Full Text

1 Introduction A number of US overseas bases have been deployed around the world to protect allies and maintain regional peace. During the Cold War period, US overseas bases contributed to the development of economies in Western allies because US overseas troops reduced the future uncertainty and encouraged Americans to invest in the major sectors of its partner countries. US overseas bases have been essential for securing resource flows especially from the Middle East since the 1970s. After the 9/11 incident, US foreign bases have been centered on anti-terrorist operations. US overseas bases have been playing the critical role in formulating the world order. However, this is a rare practice in the history of international relations and, except for the formal empire, major powers had not stationed overseas bases for the long-term until World War II. The problem of base politics is in the delicate interactions between the United States and the host partners (Calder, 2007, Chapter 1). The United States does not always maintain long-term overseas bases in the host countries. For example, while Germany and Japan have been hosting US bases for more than fifty years, the Unites States has withdrawn overseas bases from South Vietnam (1975), Iran (1979), and Philippines (1991) in the face of strong local opposition. Therefore, understanding local support for US overseas bases is indispensable for managing alliance politics and pursuing US grand strategy. This article addresses base politics in Okinawa. Okinawa is located in the Southeast region of Japan and consists of an archipelago of small islands close to Taiwan. Okinawa is the only one prefecture of Japan, but it plays the pivotal role in Japan’s national security policy as more than 70% of US bases in Japan are stationed in Okinawa to deter challenges to the 92 Koji Kagotani and Yuki Yanai at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from status quo and protect Japan under the US–Japan security treaty. The number of American military personnel in Japan was roughly 39,000 during the 2000s, but it increased to 50,000 in 2012 in response to recent security challenges from China.1 Okinawa hosts the majority of American troops stationed in Japan whose function is to engage in military operations to secure Japan and maintain peace in East Asia after the closure of Clark Airfield and Subic Bay in the Philippines in 1991 and 1992, respectively (Calder, 2007). On the other hand, Okinawans are exposed to the risks of airplane crashes, environmental and noise pollution, and rape incidents associated with US stationed troops. Since the reversion of Okinawa to Japan in 1972, US bases have been a prime political issue and an anti-base candidate generally competed with a pro-base candidate in every election for the Okinawa governor. Okinawa is a special case where local politics is strongly linked with Jpan’s national security policy. Thus, the analysis of Okinawa base politics will provide policy implications for managing the US–Japan security treaty. The Okinawa case provides an interesting disjuncture between Okinawans’ anti-base feelings and vote choice. In Okinawa, three historical episodes form its antimilitarist culture: Okinawa’s forced annexation by Japan’s feudal government in the early Edo period (in the early seventeenth century), its experience of scorched battlefield at the end of World War II, and its occupation by US military forces since 1972 (Cooley and Marten, 2006, pp. 568–570). Okinawans hold anti-base feelings due to the possible clash of fighter jets and helicopters, noise pollution, and crimes associated with US bases and soldiers. For example, the 1995 Okinawa rape incident that provoked widespread grievance called into question the continued presence of US bases in Okinawa (Funabashi, 1999; Yeo, 2011, Chapter 3). Also, US bases can invite foreign challenges because they are essential for military operations to counter adversaries’ hostile behavior. Thus, strong opposition to US bases has to been prevalent in Okinawa, and US base policy has proven to be a salient issue in the Okinawa gubernatorial elections. Nevertheless, in the 1972–2006 Okinawa gubernatorial elections, locals supported pro-base candidates six out of ten times contrary to their anti-base preferences. 1 US Department of Defense provides the historical data on active duty military personnel by service by region/country. URL: siadapp.dmdc.osd.mil/personnel/MILITRY/miltop.htm (7 March 2013, date last accessed). External threats and voting in Okinawa 93 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from Why can a pro-base candidate win an election in spite of Okinawans’ strong anti-base feelings? A few studies on base politics point to the importance of material interests such as financial transfers and base-related subsidies. In order to consolidate pro-base support, the central government provided more public works projects to mitigate unemployment, set up provisions for preventing military accidents and reducing noise pollution, and paid the rent for military reservation (Calder, 2007, Chapter 6). Thus, these selective incentives form a base for pro-base feelings. Even after the 1995 rape incident, these material interests encouraged locals to support a pro-base candidate in the gubernatorial election and to abstain from a local referendum regarding opposition to US bases (Eldridge, 1997; Cooley and Marten, 2006). This is one of the arguments that we are going to test statistically. Also, we tackle this puzzle in the nature of base politics by focusing on Okinawans’ reaction to external threats. We analyze the gubernatorial elections as the opportunities for Okinawans to covey their support for or opposition to Japan’s current security policy since US bases in Okinawa are critical to Japan’s security. We find that external threats encourage Okinawans to support pro-base candidates but the effect of external threats on vote choice is moderate. Moreover, physical and psychological costs such as airplane crashes, environmental and noise pollution, and rape incidents seem to influence election outcomes rather than material benefits such as the fiscal transfers and base-related subsidies, which is contrary to the conventional view. The connection between foreign crisis and public opinion has been attracting a lot of scholars in international relations. Even though there is no consensus over the evidence, some have confirmed the rally-’round-the-flag effect (or simply the rally effect) in the United States and the United Kingdom (Parker 1995; Edwards and Swenson, 1997; Nickelsburg and Norpoth, 2000; Lai and Reiter, 2005). However, they have yet to sufficiently address vote choice itself and only a few studies present the mixed results on the influence of foreign policy outcomes on vote choice in the United States. For example, Anand and Krosnick (2003) found that in the 2000 presidential election, Americans had different attitudes towards foreign policy goals and supported the candidate who pursued their preferred goals. According to Gelpi et al. (2005/2006), in the 2004 presidential campaign, one-third of the voters cited foreign policy issues as the most important determinant of their vote choice and their 94 Koji Kagotani and Yuki Yanai at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from attitude towards the Iraq war had a significant impact on vote choice. On the other hand, Wlezien and Erikson (2005) show that the Iraq war did not work against President Bush’s electoral performance. Thus, we seek to contribute a non-US/UK case to the rally-’round-the-flag literature and extend base politics research by investigating the relationship between external threats and vote choice.2 As discussed in more detail later, we begin by examining the relationship between external threats and vote choice. In Section 3, we explain Okinawans’ attitudes towards US bases and the US–Japan security treaty. In Section 4, we present a statistical model to analyze the 1972–2006 Okinawa gubernatorial elections. In Section 5, we show the results that external threats bolster the vote share of pro-base candidates, but the effect of external threats is moderate. The results also demonstrate that physical and psychological costs determine Okinawans’ voting behavior rather than material interests. We discuss the implications for managing the US–Japan security treaty in our concluding remarks. 2 Theory Since World War II, the baseline of Japan’s security policy has been based on its Self-Defense Force in collaboration with US stationed troops. Okinawa has been hosting >70% of US overseas bases in Japan. Since the reversion of Okinawa to Japan in 1972, the US base issue has been politicized in the Okinawa gubernatorial elections. A pro-base candidate competed against an anti-base candidate in the elections and the electoral outcomes changed the difficulties in consolidating pro-base support to pursue military operations steadily under the US–Japan security treaty. Thus, Okinawa gubernatorial elections are strongly tied to national politics and provide Okinawans with an opportunity to indicate their support for or opposition to the current security policy of Japan. Electing a pro-base candidate represents Okinawans’ approval of national security policy, whereas electing an anti-base candidate implies Okinawans’strong opposition towards the policy position of national leaders. Therefore, this special characteristic of Okinawa base politics allows us to discuss a local election in the context of the politics of national security. 2 SeeAldrich et al.(2006)for the review. External threats and voting in Okinawa 95 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from Okinawans supported the pro-base candidates in six out of ten gubernatorial elections in 1972–2006 contrary to their anti-base preferences. We explore the determinants of an increase in pro-base support by examining Okinawans’reactions to external threats. A surge of pro-base support in an election can be considered to represent both Okinawans’ emotional and rational reaction to an increasing number of security-related risks surrounding Japan. The rally effect explains Okinawans’ emotional reaction to external threats. The literature predicts that external threats will increase support for the national leader or national security policy and thus will increase the vote share of pro-base candidates. One implication of this theory is that voters’reactions are reflexive. We have several reasons for the rally effect. First, external threats wave the flag and patriotism encourages voters to support their national leaders (Mueller, 1970). In general, this phenomenon is explained by the in-group/out-group argument; more conflicts with an out-group may enhance ties within the in-group (Levy, 1989). Second, less criticism by the political opposition helps public opinion align with the national leaders in the time of crisis (Brody, 1991). The opposition may also support their national leaders only when conflict with adversaries is important to national interests (Schultz, 2001). Third, external threats are more likely to induce more favorable views of the state of the economy, politics, and society in public mind. Voters with such favorable views are more likely to support their national leaders (Norpoth, 1991; Parker, 1995). Therefore, the rally effect can increase Okinawans’ approval of national security policy and encourage more Okinawans to support a pro-base candidate. In contrast, the demand for national security explains Okinawan’s rational reaction to external threats. The literature on military spending is helpful for understanding the Okinawans’rational calculations. The panel analysis of 165 countries in the period 1950–2000 shows that the increasing likelihood of interstate conflict raises military expenditure in a country (Nordhaus et al., 2012). During the cold war, military build-up by the Soviet Union increased Japan’s and US budget allocations to military spending, particularly affecting Japan, which was more sensitive to the threat posed by the Soviet Union than the Unites States was (Okamura, 1991). Furthermore, US military spending even increased in response to public anxiety about national security (Hartley and Russett, 1992). Thus, external threats generate the demand for more national security and 96 Koji Kagotani and Yuki Yanai at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from encourage the public to, in turn, influence national security policy. In the context of Okinawa base politics, we predict that if Japan is confronted with an increasing number of security-related risks, the demand for US bases and stationed troops will prevail among Okinawans who will be willing to support a pro-base candidate in the gubernatorial election to enhance military operations in the US–Japan security treaty. Therefore, both Okinawans’ emotional and rational reactions to external threats can determine the trend of pro-base support over time and we have the following hypothesis. H1: The more external threats Japan is confronted with, the more voters will support a pro-base candidate in the Okinawa gubernatorial election. 3 Okinawans’ attitudes For the empirical analysis, we begin by examining Okinawans’ attitudes towards US bases in Okinawa before testing H1. Confirming the degree of Okinawans’ opposition to US bases allows us to understand the difficulty inherent in securing support for pro-base candidates in gubernatorial elections. The cabinet office of Japan conducted the Okinawans’ Attitude Survey in 1973, 1975, 1976, 1977, 1981, 1985, 1989, 1994, and 2001. The Cabinet Office had asked only about living circumstances before 1981, whereas it also asked about their attitudes towards US bases in Okinawa after 1985. It asked the question, ‘What is your view of the importance of U.S. bases stationed in Okinawa? Please choose the alternative that is the most close to your view’. Okinawans choose one of five alternatives: (i) US bases are imperative for Japan’s security, (ii) US bases are important for Japan’s security, (iii) US bases are not necessary for Japan’s security, (iv) US bases deteriorate rather than help Japan’s security, and (v) Do not know. The sample size of each survey was 2000 respondents, and the response rate was
77% on average in the designated time frame.3 Figure 1 shows the trend of Okinawans’ attitudes towards US bases. The share of those who supported US bases represents people who chose answers (i) and (ii), whereas the share of those who opposed the US bases represents people who answered (iii) and (iv). Near the end of the Cold 3 You can download survey questions and results from the website of Japan’s Cabinet Office. See Okinawa Kenmin ni kansuru Seron Ch
osa [Surveys about Okinawans’ feelings]. URL: http://www8.cao.go.jpsurveyindex-all.html(6 May 2012, date last accessed). External threats and voting in Okinawa 97 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from War, Okinawans switched from ‘Support’ – answering (i) or (ii) – or ‘Do not know’ – answering (v) – to ‘Oppose’ – answering (iii) or (iv). Opposition to US bases increased to 60%, whereas support for US bases decreased to 30%. In the post-Cold War period, opposition to US bases decreased to 44.4%, whereas support for US bases increased to 45.7%. This indicates that US bases have historically confronted strong opposition. Note that the share of those who answered (iv) remained at
30% in the 1990s and then dropped to 23.8% in 2001. Therefore, Okinawans understood the utility of US bases for securing Japan whereas they also knew the US bases-related risks. The survey results of Okinawans’ attitudes show that US bases historically confronted strong opposition, but the degree of opposition decreased in the post-Cold War period. We further explore the reasons for this trend by focusing on Japan’s security-related risks. The Cabinet Office of Japan conducted the National Survey on the Japan Self-Defense Force and Security Issues every three years in the period 1972–1984 and post-1988. The Cabinet Office asked the question, ‘Does Japan face any risk of adversaries waging war against Japan or of being dragged into war with another country?’ Japanese respondents could choose one of five alternatives: (i) There is a definite risk, (ii) There is a likely risk, (iii) There is an unlikely risk, (iv) There is no risk, and (v) Do not know. The sample size of each Figure 1 Okinawan’s opinion on US bases in Okinawa 98 Koji Kagotani and Yuki Yanai at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from survey was 3000 respondents except for the year 2000, and the response rate is
73% over time.4 Figure 2 shows the trend of Japanese perceptions of Japan’s security risks. The share of those who acknowledged the existence of a substantial security risk represents people who answered (i) and (ii), whereas the share of those who acknowledged little to no security risk reflects people who answered the (iii) and (iv). In the 1980s, more Japanese respondents acknowledged a heightened security risk as tensions were raised between the Soviet Union and the Unites States and Japan with the onset of the new Cold War. The share of the group perceiving a security risk dropped to <50% in 1994 after which it steadily increased. It rose to 80% in 2003 due to effects of the North Korean missile experiment in 1998 and the global War on Terror. The share of the group acknowledging a security risk has gradually been decreasing since 2003 but has not gone below 65%. Therefore, the survey results show that more Japanese became sensitive to security-related risk in the post-Cold War period. Figure 2 Public opinion on Japan’s security-related risks 4 You can download survey questions and results from the website of Japan’s Cabinet Office. See Jieitai to B
oeimondai ni kansuru Seronch
osa [Surveys on Self Defense Force and Defense Issues]. URL: http://www8.cao.go.jpsurveyindex.html (28 October 2011, date last accessed). The sample size is 5000 respondents for the year 2000. Survey questions and results are available online. External threats and voting in Okinawa 99 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from The survey results above imply that in the post-Cold War period, external threats surrounding Japan had been increasing whereas more Okinawans became supportive for US bases. The increasing level of Japan’s security risks may have raised Okinawans’support for US bases. In the following section, we conduct a regression analysis of Okinawan gubernatorial elections using the municipal level data in order to investigate the influence of Japan’s external threats on local vote choice. We also examine the effect of base policies such as base-related subsidies from Japanese government and state of the base economies brought about by the US occupation on pro-base support. 4 The statistical model The regression model examines whether external threats influence the outcomes of Okinawa gubernatorial elections. We analyze whether perceived security-related risks surrounding Japan increased the vote share of a pro-base candidate in the gubernatorial elections from the period 1972– 2006. The effect of external threats depends on Okinawans’ understanding of security environments. We expect that both emotional and rational reactions to external threats affect the vote share of pro-base candidates. The following model captures our theoretical concerns: Vit ¼ b0 þb1 Rt þb2 Eit þb3 Bit þb4 Di þb5 Pt þb6 Cit þb7 Sit þuit: ð1Þ The unit of analysis in our statistical model is municipality i in gubernatorial election year t. Okinawa consisted of 54 municipalities when it was returned to Japan in 1972. However, many municipalities merged in the period 1972–2006 and we had 41 municipalities at the 2006 election. In order to keep constant the number of municipalities over elections, we use the 41 municipalities that existed at the 2006 election. We treated the municipalities that merged between 1972 and 2006 as if they were already merged by the 1972 election. For these municipalities, we computed the weighted average of constitutive municipalities as merged values for each variable using population as weight. We choose the latest number of municipalities, 41, which is smaller than the original 54, to avoid inflating the number of observations and obtaining incorrectly small standard errors. Thus, we have data for 41 municipalities in each of ten Okinawa gubernatorial elections, and the total number of observations is 410. 100 Koji Kagotani and Yuki Yanai at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from The dependent variable Vit is the vote share of the pro-base candidate in municipality i in yeart. This variable represents Okinawan voters’support for US bases stationed in Okinawa. A pro-base candidate ran for the governor of Okinawa prefecture in every election between 1972 and 2006, and he ran against an anti-base rival in every case barring one.5 Since candidates declared their stances on US bases in every elections, the votes for a pro-base candidate can be interpreted as support for or at least acceptance of the US bases and Japan’s security policy.6 β 0 is the constant term and uit is the stochastic term. The independent variable Rt represents the degree of security-related risks surrounding Japan in year t. To operationalize the level of external threats or foreign challenges against Japan, we use the perceived risks of war in a series of national surveys, which are discussed in Section 3.7 This variable indicates what percent of respondents acknowledged the risk of war in each national survey. The higher value of this variable implies that people believed that Japan confronts more or more potent or both foreign threats. We assume that voters’ evaluations of security-related risks reflects the existence of foreign challenges against Japan and that perceived risks influence vote choice rather than actual risks. Of course, the results of national surveys do not necessarily capture Okinawans’ perception of security-related risks. However, Okinawans are more aware of security issues than people in the mainland of Japan as they live close to US bases and are more directly exposed to the moves of US military troops. The national survey data are likely to understate the amount of awareness, which will make it harder to obtain results supporting the argument that external threats increase pro-base support. Conversely, the data do not overstate Okinawans’risk perception. Thus, this variable can serve as the proxy for Okinawan’s security-related risk perception. Note that the national surveys were not always held in the same year as the Okinawa gubernatorial 5 The only one exception is the 1998 election in which a pro-base candidate competed with two anti-base candidates. However, one of the anti-base candidates was fringe and won the very small number of votes. Thus, we see this election as a two-horse race between pro-base and anti-base candidates. 6 We refer toLanning (2011)for the vote share data. 7 You can download survey questions and results from the website of Japan’s Cabinet Office. See Jieitai to B
oeimondai ni kansuru Seronch
osa [Surveys about Self Defense Force and Defense Issues]. URL: http://www8.cao.go.jpsurveyindex.html (28 October 2011, date last accessed). External threats and voting in Okinawa 101 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from elections. We used linear interpolation to estiamte the values for the elections in which the values are not available.8 In addition, we include several possible control variables in the model. First, we address the size of the local economy Eit as a possible determinant of support for pro-base candidates. Local economies in the region have been growing since the reversion of Okinawa to Japan, with most of the local governments still relying more on fiscal transfers and subsidies from the central government as the main sources of their annual revenue. With this dependence relationship, the central government can make a threat to withdraw or a promise to increase public works projects and financial support in order to enhance its political leverage. Since poor municipalities are particulary vulnerable to such carrot-and-stick policies, they are forced to support the status quo security policy: the continued presence of US bases in Okinawa and the US–Japan security treaty. This tendency becomes more significant when Okinawa experiences economic recessions (Mikanagi, 2004, pp. 104–105). Therefore, the size of the local economy can determine the degree of municipal dependence on the central government and can influence their support for the pro-base policy. To operationalize the size of the local economy, we use the real income tax per capita as a proxy because the municipality-level data of GDP per capita are available since 1982, and there is a strong positive correlation between the size of income tax revenue and the size of the local economy. We expect that richer municipalities have lower support for the pro-base candidate and his policies than poor municipalities. We use the logtransformed value to mitigate the skewness of the distribution.9 Second, we include base dependency Bit in the model to examine Japan’s base-related policy. The central government have been using different compensation schemes to consolidate support for US bases and stationed troops. The fiscal transfers and subsidies are used for improving the welfare of locals. Local governments receive subsidies for setting up 8 The national surveys and the Okinawa gubernatorial elections were not held in the same year except for 1978, 1994, and 2006. That is, we estimated the value of perceived risks for seven out of ten Okinawa gubernatorial elections in the period 1972–2006. Since the first survey was conducted in 1976, the value of 1972 was extrapolated from the values of 1976 and 1978. For other missing values, linear interpolation estimated the values of perceived risks using the values in the national surveys before and after a given election. We discuss the data of perceived risks in the online supplementary material. 9 To compile the data, we used Okinawa Toukei Nenkan [Okinawa Statistical Year Book] by the Statistics Division, Department of Planning, Okinawa Prefecture. 102 Koji Kagotani and Yuki Yanai at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from provisions for preventing military accidents and reducing environmental and noise pollution (Calder, 2007, Chapter 6). Unlike the mainland, where 90% of the land for military reservation is government owned, 60% of land for US bases is owned by local governments and locals. They receive rental payment from the central government (Cooley and Marten, 2006). Also, the US military is the second biggest employer in Okinawa; 9000 locals have been working on the US bases since 2005 (Maedomari, 2011, Chapter 4). Their salaries are covered by the base-related subsidies. Therefore, locals living near military facilities have incentives to support US bases to retain such material benefits. To operationalize base dependency, we use the ratio of the base-related income to the revenue of a municipality i in yeart. Base dependency takes a percentaged value.10 In 2006, only 21 municipalities received base-related subsidies. The five highest base-dependent municipalities are Ginoza-son (35%), Kin-cho (32%), Onna-son (28%), Kadena-cho (17%), and Chatan-cho (16%) in Northern Okinawa, and US bases occupy the majority of their territories. The level of base dependency of all other municipalities is <10% with 20 municipalities receiving no base-related subsidies. We expect that a more base-dependent municipality will provide more support for pro-base candidates. Third, we control the distance between a municipality and a US base Di to capture costs incurred due to the presence of US troops. Military bases cause inconveniences such as environmental and noise pollution and the risk of airplane crashes, which can deteriorate the quality of life for people living close to the bases. Also, it increases the possibilities of soldiers-related crimes such as hit-and-runs, intrusions, rape, and theft. The US–Japan Status-of-Forces does not allow Japan to take suspected soldiers to trial in local courts, which can cause further grievance for locals. Since a shorter distance between local communities and US bases increases contact and interchange between locals and soldiers, locals living near US bases are more subject not only to a physical but also to a psychological burden. Thus, voters living near US bases are less likely to support a pro-base candidate. In Okinawa, there exist two largest US bases in Japan, Futenma, and Kadena bases, which house substantial troops and exert a considerable 10 We use the base-related subsidy data by Lanning (2011) and the municipality-level revenue data compiled by the Okinawa prefecture to calculate the base dependency ratio. External threats and voting in Okinawa 103 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from impact on the lives of locals. On these bases, US military troops engage in takeoff and landing drills round the clock and engage in other disquieting military operations. We choose Marine Corps Air Station Futenma (hereafter referred to as Futenma base) as the most significant US base because it is one of the largest bases in Okinawa and the most controversial base due to crashes of fighter jets and helicopters. When two marines and a navy man kidnapped and raped a twelve-year-old school girl in 1995, Okinawans attempted to reclaim the land occupied by the Futenma base. The rape incident triggered widespread Okinawan’s grievance and politicized the relocation issue of the Futenma base (Funabashi, 1999). Thus, the Futenma base is an appropriate base to control for the physical and psychological burden.11We calculate the distance between each municipal office and the Futenma base using Google Maps and use the logtransformed values in kilometers to mitigate the skewness of the distribution. We expect that municipalities far from the Futenma base will have more support for pro-base candidates.12 Fourth, we control for partisan support for pro-base candidates. The liberal democratic party (LDP) has been taking a pro-base policy stance to maintain the US–Japan security treaty. The LDP maintained the majority of the Lower House seats in the period of 1955–1993, but after 1993, it lost several national elections. If the LDP is more popular at the national level, a pro-base candidate who runs an election expects to receive more support. Such a partisan mood can determine the baseline of support for the pro-base policy. To operationalize partisan support Pt , we use two different measures: the incumbent variable in yeart and the vote share of the LDP in the last national election. The incumbent variable explains whether the incumbent politician is running the Okinawa gubernatorial election. We assume that 11 There are twenty-five bases exclusively used by the United States: two air stations (Futenma and Kadena), three communications bases, one harbor facility, four garrisons, two logistical bases, eleven exercise areas, one medical center, and one welfare (recreation) facility. Among US bases in Okinawa, only Futenma and Kadena are air stations, which are the main sources of the physical costs such as noise and airplane crashes. Other bases are relatively small, and some of them are logistical and communication bases, which are not necessarily dangerous in itself. 12 To check robustness, we run the same regression models with the distance between each municipal office and the other large base, Kadena base. The basic results remain the same. Also, we do not use the dummy variable indicating if a municipality hosts a base because this binary variable captures what the base dependency variable measures. We provide the further explanation of this point in the online supplementary material. 104 Koji Kagotani and Yuki Yanai at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from a pro-base incumbent runs an election when the LDP receives strong national support. We also assume that an anti-base incumbent runs an election when the LDP receives weak national support. Since the incumbent can be either a pro-base or an anti-base politician, we need to control the direction of partisan attributes. Thus, the incumbent variable equals 1 if the pro-base incumbent is running an election, −1 if the anti-base incumbent is running an election, and 0 otherwise. Moreover, we focus on the vote share of the LDP under the proportional representative system in the last national election.13 We use the percentaged value of the LDP vote share as an alternative measure of partisan support. As shown in Figure 3, both measures of partisan support show similar patterns, which validates the assumption regarding the incumbent variable. To avoid multicolinearity, we include these measures separately in the model.14 We expect that these measures will have a positive correlation with the vote share of pro-base candidates. Finally, we also take the industrial structure of the municipal economies into consideration. In the pre-war period, 70% of Okinawans engaged in Figure 3 Incumbent and national support for the LDP 13 A national election can be either the Lower House or the Upper House election. 14 The correlation coefficient of these measures is 0.6. External threats and voting in Okinawa 105 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from agriculture. However, the ratio of employed workforce in agriculture dropped to 5% in the postwar period as US military took over private-owned land and Okinawans lost assets such as fishery boats and equipments for household industry due to World War II. Immediately after Okinawans were released from the prisoner-of-war camps, they had no other choice but to work in the military-related construction industries, restaurants, cleaning shops, clothing stores, and hock shops. The Okinawan economy still relies on the tertiary industry to produce 90% of its gross prefectural product. Employment creation is highly dependent on the base-related construction and services, public works projects, and tourist business (Maedomari, 2011, Chapter 4). Therefore, the municipal industrial structure can form material incentives for locals to support for pro-base candidates. We use the share of employees in the construction and service industries relative to the total employees in each municipality, Cit and Sit, respectively.15 Workers in the construction industry benefit from accepting baserelated construction orders such as maintenance and expansion work. Similarly, most of people in the service industry work for a base or sell products to military personnel and their families. Therefore, employees in these industries are more likely to support pro-base candidates as well. We expect that a municipality with a higher share of construction and service industries will show more support for pro-base candidates. In addition to the model description, we briefly discuss the model speci- fication test and two possible correlation over observations that require us to fix the size of standard errors. We report the random-effects model because a Hausmann specification test found no systematic difference between the municipality-specific random-effects and municipalityspecific fixed-effects models. We also consider the correlation between observations because it generates smaller standard errors. First, municipalities have different industrial and fiscal structures, which may cause the municipality-specific correlation (autocorrelation) over observations. Second, voters confront the same level of Japan’s security-related risks in year t, which may cause the year-specific correlation over observations. Thus, errors uit might have correlations among different election years in 15 To compile this data, we used Jigy
osho Kigy
o Toukei Ch
osa Houkoku [Establishment and Enterprise Census] by Statistics Bureau, Management and Coordinate Agency, Government of Japan. 106 Koji Kagotani and Yuki Yanai at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from the same municipality and among different municipalities in the same year, which make the standard errors smaller. We use the two-dimensional clustering of municipality i and year t to fix the size of standard errors (Cameron et al., 2011).16 To estimate the random-effects model, we run the pooled OLS model (Equation 1) with the two-dimensional clustered standard errors because the pooled OLS estimator is consistent with the random-effects estimator if and only if the error term is uncorrelated with the covariates.17Table 1 shows the summary statistics of the variables used in the random-effects model of the vote share of pro-base candidates in the 1972–2006 Okinawa gubernatorial elections. 5 Results Table 2 shows the estimated results. Model 1 is the benchmark model, and Model 2 includes additional controls for the industrial structure. These models include the incumbent variable in the model to control for partisan support for the LDP. Models 3 and 4 are for replacing the incumbent variable with the vote share of the LDP in the last national election. We focus only on the results of Models 1 and 3 because the industrial structure Table 1 Summary statistics Variable N Mean Std. Dev. Minimum Maximum Pro-base vote share 410 53.56 14.22 7.08 92.08 Japan’s perceived security risk 410 56.20 11.71 43.60 77.60 ln (real tax per capita) 410 3.76 0.62 1.68 6.53 Base Dependency 410 2.97 6.32 0.00 38.00 ln (distance) 410 3.65 1.50 0.88 6.67 Incumbent 410 0.20 0.75 −1.00 1.00 National support for the LDP 410 39.05 5.75 25.17 46.14 Share of construction industry 410 15.88 10.23 0.00 63.48 Share of service industry 410 25.92 12.10 0.00 81.02 16 We estimate the coefficients of the model using the user-package cluster2 on Stata version 11.2. 17 Let ai and 1it denote the municipal-specific random effect and an idiosyncratic error. We can rewrite the error term in equation 1: uit ¼ ai b0 þ1it. This is consistent with the random-effects model. See Cameron and Trivedi (2009, p.254). External threats and voting in Okinawa 107 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from Table 2 Random-effects model of the vote share of pro-base candidates Model 1 (SE) Model 2 (SE) Model 3 (SE) Model 4 (SE) Japan’s perceived security risk 0.37** (0.17) 0.38** (0.16) 0.55** (0.23) 0.56** (0.22) ln (real tax per capita) −2.58 (2.53) −2.99 (2.70) −4.38 (2.83) −4.71 (2.94) Base dependency 0.22* (0.12) 0.22* (0.13) 0.11 (0.14) 0.11 (0.14) ln (distance) 2.45*** (0.95) 2.33** (0.97) 2.26** (0.92) 2.18** (0.94) Incumbent 5.22** (2.36) 5.16** (2.29) National support for the LDP −0.04 (0.13) −0.00 (0.14) Share of construction industry 0.12 (0.12) 0.12 (0.13) Share of service industry 0.07 (0.09) 0.09 (0.11) constant term 31.78*** (9.74) 29.32*** (9.27) 32.21** (14.77) 27.24* (14.46) Adjusted R-squared 0.25 0.25 0.18 0.19 F statistic 31.17 23.58 21.34 16.65 Number of municipalities 41 41 41 41 Number of elections 10 10 10 10 Number of observation 410 410 410 410 *P < 0.1, **P < 0.05, ***P < 0.01. 108 Koji Kagotani and Yuki Yanai at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from variables have no significant effects and the inclusion of those variables does not change the coefficients of the other explanatory variables.18 Model 1 supports the hypotheses about and show significant effects of Japan’s perceived security risk, the distance from the Futenma base, and partisan support. If the national survey shows a one-point increase in Japanese risk perception of war, the vote share of pro-base candidates increases by 0.37 points. This implies that when an additional 10% of the respondents acknowledges Japan’s risk of war, a pro-base candidate receives an additional support of 3.7 points in the gubernatorial election. Since such an increase in risk perception reflects a significant change in Japan’s security environment, we can conclude that the effect of external threats is moderate. Note that the data are likely to understate the amount of Okinawan awareness, which will make it harder to obtain results supporting the rally argument that threat perception increases Okinawan support for US bases. Next, the size of the municipal economies or revenues does not affect the election outcomes. A one-point increase in base dependency causes a 0.22-point increase in the vote share of pro-base candidates, but its effect is marginally significant at the 0.1 level. We cannot strongly claim that base dependency positively correlates with the vote share of pro-base candidates. In contrast, the distance of the municipality from the Futenma base significantly affects the vote share of pro-base candidates. A 1% increase in the distance from the Futenma base generates a 2.45-point increase in the vote share of pro-base candidates. This means that the effect of the distance is non-linear and diminishing. When the distance increases from 5 to 15, its effect changes from 3.94 to 6.63. When the distance increases from 15 to 25, its effect changes from 6.63 to 7.89. When municipalities are relatively near to the Futenma base, even a small difference in the distance causes a huge impact on the vote share of pro-base candidates. The incumbent variable also shows a remarkable effect on the vote share of pro-base candidates. A pro-base incumbent enjoys a 5.22-point advantage in the election, whereas an anti-base incumbent gives a 5.22-point disadvantage to his pro-base rival. Therefore, we confirm the effect of external threats in the Okinawa gubernatorial elections, but we find no evidence that material interests such as the size of municipal economies and base dependency form incentives to raise support for pro-base candidates. 18 We cannot reject the joint hypothesis that coefficients regarding the shares of construction and service industries are equal to zero (p= 0.57 and p= 0.61 for Model 2 and Model 4, respectively). External threats and voting in Okinawa 109 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from Model 3 shows similar results except on two points; the effect of Japan’s perceived security risk is slightly greater and national support for the LDP has no influence on the vote share of pro-base candidates. Our concern then becomes whether Model 1 or 3 is more appropriate for explaining the vote share of pro-base candidates. We examined whether we can reject the joint null hypothesis regarding the partisan support variable of each model. We can reject the joint null hypothesis regarding the incumbent variable in Model 1 (P = 0.03), whereas we cannot do so regarding the national support variable for the LDP in Model 3 (P= 0.77). These results imply that the model with the incumbent variable is more appropriate than the one without the national support variable. Therefore, we can adopt the results of Model 1 and re-confirm that the effect of Japan’s perceived risk is moderate. In summary, our findings clearly reveal Okinawans’ political calculation in the gubernatorial elections. First, we find the effect of external threats in the Okinawa gubernatorial elections to be present, but the effect is more moderate than the effects of the other variables. Changes in Japan’s security risk are exogenous shocks to Okinawans, which is consistent with the fact that Article 9 of the Constitution prohibits Japan from initiating a military conflict and only foreign challenges change Japan’s strategic environments. Second, it is surprising that the vote share of pro-base candidates is not affected by material interests such as the size of the municipal economy and the industrial structures of municipalities. Since these structural differences among municipalities remain almost the same overtime, they might determine the baseline support level but is not helpful for explaining the effect on temporal changes. Third, Japan’s base-related fiscal transfer and subsidies have a marginal effect on the vote share of pro-base candidates, which is rather contradictory to the conventional view. However, if the distribution of these financial compensation remains almost the same over time, base dependency can be helpful for consolidating the baseline support level but not for enhancing pro-base support temporarily. In the post-Cold War period, certain groups of municipalities have been receiving more base-related subsidies than before.19Such a selective incentive will further consolidate pro-base support. 19 In the post Cold War period, the base-related subsidies have been more targeted at the following municipalities: Ginoza-son, Ie-son, Yomitan-son, Kadena-cho, Kin-cho, Nago-shi and Onna-son. 110 Koji Kagotani and Yuki Yanai at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from Fourth, the distance of a municipality from the Futenma base captures physical and mental costs imposed on locals as a significant determinant of Okinawans’ voting behavior. Such costs influence support for pro-base candidates more than material benefits provided by the base-related policies. Finally, the gubernatorial election can be highly influenced by the partisan support or the political mood encouraging the incumbent to run an election. Both the pro-base and anti-base groups can influence the election outcomes by manipulating the base issue. Fifteen out of forty-one municipalities had five or more competitive elections in which the vote share of pro-base candidate was on the margin (between 45 and 55%). The impact of partisan support suggests that political maneuvers can influence policy mood and generate the fluctuating trend of pro-base support over municipalities and time. 6 Conclusion In the postwar period, Okinawans have been engaged in anti-base activities, but anti-base candidates did not always win the gubernatorial elections. To understand the dichotomy between Okinawans’ anti-base sentiments and the pro-base voting record, we investigated the Okinawa gubernatorial elections in the period 1972–2006. Japan’s perceived security risk acts as a driving force behind the evolution of Okinawans’ vote choice because external threats encourage Okinawans to support a pro-base candidate. The empirical analysis provides three important findings. First, the effect of external threats determines the trend of Okinawa gubernatorial election outcomes although the size of the effect is moderate. When Okinawans perceive more security risks facing Japan, patriotic sentiment or the demand for national security encourages them to support a pro-base candidate. Okinawans behave in response to changes in strategic environments surrounding Japan. Second, the distance from the controversial Futenma base determines the risks of physical and mental costs such as airplane crashes, noise pollution, rape incidence, and other crimes, which significantly reduces pro-base support. On the other hand, Japan’s base-related fiscal transfers and subsidies only marginally enhance pro-base support, which is inconsistent with the conventional view. However, we cannot exclude the possibility that material interests form the baseline of pro-base support. In future research, we will investigate the External threats and voting in Okinawa 111 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from impact of the sudden increase in the base-related income in the municipalities with US military facilities after the end of Cold War. Third, partisan support or political mood encouraging the incumbent to run an election can determine the election outcomes, especially in the municipalities where the vote share of the pro-base candidate is on the margin (or between 45 and 55%). This implies that both pro-base and anti-base activists can influence the election results by manipulating the base issue. In short, external threats appear to form Okinawan’s attitudes, whereas baserelated costs and partisan mood determine fluctuations of pro-base support. The results contribute to our understanding of international relations theory and base politics in Okinawa. We examined whether external threats cause the rally effect and whether they determines vote choice. The results confirm that Okinawans’ voting behavior reflects their perception of Japan’s security risks. Our findings also show that, unlike the mixed results of US cases, the Okinawa gubernatorial elections show a clear but moderate connection between foreign policy outcomes and vote choice because Japan’s security policy directly affects Okinawans’ security risks (but US military operations do not always change Americans’ security risks due to the global commitment of US troops). In short, with careful consideration of their security risks, Okinawans are prudent with casting their votes based on material calculations and the partisan mood surrounding the gubernatorial elections. These results extend our understanding of the fluctuations in the support for US overseas bases from the locals of the host countries. The results also provide insightful policy implications for the management of the US–Japan alliance in the post-Cold War period. According to the surveys by the Cabinet Office of Japan, Japanese security risk perception decreased from 60 points in the early 1980s to 48 points in the mid-1990s and then increased to 78 points in the mid-2000s. The pro-base vote share in the gubernatorial elections decreased from 56% in 1986 to 40% in 1994 and then increased to 52% in 2006. The end of the Cold War decreased the demand for securing Japan, given the expectation that the collapse of the Soviet Union would eliminate regional security risks in East Asia. Japan’s security demand then increased due to North Korea’s subsequent missile experiment and nuclear development and China’s increasing territorial ambitions. Okinawans responded to such a change in their security environments by changing their vote choice. They appreciate 112 Koji Kagotani and Yuki Yanai at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from the deterrent effect and utility of the US–Japan security treaty and are willing to face the base-related costs during these times. In 2009, Prime Minister Hatoyama blindly politicized but failed to settle the relocation issue of the Futenma base, which provoked Okinawans’ anger and sparked anti-base movements. This caused obstacles against the management of the US–Japan alliance. Populists should not resort to blindly using the Okinawa issue to raise their popularity without understanding base politics in Okinawa. Supplementary material Supplementary material is available at International Relations of the Asia-Pacific online. Acknowledgements We acknowledge William Lanning for providing his data of the municipallevel voting results in the Okinawa gubernatorial elections and the baserelated subsidies. For helpful comments, we are grateful to Han Dorussen, Kentaro Fukumoto, Erik Gartzke, Benjamin E. Goldsmith, Yusaku Horiuchi, Takeshi Iida, Kosuke Imai, Rene Lindstaedt, Farheen Momen, Toshio Nagahisa, Jun Saito, Arthur A. Stein, and Teppei Yamamoto. References Aldrich, J.H., Gelpi, C., Feaver, P., Reifler, J. and Sharp, K.T. (2006) ‘Foreign policy and the electoral connection’, Annual Review of Political Science, 9, 477–502. Anand, S. and Krosnick, J.A. (2003) ‘The impact of attitudes toward foreign policy goals on public preferences among presidential candidates: A study of issue publics and the attentive public in the 2000 U.S. presidential election’, Presidential Studies Quarterly, 33(1), 31–71. Brody, R.A. (1991) Assessing the President: The Media, Elite Opinion, and Public Support. Stanford, CA: Stanford University Press. Calder, K.E. (2007) Embattled Garrisons: Comparative Base Politics and American Globalism. Princeton, NJ: Princeton University Press. Cameron, A.C. and Trivedi, P.K. (2009) Microeconometrics Using Stata, Revised edn. College Station, TX: Stata Press. External threats and voting in Okinawa 113 at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from Cameron, A.C., Gelbach, J.B. and Miller, D.L. (2011) ‘Robust inference with multiway clustering’, Journal of Business and Economic Statistics, 29(2), 238–249. Cooley, A. and Marten, K. (2006) ‘Base motives: The political economy of Okinawa’s antimilitarism’, Armed Forces and Society, 32(4), 566–583. Edwards, G.C., III and Swenson, T. (1997) ‘Who rallies? The anatomy of a rally event’, Journal of Politics, 59(1), 200–212. Eldridge, R.D. (1997) ‘The 1996 Okinawa referendum on U.S. base reductions: One question, several answers’, Asian Survey, 37(10), 879–904. Funabashi, Y. (1999) Alliance Adrift. Washington, DC: Council on Foreign Relations Press. Gelpi, C., Feaver, P.D. and Reifler, J. (2005/2006) ‘Success matters: Casualty sensitivity and the war in Iraq’,International Security, 30(3), 7–46. Hartley, T. and Russett, B. (1992) ‘Public opinion and the common defense: Who governs military spending in the United States?’, American Political Science Review, 86(4), 905–915. Lai, B. and Reiter, D. (2005) ‘Rally ‘round the union jack? Public opinion and the use of force in the United Kingdom, 1948–20010 , International Studies Quarterly, 49(2), 255–272. Lanning, W.R. (2011) U.S. base dependence in Okinawa: The relationship between U.S. bases and voting in Okinawa. Master’s thesis. Waseda University. Levy, J.S. (1989) ‘The diversionary theory of war: A critique’, in M.I. Midlarsky (ed.), Handbook of War Studies, pp. 259–288. MA: Unwin Hyman. Maedomari, H. (2011) Okinawa to Beigun Kichi [Okinawa and US Bases]. Tokyo: Kadokawa Shoten. Mikanagi, Y. (2004) ‘Okinawa: Women, bases, and US-Japan relations’, International Relations of the Asia-Pacific, 4(1), 97–111. Mueller, J.E. (1970) ‘Presidential popularity from Truman to Johnson’, American Political Science Review, 64(1), 18–34. Nickelsburg, M. and Norpoth, H. (2000)‘Commander-in-chief or chief economist?: The president in the eye of the publics’,Electoral Studies, 19(2-3), 313–332. Nordhaus, W., Oneal, J.R. and Russett, B. (2012) ‘The effects of the international security environment on national military expenditures: A multicountry study’, International Organization, 66(3), 491–513. Norpoth, H. (1991) ‘The popularity of the Thatcher government: A matter of war and economy’, in H. Norpoth, M.S. Lewis-Beck and J.-D. Lafay (eds), Economics and Politics: The Calculus of Support, pp. 141–160. Ann Arbor, MI: University of Michigan Press. Okamura, M. (1991) ‘Estimating the impact of the Soviet Union’s threat on the United States–Japan alliance: A demand system approach’, Review of Economics and Statistics, 73(2), 200–207. 114 Koji Kagotani and Yuki Yanai at Columbia University Health Sciences Library on January 30, 2014 http://irap.oxfordjournals.org/ Downloaded from Parker, S.L. (1995) ‘Toward an understanding of “rally” effects: Public opinion in the Persian gulf war’,Public Opinion Quarterly, 59(4), 526–546. Schultz, K.A. (2001) Democracy and Coercive Diplomacy. Cambridge: Cambridge University Press. Wlezien, C. and Erikson, R.S. (2005) ‘Post-election reflection on our pre-election predictions’,PS: Political Science & Politics, 38(1), 25–26. Yeo, A. (2011) Activists, Alliances, and Anti-U.S. Base Protests. Cambridge: Cambridge University Press