoral system made the costs of paying attention to national security prohibitive, and it was only when their electoral strategies changed after electoral reform in 1994 that the costs were lowered, contributes a new theoretical insight to the sub-field of international relations: that electoral systems influence security policy by influencing the costs of making security policy. This suggests that theory and careful hypothesis-testing need to go hand-in-hand, for it is only by elucidating and carefully testing hypotheses that new theories may be advanced.

The contributions of the book span the sub-fields of international relations and comparative politics, as well as the study of Japanese politics and Japanese security policy. For international relations, the book demonstrates that electoral systems influence security policy by influencing the amount of attention politicians can pay to it and the costs of changing it. Crucially, this influence is exercised independently of the level of concern politicians or their voters might have about national security and independently of any policy preferences or ideological inclinations they might hold. This confirms the findings of a growing body of work in international relations that establishes the impact of domestic political institutions on security policy (e.g. Bueno de Mesquita et al. 1999, 2003; Schultz 1999; Kaarbo and Beasley 2008; Davis 2008; Hymans 2011). It also demonstrates that democracies, like autocracies (Weeks 2008), vary in the extent to which people can hold leaders accountable for the provision of national security and other policies.

Chapter 1 described the puzzling turnaround in attention to national security by conservative Japanese politicians in 1997. Chapter 2 argued that this is best explained by a shift in their electoral strategies from pork for groups of voters in their respective districts to broad policy issues like national security, brought about by electoral reform to the House of Representatives (HOR) in 1994. Chapter 3 explained how candidate election manifestos and quantitative text analysis were used to obtain measures of the degree to which each of the 7,497 serious candidates competing in the eight HOR elections between 1986 and 2009 were relying on pork, policy, and within the policy category, national security policy. Chapters 4, 5, and 6 used those measures to test four hypotheses derived from the argument. Chapters 5 and 6 derived eight categories of alternative hypotheses that could plausibly account for both the shift in strategies and the turnaround in attention to national security, and tested those using measures of the ideological positions adopted by candidates during these eight elections, the topics of 126,275 voter petitions presented to the Diet during this time, public opinion polls, and other new data. Together, these chapters provide compelling evidence that conservative Japanese politicians began paying more attention to national security in 1997 because their electoral strategies changed, and not because of any other variable.

This begs the question, what influence has their new attention to national security had on Japanese security policy? Before answering this question, it is important to think about why it is difficult to evaluate the relationship between their new electoral strategies and Japanese security policy. This is because we cannot observe what Japanese security policy would have looked like in the absence of their new electoral strategies. The standard approach for scholars interested in evaluating the causal impact of a treatment like electoral reform on an outcome like security policy is to identify a set of countries that resemble Japan, some of which received the treatment and some of which did not, and examine what their security policies look like after some received the treatment. If the security policies across the two sets of countries were then found to vary, we could attribute that variation to the effects of the treatment because we know the countries were similar in other regards before they received the treatment.