Intergenerational Conflict Over Fiscal Consolidation: Theory and Evidence from Japan

Transcription

1 Intergenerational Conflict Over Fiscal Consolidation: Theory and Evidence from Japan Real Arai Ryosuke Okazawa Katsuya Takii Last Revised: February 2, 2018 Abstract This paper analyzes the determinants of voter preferences on fiscal consolidation policies using an opinion survey of Japanese citizens. We address the following questions: (i) Is there a huge intergenerational conflict of interest surrounding fiscal consolidation policies? (ii) Can standard political economic theory explain the observed differences in political preferences among generations? Contrary to the prevailing view, we find that older voters are more likely to support fiscal consolidation policies. We build a simple political economic model and provide several hypotheses that explain the observed fact. We find that financial asset holdings and political knowledge are significantly related with citizen s preferences for fiscal consolidation. Moreover, we find that nearly half of intergenerational conflict surrounding consumption tax policy can be explained by the differences in these factors. We are grateful to Nobuo Akai, Yu Awaya, Shunichiro Bessho, Takero Doi, Makoto Hasegawa, Masayuki Kudamatsu, Kazuki Onji, Shinpei Sano, Kimiko Terai, and the participants of the seminar in Osaka University, Kochi University of Technology, Tokyo Metropolitan University, GRIPS, Kansai Public Economics workshops. Of course, we alone are responsible for both the content and any errors. Department of Management, Kochi University of Technology. Graduate School of Economics, Osaka City University. okazawa@econ.osaka-cu.ac.jp Osaka School of International Public Policy (OSIPP), Osaka University.

2 1 Introduction Since the outbreak of financial crisis, many developed countries, are suffering from enormous government debt. Figure 1 illustrates the recent amounts of general government liabilities per GDP in selected OECD countries. It shows that many countries have experienced government liabilities per GDP after financial crisis. Although fiscal consolidation is one of urgent macroeconomic issues for most developed countries, the processes of fiscal consolidation do not seem to continue well in many countries 1. Why is fiscal consolidation so difficult? [Figure 1 here] The political process is often considered a cause of persistent government deficit. In the classical work, Buchanan and Wagner (1977) argue that fiscal consolidation policies, such as tax increases or cuts in government spending, tend to be postponed in democratic societies because they burden the current electorate although they are beneficial for future generations. Since future generations do not have political influence on current policy decisions, fiscal policy often causes enormous accumulation of government debt. 2 Figure 2 shows that the median age of voters consistently increases in advanced countries in periods when government debt expansion is observed. The changes in the age distribution of the electorate in recent years suggests that the intergenerational conflict might be one explanation for recent government debt expansion. 3 [Figure 2 here] This paper analyzes the determinants of voter preferences for fiscal consolidation policies using an opinion survey conducted among Japanese citizens. We examine how the age of voters influences the preferences for fiscal consolidation policies. Because, as Figure 1 and 2 show, Japan experiences a rapid increase in the median age of voters as well as that in government debt per GDP, Japan is an ideal country for our research purpose. A popular view on a cause of a huge government debt can be summarized by Ihori (2016): As the difference in preference by age increases, the current beneficiary tends to hope for an increase in benefit payments but is reluctant to bear their costs, and the reforms tend to be postponed. As a result, the fiscal burden is imposed on the future beneficiary. If the policy is determined by silver democracy, it causes the accumulation of fiscal deficits and an increase in social security benefits. Using an opinion survey conducted in Japan, this paper examines whether the growing political power of older generations threatens fiscal sustainability. 1 Alesina et al. (2015) analyze the features of fiscal consolidation policies and examines their consequence on the economy in the period There is also substantial literature that analyzes the debt management policies of the government from a positive perspective (e.g., Alesina and Passalacqua (2016)). Intergenerational conflict over fiscal policies is often noted in the literature. 3 Since aging populations typically place a financial burden on the government through an increase in medical spending or pension benefit payments, the role of intergenerational political conflict is reflected only by the observed co-movement of an aging population and the accumulation of government debt. However, it seems natural that politicians in democratic countries should heed the interests of older voters as the median age of the voters increases. 1

3 The first and most important contribution of this paper is to provide robust evidence that the older voter is more likely to support fiscal consolidation policies. As Figure 3 shows, the share of those who support a consumption tax hike monotonically increases with age. A similar pattern is observed in opinions concerning trade-off between fiscal stimulation and consolidation policies. The differences in preferences across generations are statistically significant even if we control for differences in education, income, and family structure. Moreover, we show that it is difficult to explain the observed differences in policy preference by the altruism of older citizens toward future generations. These findings are contrary to the prevailing view that older voters hinder the implementation of fiscal consolidation policies. [Figure 3 here] Our second contribution is to provide several hypotheses that explain the reasons why older voters are likely to support fiscal consolidation policies. We build a simple political economic model where the consumption tax rate and level of government debt are determined by the political process. Our model captures the basic insights of the common view of intergenerational conflict over fiscal consolidation policy. In the benchmark model, the electorate is only interested in how the tax burden is allocated intergenerationally. Since the older voter does not have incentive to approve an increase in the current tax level, the benchmark model cannot explain the observed fact. To explain the fact, we reconsider the underlying assumptions of a benchmark model. We propose three possible explanations. First, we consider the case where the tax rate and the amount of government spending are jointly determined by the political process. In this case, the older citizen has an incentive to increase the tax rate if it increases public good provision. If older citizens receive greater benefit from government spending, older citizens might be more likely to accept a tax increase to finance additional government spending. Second, we consider the possibility that the government can raise a tax on domestic savings returns of older citizens after saving decision is made. Government can implement this policy not only by a default on sovereign debt, but also more implicitly by an inflation tax with an assist of central bank. 4 In this case, the older voter faces a trade-off between a tax on consumption and a tax on current saving. Since this ex post tax on saving implies redistribution from older citizens who hold more domestic assets compared to younger citizens, the older voter might accept a consumption tax increase to prevent heavy taxation on saving. Third, we relaxes the assumption of perfect knowledge by the voter. Although our model assumes that the voter has perfect knowledge concerning the current fiscal situation and their impact on future policies, the voter might have only limited information in reality. If the lack of political and economic knowledge causes citizens to underestimate the problem of fiscal sustainability, citizens knowledge would be correlated with their support for fiscal consolidation policies. According to this argument, the older voter might be more likely to 4 In facts, there are some opinions that the main purpose of a monetary easing in Japan since Abe Cabinet started in 2012 might be monetization. 2

4 support consolidation policies if the older voter has more political concerns and takes the current fiscal situation more seriously than the younger voter. In the latter part of the paper, we test the validity of these hypotheses using policy preference data, asset holdings, and citizens knowledge. We show that the coefficients of ages on the preferences for the rise in consumption tax does not change even after controlling several proxies for the preference for the amount of government spending. On the other hand, we find that the citizen who holds bonds and reads newspapers more frequently is likely to support a consumption tax increase. Interestingly, holding foreign stock does not have any significant effect on the preference on the consumption tax increase. Moreover, when we control the effect of asset holdings and the frequency of reading newspapers, the correlation between age and support for a consumption tax hike becomes weaker. Although we cannot derive a definite conclusion on the channel through which age influences policy preferences from this evidence, older citizens concern for their assets value or their political knowledge are likely to explain the parts of reasons why an old supports fiscal consolidation in Japan. This study is related to several studies that analyze the determinants of fiscal policy preference using survey data. Many studies analyze the determinants of preference on redistributive policy (e.g., Corneo and Gruner 2002, Alesina and La Ferrara 2005, Alesina and Fuchs-Schündeln 2007, Alesina and Giliano 2011, Giuliano and Spilimbergo 2013). 5 On the other hand, a few studies analyze preferences for fiscal consolidation policies. But, an existing few studies suggest that Japan is not only country that old support fiscal consolidation more. Blinder and Krueger (2004) investigate the process of public opinion formation using a telephone survey of a sample of the US population. Although they do not address fiscal consolidation policies directly, the authors show that older respondents are more likely to consider the fiscal deficit a serious problem while they are less likely to consider raising tax an appropriate way to reduce the deficit. Heinemann and Hennighausen (2012) and Hayo and Neumeier (2017) also find that older citizens are more likely to demand public debt reduction using German Survey data. Walter et al. (forthcoming) also find that older voters tend to agree with austerity programs rather than young people using the original survey data in Greece. Although these authors find similar evidence as we find in Japan, they do not analyze the reason why older citizens prefer debt reduction because the intergenerational conflict over consolidation policy is not the main focus of their analysis. 6 This paper goes further ahead. We try to investigate the reason why older citizens prefer consolidation using a simple political economic model and Japanese data. In particular, to 5 While several studies analyze fiscal policy preferences(e.g., Ohtake and Tomioka 2004, Hayo and Ono 2010, Yamamura 2012, 2014) using survey data from Japanese citizens, there is no empirical research that analyzes Japanese citizens preferences for fiscal consolidation. 6 The intergenerational conflict over policy preference is studied more substantially in research that analyzes demand for government spending. Much research examines the differences in opinion between generations over public education spending. The empirical research using survey data typically confirms that the older voter is less likely to support spending on public education, which is consistent with the standard economic theory since government spending on public education is considered redistribution from older to younger generations (e.g., Bruner and Baldson 2004, Cattaneo and Wolter 2009, Rattsøand Sørensen 2010, Sørensen 2013, Bruner and Johnson 2016). 3

5 the best of our knowledge, no previous study analyzes the relationship between financial asset holding and fiscal policy preferences. Our explanation on the reason why older citizens prefer debt reduction can provide an important lesson for other countries because the arguments that the aging electorate causes political bias toward the interests of the older electorate are commonly observed in many advanced countries. In countries where the accumulation of sovereign debt is outstanding, such as Italy and Greece, loose financial administration is often attributed to political rule by the older generation. For instance, in an article in the Financial Times, the common view of the young Italian is described as follows (young people) feel stifled by Silvio Berlusconi s older generation of political leaders who drove Italian public debt to crippling levels and kept the top jobs to themselves for decades. 7 Our results suggest that there is a mechanism that an old rather prefers a reduction of debt to save their assets. Our finding is also related to the recent discussion on political reforms in aging societies. In developed countries where there is a tendency for the political power of older citizens to increase, some people insist on election system reform to ensure equality between generations. Giving suffrage to the youth is a standard reform plan, which has been introduced in some regions and countries in recent years. For instance, the voting rights in the parliamentary election in the United Kingdom was extended to 16 and 17-year-old citizens in In Japan, the voting age has recently been lowered to 18. The Demeny voting system, which gives political influence to children by allowing parents to vote on behalf of their children, is being discussed as an alternative reform plan in several countries such as France, Germany, Hungary, and Japan. 8 Although there are additional aims of such electoral reform plans, the concern for fiscal sustainability is included in the reform plan objectives. However, our result casts doubt on the argument that such electoral reforms will enhance fiscal sustainability. The remainder of the paper is organized as follows. In Section 2, we explain the background of the recent fiscal policy in Japan. In Section 3, we explain the characteristics of our data. Section 4 provides the basic empirical results. In Section 5, we build a simple political economic model for government debt to consider the implications of our findings theoretically. We propose several hypotheses that explain our empirical findings. We test the validity of these hypotheses using data in Section 6. Finally, we provide some discussion and conclusions in Section 7. 7 Italy s Generation X hopes to loosen grip of gerontocracy Financial Times, December 2, In Japan, several other electoral reform plans are being discussed. For instance, Doi and Ihori (1998) propose electoral reform that reallocates representatives seats to each generational group. By assigning a fixed number of seats to each generation, it solves the problem whereby the aging population causes political bias toward the older voter through an increase in the age of the median voter. See also Kato and Kobayashi (2017). 4

6 2 Background In this section, we briefly explain the recent history of consumption tax and public finance in Japan. The amount of public debt in Japan has been ballooning for some decades. While the government began issuing government bonds in 1966, the gross liability of Japan s general government has grown since 1966 and reached over 220% in Faced with accumulating public debt, the sustainability of the fiscal system in Japan has been doubted since the mid 1990s. In response to the severe fiscal situation, the government has attempted to enact fiscal consolidation. In 2001, the government developed the basic plan for economic fiscal policy (Basic Policies for Economic and Fiscal Policy Management and Structural Reform) and set the following policies to restore fiscal sustainability: (i) The government restricts the issue of new government bonds to a maximum of 30 trillion yen. (ii) The government will transform the primary balance of the budget into a surplus. Although the fiscal balance gradually improved in the mid-2000s toward a surplus by the early 2010s, the primary balance again deteriorated due to the financial crisis in After the change in government on September 9, 2009, the administration of the Democratic Party of Japan formulated the Basic Frame for Fiscal Consolidation: Medium-term Fiscal Plan in 2010 and emphasized the aim of improving the primary balance and persistently reducing the amount of government debt to GDP ratio. In the plan, the government aimed to reduce the primary deficit of national and local governments to GDP ratio by half from 2010 to 2015 and bring the primary balance of the budget into surplus by The coalition government composed of the LDP (Liberal Democratic Party) and Komeito decided to take over the target of fiscal consolidations after the change of government on December in However, it will be difficult to achieve a primary surplus for national and local governments by 2020 even if increases in the consumption tax to 10% can be implemented. In 2017, Prime Minister Shinzo Abe recognized that the fiscal situation is such that the goal will be difficult to attain. The comprehensive reform of social security and tax systems have played a key role in the fiscal consolidation plans of the government since The reform proposal in 2011 stated that the consumption tax rate would be raised gradually to 10% by the middle of Consumption taxation is considered an important channel for increasing primary revenues and attaining fiscal consolidation. 11 Additionally, in the Basic Frame of Fiscal Consolidation, the government plans to increase the consumption tax rate taking economic conditions and other factors into account. The consideration of economic conditions allows discretion and make the implementation 9 Data source: OECD Economic Outlook 2017 (June). The consumption tax was introduced in Japan in Then, the tax rate was 3%. In 1998, the consumption tax rate was raised from 3% to 5%. 11 For instance, the IMF (2011) reported that Japan faces a severe situation regarding fiscal sustainability and should raise the consumption tax rate as a main revenue source to avoid a fiscal crisis. Many academic studies also investigate sustainable fiscal policy and show a similar result that implies that Japan must adjust its fiscal policy and fiscal sustainability system (e.g., Ihori et al., 2005; Imrohoroglu and Sudo, 2011; Arai and Ueda, 2013; Hansen and Imrohoroglu, 2016; Imrohoroglu et al; 2016). 5

7 of fiscal consolidation politically difficult. In 2014, the consumption tax rate was raised from 5% to 8% as planned. However, in 2014, a further increase in the consumption tax rate from 8% to 10%, which was scheduled in October 2015, was postponed to April 2017 by the political decision of the prime minister Abe. Moreover, it was postponed again to October 2019 in June 2016 because of economic stagnation and sluggish domestic demand. Political consideration underlies delays in fiscal consolidation. The consumption tax is unpopular with voters and, hence, it is necessary to avoid tax increases to win the election. The opposition to tax increases remains strong at present, and there are concerns of a further delay in fiscal reform. 3 Data In this section, we discuss our investigation of intergenerational differences in attitude toward consolidation policies. We describe our data and the measurement of preferences for the consumption tax rate from the data. We use survey data from the Japanese General Social Survey (JGSS) and opinion poll data from the University of Tokyo/Asahi Shimbum Survey (UTAS). The JGSS is a Japanese version of the General Social Survey and collects data including demographic characteristics such as age, sex, family structure, education, and income. The JGSS also collects opinions, preferences, and data on the sense of values of a wide range of respondents. 12 While the survey has been conducted periodically since 2001, and most of the survey data are available, we use the data from 2010 (JGSS2010) and 2012 (JGSS2012) because these surveys investigate opinions on the desirable consumption tax rate in the context where public discussion on the consumption tax increase attracts much attention at the time when the two surveys were conducted. The government decided on the consumption tax increase in 2013, and the tax rate was increased from 5% to 8% in Our data contain citizens opinions on the desirable level of consumption tax at the time before the government eventually decided on the consumption tax increase. While the JGSS contains cross-sectional data, and respondents are not the same in the 2010 and 2012 surveys, these two surveys use the same method and contain many overlapping question items. Therefore, we pool the data from the two surveys and analyze preferences for fiscal consolidation using the pooled data. As secondary data source, we use data from the University of Tokyo/Asahi Shimbum Survey (UTAS), which has jointly conducted public opinion polls since the 46th Lower House 12 The JGSS is designed and conducted by the JGSS Research Center at Osaka University of Commerce (Joint Usage/Research Center for Japanese General Social Surveys accredited by the Ministry of Education, Culture, Sports, Science and Technology) in collaboration with the Institute of Social Science at the University of Tokyo. The JGSS adopts both interview and detention methods using a questionnaire to collect survey data. The questions on subjects that would relate to social justice or morals are posed using the detention method. The survey population is composed of men and women from 20 years old to 89 years old. The survey subjects are selected using a stratified two-stage sampling method. 13 Although a further tax increase from 8% to 10% was supposed to take place in 2017, the Japanese government decided to delay it until

8 general election in The prominent characteristic of this data are that it includes data on the political preferences of candidates running for the Diet, although we use only the opinion data of citizens. The survey was conducted nationwide and questionnaires were sent to 3,000 voters; responses were received from 1,900 voters, and data on the attitudes toward various policy dimensions including diplomatic issues, political reform, trade policies, and fiscal policies were obtained. The advantage of the UTAS data is that the data reflect respondents opinions on tax policies and on the desirable amount of government spending. On the other hand, we find more demographic variables, such as the income of respondents, in the JGSS. Therefore, we use the UTAS to back up our findings obtained in the JGSS. 3.1 Preferences for the Consumption Tax Rate Let us first discuss how to measure the preferences for fiscal consolidation using the JGSS survey data. We measure preferences for fiscal consolidation using the following question on the consumption tax rate: At what level do you think the consumption tax rate should be? This question was posed to 2,507 respondents in the JGSS 2010 and 4,667 in the JGSS The respondents were asked to choose their answers from six alternatives, which are composed of two categories with the tax rate less than 5 %, one with a tax rate of exactly 5%, and three with tax rates higher than 5%. 16 We focus on whether the respondents choose a tax rate higher than 5%, which is the actual consumption tax rate at the time of each survey, and construct a binary variable that takes a value of 1 if the respondents choose alternatives that indicate tax rates higher than 5% and 0 otherwise. In the following analysis, we regard those who choose tax rates higher than 5% as supporters of the consumption tax increase and empirically analyze the determinants. In the UTAS, citizens are asked for their opinions on tax policies and fiscal stimulants. Concerning fiscal consolidation, the following statement is presented: For the time being, the government should increase spending to stimulate economic activities rather than cut spending for fiscal consolidation. The respondents are required to choose their answers from five alternatives: agree, somewhat agree, cannot say (between agree and disagree), somewhat disagree, and disagree. The respondents are also asked to provide their opinion of the consumption tax increase using the following statement: In the long run, it is inevitable that the consumption tax rate will be higher than 10%. The respondents choose their answers from five alternatives, which are the same as those above. We use each respondent s response to these questions as their preference for fiscal consolidation and the consumption tax increase. 14 The UTAS is conducted by Masaki Taniguchi at the University of Tokyo and the Asahi Shimbun. 15 Approximately half of the respondents in the JGSS 2010 are not asked their opinion on the consumption tax. 16 Although the questions were the same for the JGSS 2010 and 2012, there was a slight difference in their alternatives. While the six alternatives were (1) 0%, (2) 1 to 4%, (3) 5%, (4) 6 to 7%, (5) 8 to 9%, and (6) more than 10% in the JGSS 2010, The alternatives in 2012 were (1) 0%, (2) 1 to 4%, (3) 5%, (4) 6 to 9%, (5) 10 to 14%, and (6) more than 15%. To control any possible influence caused by the differences in the answer categories between the JGSS 2010 and 2012, we include a survey year dummy in the following analysis. 7

9 3.2 Independent Variables Following the literature, we include some control variables such as sex, marital status, number of children, education, employment status, and annual household income in our analysis. These variables are designed to control for economic and demographic characteristics among citizens that might influence their preference for fiscal consolidation. We classify respondents into the following four categories according to their terminal education record: elementary and junior high school, high school, some college, and university. 17 Employment status is classified into four categories in the JGSS data: regular employment (seiki-koyo), part-time employment (hiseiki-koyo), self-employment, and not working. 18 In the UTAS data, we created two different categories, public servant and student, in addition to the four categories. For household income, only JGSS asks for pre-tax household income that includes wages and pension benefits, yields on shares, and income from real estate. The respondents are asked to choose their answers from among several income categories. We record the median value of the category that the respondent chooses as household income. 19 We also include additional controls in some specifications. (1) Whether a respondent lives in a city with a population over 250,000. (2) A subjective evaluation of their health status (five grades). (3) Whether the respondent has their own house 20 (4) Whether the respondent trusts members of Congress 2122 (5) Whether the respondent donated during the past year. (6) Whether the respondent will participate in volunteer activities 23 The last two variables are included in the JGSS 2012, which was conducted in the year following the Great East Japan Earthquake. We use these two variables as a proxy for respondent altruism. [Table 1, 2 here] 17 Both the JGSS and UTAS ask respondents for their terminal education record. The some college category includes colleges of technology and junior colleges. Graduate school is included in the university category. 18 Managers and directors are included in regular employment while part-time and temporary workers are included in non-regular employment. We provide a more rigorous explanation of the variables in the appendix. 19 We assign 350 thousand yen for the respondent who chooses the lowest category, less than 700 thousand yen for the next lowest category, and 25 million yen for the respondent who chooses the highest category of more than 23 million yen. 20 Stadelmann and Eichenberger (2012, 2014) show that public debts capitalize into prices of property. House ownership might affect policy preference through the capitalization effect of public debt. 21 Stix (2013) shows that the credibility of fiscal consolidation matters with regard to attitudes toward fiscal consolidation policies. 22 In the JGSS, there are questions concerning trust in members of Congress as follows: How much do you trust members of Congress? We regard those who respond with trust a lot or trust a little as having trust in politicians, and we examine the impact. Since half of the respondents in the JGSS 2012 are not asked about trust toward members of Congress, the sample size substantially decreases when we include trust in politicians in our analysis. 23 JGSS 2012 asked respondents, Would you like to participate in volunteer activities if there is an opportunity to from now on? We regard those who respond with I d like to participate or I d like to participate if I could as having the will to participate in volunteer activity. 8

10 24 Table 1 and 2 provide the summary statistics for the variables used in our analysis. Table 1 shows that 39% of respondents of the JGSS support a consumption tax hike. Table 3 shows how consumption tax preferences vary among different demographic groups. Approximately half of all respondents choose the status quo for the consumption tax rate. While 46% of those aged in their 60s support a consumption tax increase, only 29% of those aged in their 20s support a consumption tax increase. Few people hope for a reduction in the consumption tax rate, which is natural when we consider the current fiscal situation in Japan. However, there is a tendency among younger respondents to hope for a reduction in consumption taxes compared to older individuals. Panels B and C in Table 3 compare the gender differences. The results show that female respondents are less likely to support a consumption tax increase. The tendency that the fraction of tax hike supporters increases with age is observed both in the male and female sample, although it is less obvious in the latter. These findings contradict the standard argument that the older generation would postpone the implementation of consolidation policies and impose greater fiscal burden on future generations. [Table 3 here] 4 Empirical Results This section shows our empirical results. We first show the results from the JGSS. 4.1 Japanese General Social Survey Following the literature on fiscal policy preferences, we control several demographic variables to isolate the impact of age on support for a consumption tax increase, which cannot be explained by the correlation between age and other demographic variables. We use the following regression model to clarify how tax policy preferences are related to age and other variables, Y it = β 0 + β 1 Age it + β 2 X it + δ t + ϵ it (1) where Y it refers to support for the tax increase. In the benchmark specification, Y it refers to whether the respondent i supports the consumption tax increase, and we estimate (1) using probit regression. 25 The subscript t refers to the survey year. In Model (1), our main concern is whether the coefficient of age, β 1, is positive or negative. The other basic control variables X it include the demographics of the respondent i such as sex, marital status, number of children, education, employment status, and annual household 24 There is no information on marital status and family structure in the UTAS data. 25 Even if we measure preferences for a tax increase more precisely by dividing the respondents into (i) those who choose a tax rate less than 5%, (ii) those who choose a tax rate equal to 5%, and (iii) those who choose more than 5%, our qualitative result does not change. In this case, Y it is the ordered variable that refers to the degree of preference for the consumption tax increase, and we estimate (1) using an ordered probit. The result is available upon request. 9

11 income. We always include the survey year fixed effect δ t, which captures any differences between the survey years. Table 4 provides the estimation results from our basic specification. The explained variable is whether the respondent supports a consumption tax increase, and Table 4 reports the marginal effect on such support. The important finding is that older individuals are significantly more likely to support a consumption tax increase even if we control the demographic variables. In all specifications, the impact of age is significantly positive, and the estimated marginal effect is approximately 0.5 to 6%. Therefore, aging 20 years increases the probability that the respondent will support a consumption tax increase by approximately 10%, indicating that the intergenerational difference in preference for the consumption tax should not be negligible. [Table 4 here] The results in Table 4 do not depend on the assumption of a linear relationship between age and preference for a tax increase. To allow a non-monotone relationship between age and policy preference, we estimate a model that includes the cohort dummies of respondents instead of their age. Table 5 provides the results, which show that there is no clear pattern that strongly denies the monotone relationship between age and preference. The number of supporters of the consumption tax increases consistently with age although individuals in their 60s are more likely to support the tax increase than individuals over 70. Notably, support for the tax increase is significantly large among those in their 60s, which is when Japanese workers typically retire and start to receive pension benefits. [Table 5 here] Table 4 also shows that females are significantly less likely to approve of a consumption tax increase. The large estimated impact indicates that gender is one of the most important factors explaining the differences in attitudes toward a tax increase. Moreover, family structure has an impact on preferences for a consumption tax increase. Compared with single individuals, those who are married are more likely to support a consumption tax increase although the estimated impacts are not significant in all specifications. The existence of children, however, has no significant impact on preferences for a consumption tax increase, which is contrary to the results in the previous literature. This result is robust when we allow the nonmonotone relationship between number of children and policy preference as Table 5 shows. Therefore, there would be no tendency whereby individuals with children are more likely to consider the welfare of future generations and, thus, approve fiscal consolidation policies. The relationship between education and preference for the consumption tax increase is roughly consistent with that of previous research. There is a tendency that more educated individuals are more favorable toward a consumption tax increase. Additionally, household income is a key determinant of preference for a consumption tax increase; rich individuals are significantly more likely to support a consumption tax increase, which might be because of the regressivity of the consumption tax. 10

12 Column (2) in Table 4 shows that the positive relationship between age and support for a consumption tax increase is robust even when we include the additional control variables of residential area, house ownership, and health status. Whether the respondent lives in a city with a population of more than 200,000, house ownership and the subjective health status are not related to support for a consumption tax increase. More importantly, the estimated impact of age barely changes even if we control these variables. Therefore, we cannot attribute the intergenerational differences in preferences for a consumption tax increase to differences in residual area, house ownership, and health status. Distrust in the political process might be an obstacle preventing support of fiscal consolidation policies, as Stix (2013) argues. To test the validity of this argument, we include trust in politicians as an explanatory variable. Column (3) in Table 4 shows that support for a tax increase is not significantly related to trust in politicians, and the intergenerational difference in preferences for a consumption tax increase do not disappear even if we control trust in politicians. The finding that older individuals tend to prefer a consumption tax increase seems to be robust but not consistent with the standard view that older individuals would place fiscal burden on future generations. Is this because older individuals are altruistic toward future generations? If older individuals value the welfare of society including that of future generations, they would prefer fiscal consolidation policies not for their own interest, but for the public s interest. 26 To examine whether such a preference exists and can explain the observed intergenerational differences in preferences for fiscal policies, we run the regression model that includes the proxies of altruism as additional control variables. For the proxies of altruism, we use the experience of donation during the past year and volition to volunteer. Column (4) reports the results. Altruism seems to have a significantly positive impact on support for the consumption tax increase. 27 There is a tendency for individuals who have experience with donating or volunteering to be more likely to approve of a consumption tax increase, which would be beneficial for future generations. However, the marginal effect of age barely changes even if we control the degree of altruism. This is because there is no strong relationship between age and the proxies of altruism. Therefore, altruism cannot explain the observed intergenerational differences in political preferences. While we find that older Japanese individuals tend to support fiscal consolidation policies, is there a general tendency that they are more likely to support policies for future generations? We provide evidence against this claim from different points of view. We focus on the educational policy of the government, which attracts attention when there is active 26 In the model in Section 5, we assume that each agent thinks only of their self-interests. If older individuals focus on the welfare of their offspring as if it were their own, the so-called Ricardian equivalence should hold. That is, any difference in government finance policies does not affect the welfare of citizens given the sequence of public spending. Therefore, in such a case, there would be no intergenerational conflict over fiscal consolidation policies. However, it is difficult to explain our finding based only on the altruism of older individuals because the Ricardian equivalence cannot explain the fact that older individuals are more likely to support a consumption tax increase than younger individuals. 27 The sample size decreases considerably compared with the benchmark specification because the questions regarding donation and volunteering are contained only in the JGSS

13 debate on tax reforms. The Democratic Party, which came to power in 2009, enacted a bill that provides free high school tuition to extend educational opportunity. Since government spending on education is a type of intergenerational transfer from older to younger generations, older individuals care for future generations if there is a tendency that the older generations are more likely to support such educational policy. Fortunately, the JGSS 2010 investigates respondents attitudes toward educational policies as follows: Are you in favor of policies to make the tuition of public or private school free? The respondents are asked to choose an answer from the alternatives, which are agree, somewhat agree, somewhat disagree, disagree, and do not know. Using the JGSS 2010, we estimated the determinants of a positive attitude toward public educational policies using ordered probit estimation. 28 Table 6 provides the results and shows that the coefficient of age is significantly negative. Therefore, older individuals are significantly less likely to agree with the public subsidy for education, which is consistent with previous research (e.g., Brunner and Balsdon 2004, Cattaneo and Wolter 2009, Brunner and Johnson 2016). Contrary to the case in support of the consumption tax increase, the existence of children has a significant impact on support for educational spending; however, marital status does not. While it seems natural that older individuals are less likely to support an educational subsidy, this result casts doubt on the argument that these individuals care significantly for future generations. [Table 6 here] 4.2 Opinion Poll by the University of Tokyo and Asahi Shimbum Survey To confirm the robustness of our findings, we reexamine the relationship between age and preference for fiscal consolidation policies using the UTAS data, which have the advantage of more directly investigating respondents preferences for fiscal consolidation policies. When we analyze the UTAS data, the dependent variable is the attitude toward fiscal stimulant policies or a consumption tax increase. These are ordered variables, and a higher number corresponds to stronger support for fiscal consolidation or the consumption tax increase. Therefore, we estimate using ordered probit models. [Table 7 here] Table 7 shows that older individuals are significantly likely to prefer fiscal consolidation to stimulate policies and to approve of the consumption tax increase in the long run compared to younger individuals. More precisely, those in their 60s are most likely to support fiscal consolidation policies. While females tend to disagree with a tax increase, which is consistent 28 We define the attitude toward subsidy for education by the ordered variable, which takes 4 if the respondent chooses agree as the answer to the above question, 3 if they choose somewhat agree, 2 if they choose somewhat disagree, and 1 if they choose disagree. We drop the sample of respondents who choose do not know. 12

14 with the result in the JGSS, they are also more likely to support fiscal stimulants. Therefore, it would not be true that females do not support fiscal consolidation policies, but they would prefer spending cuts rather than a tax increase. Additionally, we observe the tendency that more educated individuals support fiscal consolidation policies. Moreover, publicly employed voters and students are likely to approve of a consumption tax increase. Overall, we find no evidence that refutes our findings using the JGSS. 5 Why Do Older Generations Support a Tax Increase? In the previous section, we find that older voters tend to support a consumption tax increase. This finding is not consistent with the prevailing view on intergenerational conflict with fiscal consolidation policy. In this section, we build a political economic model for consumption tax and consider the theoretical implications of our findings in the previous section. 5.1 The Benchmark Model There are some theoretical studies in the literature that analyze the politics of intergenerational conflict over government debt policy (Tabellini 1991, Song et al. 2012). Following Song et al. (2012), we provide a simple political economic model that captures the intergenerational conflict over consumption tax policy. We consider a small open overlapping generations economy. In every period, one unit of individual is born and lives for two periods. In each period, there is a young individual and an old individual. We assume that there is no production. Each individual is given a certain unit of endowment to allocate to consumption and saving. The government provides public good revenue from a consumption tax or by issuing government debt. The lifetime utility of an individual born in period t is given by U t = log c y t + θ log g t + β(log c y t+1 + θλ log g t+1). (2) where c y t is private consumption in youth, c o t+1 is consumption in old age, and g t and g t+1 are public good provision in period t and t+1, respectively. The parameter β is the discount factor, and θ represents the degree of an individual s preference for public good. Following Song et al. (2012), we assume that there is a difference in preferences for public good between the old and the young. The relative strength of old s preference for public good is captured by parameter λ > 0. For instance, if λ > 1 holds, the older individual receives higher marginal utility from public good than the young. Th individual faces the following budget constraints, (1 + τ t )c y t + s t = e, (3) (1 + τ t+1 )c o t+1 = Rs t. (4) Equation (3) is the budget constraint when the individual is young where τ t is the consumption tax rate, s t is private saving, and e is the endowment that the individual receives 13

15 at birth. The saving is invested in an open financial market with a constant return, R. Therefore, the consumption in old age is given by (4). Given a fiscal policy sequence {g s, τ s } s=t, each individual chooses the amount of private consumption {c y t, c o t+1} and saving s t to maximize their utility, equation (2), subject to their budget constraints (3) and (4). Solving the utility maximization problem, we obtain c y e t = (1 + β)(1 + τ t ), (5) c o t+1 = β er, 1 + β 1 + τ t+1 (6) s t = β e. 1 + β (7) We assume that fiscal policy is determined through the political process in each period. In other words, we suppose that the government cannot commit to future fiscal policies. Hence the policy in period t must be determined in period t. Therefore, current citizens cannot choose the sequence of future policies directly. However, the current fiscal policy has influence on future fiscal policy because the current policy affects the future fiscal situation. For instance, the accumulation of government debt approved by current citizens would place an enormous burden on the government budget in the future, which would affect fiscal policies in the future. Since future policy depends on the state of the economy at that time, the current citizens can influence future policy indirectly through the impact of the current policy on the state of the economy in the future. 29 Fiscal policy must satisfy the government budget constraint, Rb t + g t = τ t (c y t + c o t ) + b t+1, (8) where b t is the amount of public debt at the beginning of period t. Moreover, we assume that the government cannot throw away the repayment of public debt. That is, we do not consider sovereign default. 30 This implies that the government can issue public debt only up to the amount that equals the sum of the present value of future primary surplus. Therefore, the government faces the natural debt limit b t+1 b, (9) where b is the maximum amount of public debt. We analyze the intergenerational conflict in fiscal policy by comparing the policy preferences of the young and the old agent. We assume that the public good provision of the government is exogenously given in the benchmark model. By fixing the spending size of public finance, we clarify the difference in preference toward the allocation of fiscal burden among generations. Given the amount of public good provision g t = g, the consumption tax 29 We assume that the citizens can choose fiscal policies depending on the state of the economy but not on the entire past history. This is a standard assumption in the dynamic political economic model that analyzes strategic interactions across generations, such as Hassler et al. (2003, 2005, 2007) and Song et al. (2012). 30 In the following section, we explicitly introduce wealth taxation to our model. The wealth tax can be interpreted as partial default of public debt or inflation tax. 14

16 rate, τ t, and the amount of public debt at the beginning of period t + 1, b t+1, are determined in period t by the political process. The policy that each agent prefers must maximize the following indirect utility function V (a) 31 max τ 0,b b s.t. b = Rb + g V (a) = log (1 + τ) β(1 a) log (1 + T (b )), (10) τ (1 + βr) 1 + τ 1 + β. (11) where a {0, 1} indicates the age of the agent. It holds that a = 0 if the agent is young, whereas it holds that a = 1 if the agent is old. We drop the time subscript t in (10) and (11) because our model has a recursive structure. The variables τ and b are the consumption tax rate and the amount of public debt in the current period, whereas b is the amount of public debt at the beginning of the next period. The function T (b ) represents the consumption tax rate in the next period, which depends on the amount of public debt at the beginning of the next period. As (10) shows, the welfare of the young agent depends on the consumption tax rate in the next period, T (b ). 32 Given the amount of the government debt b, the agent chooses the current fiscal policy, (τ, b ), taking their influence on the consumption tax rate in the next period into account. We assume that the function T (b ) is differentiable. Then, we can derive the change in welfare caused by the marginal increase of the current consumption tax rate as follows dv (a) dτ = τ + 1 (1 + βr)e (1 + τ) β β(1 a) dt 1 + T (b ) db. (12) The first term on the right side of the equation (12) is the utility loss due to a decrease in the current consumption by an increase in the consumption tax rate in the current period. As (5) and (6) show, the current consumption necessarily decreases when the consumption tax rate increases, which is the marginal cost of the consumption tax increase for the agent. Moreover, the change in the tax rate in the current period affects the amount of public debt at the beginning of the next period through the government budget constraint and, thereby, influences the consumption tax rate at the next period. The second term on the right side of the equation (12) captures the welfare effect of this channel. Equation (12) shows the intergenerational differences in policy preference. Since the utility of the old agent (a = 1) does not depend on the consumption tax rate at the next period, an increase in the consumption tax rate at the current period always decreases the 31 In this case, the upper bound of public debt b can be represented as b = Q g R 1, where Q = (1+βR)e 1+β. 32 Since we assume that the citizens can choose fiscal policies depending on the state of the economy, the consumption tax rate in the next period depends on the amount of public debt at the head of the next period. While we take function T (b ) as exogenously given, we can derive it by analyzing the Markov perfect political equilibrium. We provide the policy function in political equilibrium in the appendix. 15

17 utility of the old agent: dv (1) dτ = τ < 0 (13) Therefore, the old agent would like the consumption tax rate to be as low as possible. On the other hand, there would be a benefit to the consumption tax increase for the young agent (a = 0) if it holds that dt db > 0. This is because the government can reduce the current issuance of public debt if it increases the current consumption tax rate. If the decrease in the government debt yields a decrease in the consumption tax rate in the next period, the young agent can increase consumption in the next period. Therefore, if less issuance of government debt in the current period implies a lower consumption tax rate in the next period, the young individual can receive benefit from the fiscal consolidation policy. In such circumstances, it necessarily holds that dv (0) dτ > dv (1) dτ. Hence, the older individual would not prefer a higher consumption tax rate than the younger individual if dt db > 0. Proposition 1. Suppose that the amount of public good provision is exogeneously given, g t = g. Moreover, we assume that the consumption tax rate in the next period increases the amount of government debt at the beginning of the next period, dt db > 0. Then, older citizens do not prefer a higher consumption tax rate than the young. Proposition 1 captures the basic insights of the prevailing argument for the intergenerational conflict over fiscal consolidation policy. If the political interest is only focused on how to share the cost of public good provision among generations, there is no reason for older citizens to support an increase in the consumption tax rate. On the other hand, there is some benefit to a consumption tax increase for younger citizens if the future consumption tax rate increases as government debt accumulates. The assumption that the future consumption tax rate is increasing in the future amount of public debt, dt db > 0, seems natural because the large accumulation of government debt eventually makes a tax increase unavoidable. 33 Therefore, the standard theory predicts that the older citizen generally prefers a lower consumption tax rate than the younger citizen. While we provide a simple political economic model, and its prediction is consistent with the prevailing argument, our findings in section 4 do not support such a prediction. Therefore, we must consider why the benchmark model cannot explain the empirical findings. 34 In the following, we examine how the role of three assumptions underlies the benchmark model, that is, exogeneous public good provision, the perfect commitment of the government to the repayment of public debt, and perfect knowledge of agents. By reconsidering these assumptions, we propose some hypotheses that can potentially explain our empirical findings. 33 In the appendix, we derive the policy function T (b ) endogenously through the voting process. We show that the policy function T (b ) is monotonically increasing in b in a politico-economic equilibrium. 34 If we introduce the altruism of the old agent into the model, intergenerational differences in policy preferences would be reduced, but we cannot explain the fact that the older individual does prefer a higher consumption tax rate than the younger individual by the altruism of the older agent. 16

18 5.2 Preference for Government Spending In the benchmark model, given the size of government spending, each citizen chooses how much consumption tax they pay or issue a government bond to cover the cost of government spending. In such a setting, the attitude toward a consumption tax hike is entirely determined by how individuals think about intertemporal allocation of the fiscal burden. However, there would also be intergenerational differences in preferences for the extent of government spending. Particularly if the older citizen tends to prefer a larger amount of public goods provision than the younger citizen, it might hold that the older citizen prefers a higher consumption tax rate to finance substantial government spending. To consider this possibility, we assume that the consumption tax rate, the amount of government debt, and the amount of public good provision are determined by the political process. Since the utility of each agent depends on the amount of public good provision, we must rewrite the individuals indirect utility function. The policy that each agent prefers is the one that maximizes the following indirect utility V (a) where b = max V (a) = log (1 + τ) + θ[aλ + (1 a)] log g τ 0,b b,g 0 Q R 1.35 β(1 a) log [(1 + T (b )) + λθ log G(b )], s.t. b = Rb + g τ 1 + τ (14) (1 + βr)e. (15) 1 + β The second term on the right side of (14) is the welfare gain of the public good provision. Note that the marginal benefit of the public goods provision is not same between young and old citizens. 36 The old citizen obtains more utility from public goods provision than the young citizen if λ > 1. The functions T (b ) and G(b ) are the consumption tax rate and the amount of public good provision at the next period. As in the benchmark model, we assume that the fiscal policies in the next period depend on the amount of government debt at the beginning of the next period. First, we consider the optimal policy for the older citizens. Since the older citizen is not concerned with the policy at the next period, there is no cost of public debt issuance for the older citizen. Therefore, the older citizen obviously prefers b = b. Given the budget constraints of the government b = b = Rb+g τ (1+βR)e 1+τ 1+β, the older citizen chooses a consumption tax rate and amount of public good provision to maximize V (1) = log (1 + τ)+θλ log g. The optimal consumption tax rate for the old citizen is given by { } θλq ( b Rb) τ = max Q + ( b Rb), 0 (16) The equation (16) implies that if the older generation has a strong preference for public goods provision, they prefer a positive consumption tax rate. Moreover the consumption tax 35 Since the government can choose g = 0, the natural debt limit is not the same as that in the benchmark model. 36 V (a) The marginal benefit of public good provision at the current period is given by = θ[aλ+(1 a)]/g, g which depends on the age of individuals as long as λ 1. 17

19 rate that the old citizen prefers increases in the current levels of government debt. 37 Meanwhile, younger citizens would take the influence of the accumulation of government debt on fiscal policy in the next period into account. Let us suppose that both policy functions T (b ) and G(b ) are differentiable, and the consumption tax rate in the next period is increasing in the debt level, T (b ) > 0, and public good provision is decreasing in the debt level, G (b ) < 0. Under this assumption, the marginal increase in government debt db reduces the utility of young citizens by [ V (0) β dt b = 1 + T (b ) db βλθ ] dg G(b ) db through change in fiscal policy in the next period, which is the marginal cost of issuing the government debt. On the other hand, from the budget constraint (15), the government can reduce the consumption tax rate by (1 + τ) 2 1+β (1+βR)e by issuing one unit of government debt. The marginal benefit for young citizens of reducing one unit of the consumption tax rate is given by V (0) = 1 τ 1 + τ Therefore, if we assume the interior solution, the condition for intertemporal optimization for young citizens is given by 1 + β (1 + τ) (1 + βr)e = β dt 1 + T (b ) db βλθ dg G(b ) db (17) The left side of equation (17) represents the marginal benefit of additional bond issuing, and the right side is the marginal cost. Equation (17) implies that if the fiscal policy in the next period responds to debt accumulation more strongly, the young citizen prefers a higher consumption tax rate. In such a case, issuing the additional debt changes fiscal policy in the next period in an unfavorable direction for the younger citizen, hence, the young prefer fiscal consolidation. The fiscal policy that younger citizens prefer must satisfy the condition (17), budget constraint (15), and the following optimality condition. (1 + τ)g θ = (1 + βr)e 1 + β (18) Equation (18) implies that the marginal rate of substitution between consumption tax and public good provision must equal the marginal rate of transformation between the tax and public good provision. While we cannot derive optimal policy for the young citizen explicitly without specifying the form of policy function T (b ) and G(b ), the following proposition holds if we assume dt/db > 0 and dg/db < If condition θλq > b Rb is not satisfied, older individuals prefer a zero-consumption tax rate. The situation implies that there are enough available resources for the government to finance public good provisions. Thus, in these circumstances, older individuals do not need to impose consumption tax. 38 We show that the assumption dt/db > 0 and dg/db < 0 holds in a political economic equilibrium where the policy function is endogenously determined. See appendix. 18

20 Proposition 2. If the value of public good for the older citizen (θλ) is sufficiently large, then the older citizen prefers the positive consumption tax rate. Moreover, it would be the case that the older citizen prefers a higher consumption tax rate than the younger citizen if λ is sufficiently large, and fiscal policy in the next period is sufficiently inelastic with respect to the level of government debt. 5.3 Taxation on Savings Returns In the benchmark model, we assume that the government can perfectly commit to repayment of the government debt,and it must cover the cost of the public good provision with a consumption tax or the public debt. In this situation, we show that older citizens necessarily prefer to issue debt over a tax hike since an increase in the consumption tax rate directly reduces the welfare of the current generation. If there is an alternative way to raise government revenue, how would the attitude toward fiscal consolidation policy change? Here, we consider the case where the government places additional tax on savings returns as well as consumption. We consider the tax on savings returns as simply property taxation. Alternatively, it could be considered a reduction in the real value of the debt, which is equivalent to tax on the debt holder. In some cases, it would be difficult to achieve fiscal reform by cutting government spending and/or raising tax in a spite of a large amount of public debt because these policies are generally unpopular with electorates, as Buchanan and Warner (1977) argue. In this case, the fiscal imbalance may be solved by reducing the real burden of public debt through partial sovereign default or, more moderately, by inflation. In either interpretation, if there is a possibility that the government may levy a tax on savings, older citizens have an incentive to approve a consumption tax hike to avoid taxation on savings because they typically hold more financial assets, such as government bonds, than younger citizens. 39. We consider this argument by allowing taxation on saving in the benchmark model. In the benchmark model, we do not make an explicit distinction between domestic assets and foreign assets because we assume an open financial market and the return of all assets is given by R. While the real return of assets would be equalized by the arbitrage condition, we clearly distinguish between these assets to analyze the relationship between asset holding and preference on consolidation policy. We assume that there are two assets, foreign assets and the domestic government debt. Let R be the real return of foreign assets, which is exogeneously given and time-invariant. The government cannot levy tax on the foreign asset. The agent can also purchase the government debt which provides one unit of consumption good in the next period. Let p denote the price of debt. Then, the nominal return of government debt is given by 1/p, but the real net return of government debt is (1 ρ )/p since the government can levy tax on the return of the government debt in the next period 39 The lifecycle property illustrates that old people tend to hold more financial assets, including government bonds, than young people. According to the National Survey of Family Income and Expenditure, individuals aged 60 to 69 years old hold 33.9% of the total financial assets while individuals aged 30 to 39 years old hold only 6.3% of the total financial assets in Japan. 19

21 where ρ is the tax rate on the return of government debt in the next period. 40 Due to the log utility, the total savings amount is independent of the asset return and given by β 1+β e. Each agent chooses the total amount of savings and to hold the government debt or foreign asset as a saving method. Since we assume the open asset market, the price of debt p at the current period must satisfy the following arbitrary condition R = (1 ρ ) p (19) shows that the price of debt at the current period depends on the expectation for taxation on assets in the next period. The price of debt decreases if there is an increase in tax on the return in the next period. The arbitrary condition implies that there is no ex ante difference in real returns between foreign assets and government debt. Whatever the expected policy in the next period, the expected net return of the debt would be equal to R through the adjustment of the price of the debt. However, this does not guarantee that the return of these assets will be the same ex post, as we argue below. Using the extended model, we analyze how asset holding affects policy preference of citizens. First, we consider the optimal policy for older citizens. Since older citizens chose their asset portfolio in the previous period, we take their holding of government debt and foreign assets as given. Suppose that the older citizen has allocated a fraction ξ of their savings to the purchase of foreign assets and the remaining fraction 1 ξ to the holding of government debt. In this case, they receive β ξer amount of return from foreign assets. On β 1+β 1+β (1 ξ)(1 ρ)e p the other hand, they also receive amount of net return from government debt where p is the price of government debt in the previous period. 41 Then, the consumption of the older citizen is given by c o = 1 [ βe ξr + (1 ξ) 1 ρ ] 1 + τ 1 + β p Therefore, the optimal policy for older citizens must maximize the following function [ V (1) = log(1 + τ) + log ξr + (1 ξ) 1 ρ ] p (20) implies that an increase in the asset tax rate reduces the welfare of the older citizen who holds government debt. Although any expected changes in asset tax in the previous period has no impact on the real net return of debt because it has been reflected in the price of debt in the previous period, a change in the asset tax rate in the current period does have an impact on the return of the government debt. 42 The government budget constraint is replaced by [ 1 + β κr + β(1 κ) 1 ρ p (19) (20) ] + pb. (21) (1 ρ)rb + g = τ e 1 + τ 1 + β 40 If we regard asset taxation as inflation tax, we can express the tax rate ρ by 1 ρ = 1 where π is the 1+π inflation rate of the economy. 41 Since the saving allocation is sunk at the stage of old age, they cannot change their asset holdings depending on the asset taxation. 42 If there is a difference between the expected tax rate and the actual tax rate, then the net return of the government debt does not equal that of the foreign asset. 20

22 where κ is the aggregate share of the savings converted to purchase the foreign asset. There is no natural debt limit for public debt since the government can reduce the real burden of debt ex post. 43 As in the benchmark model, the older citizen wishes to issue the debt as much as possible since their payoff does not directly depend on the policies in the next period. However, accumulation of government debt b might increase the taxation on assets in the next period and thereby reduce the current price of government debt. In this case, an increase in government debt does not necessarily increase government revenues. Let T (b ) and P (b ) be the consumption tax rate and asset tax rate in the next period, respectively, which depend on the level of government debt at the beginning of the next period. Naturally, we assume that both T (b ) and P (b ) are increasing in b. By substituting the arbitrage condition (19) into (21), we obtain (1 ρ)rb + g = τ [ e 1 + β 1 ρ ] + [1 P (b )]b (22) 1 + τ 1 + β p R where we assume that the foreign assets holding is negligibly small κ 0 for simplicity. The last term on the right side of the government budget constraint (22) is the revenue from issuing government debt. Since the older citizen wishes to maximize revenue and not the amount of government debt, they obviously choose b = b max b [1 P (b )]b. Then, the optimal policy for older citizens must maximize (20) under the budget constraint (22) and b = b. Given constraint (22) and b = b, older citizens face a trade-off between the burden of consumption tax and the burden of asset tax. Since older citizens who hold a substantial amount of domestic government bonds would like to reduce asset taxation, we have the following. Proposition 3. Older citizens who hold more domestic government bonds prefer a higher consumption tax rate than older citizens who hold fewer domestic government bonds. Next, we consider the policy preference of younger citizens. Even if we assume that there are two assets, the consumption of younger citizens in the current period and the expected consumption in the next period are given by (5) and (6), which is independent of fiscal policy. This is because the expected net return of the asset is always equal to R by the arbitrary condition. Therefore, the indirect utility of the younger citizen, which does not depend on the asset tax rate in the next period P (b ), is given by V (0) = log (1 + τ) β log (1 + T (b )) (23) The optimal policy for younger citizens must maximize (23) under the government budget constraint (22). Since there is no welfare cost of increasing asset taxation in the current period for younger citizens, the younger citizen typically prefers asset taxation to consumption tax, On the other hand, there is incentive for the younger citizen to reduce issuing of government debt because it would decrease the consumption tax rate in the next period. If 43 If the government can choose the tax rate ρ [0, 1] freely, for any amount of public debt the government can always repay the outstanding debt through an increase in asset taxation. 21

23 the latter effect is small, and the older citizen holds large government debt, there would be a case where the older citizen would prefer a larger consumption tax rate than the younger citizen. Proposition 4. The older citizen would prefer a higher consumption tax rate than the younger citizen if the government debt at the current period b is sufficiently large, and fiscal policy in the next period is sufficiently inelastic with respect to the level of government debt. Proof. See Appendix. 5.4 Perfect Knowledge In the previous sections, we assumed that the citizen had perfect knowledge of the government s fiscal situation. In the model, we suppose that each citizen chooses policy given the government s budget constraint and policy function, which determines the relationship between the government debt and policy in the next period. In reality, however, many citizens may not recognize the government budget or the influence of the accumulation of government debt on policies in the future correctly since the government budget structure is complex. Particularly, if the citizens underestimate the size of the government deficit or their influence on future policies, they may prefer a lower consumption tax rate simply because they hold an optimistic view. There is some evidence that implies that citizens knowledge of the economic and political situation affects their preference on fiscal consolidation policy. Hayo and Neumeirer (2017) more directly show that citizens economic knowledge on subjects, such as the size of the government deficit and current interest rates, affects their attitude toward consolidation. The respondent who has better knowledge of government debt is more likely to dislike the accumulation of the debt. Boeri and Tabellini (2012) examine how information affects support for pension system reform in Italy with a controlled experiment. The authors randomly selected a subsample of individuals who read a brief descriptive note on the Italian pension system and found that citizens who read the note were more likely to support pension reform. As a general implication, the authors argue that a lack of information among voters would have a systematic impact on their policy preferences. Gavazza and Lizzer (2009) build a dynamic model of electoral competition where voters have imperfect information on aggregate government spending and revenues. The authors show that the decision of voters depends on their information on the size of aggregate policy because the size of aggregate policy influences future fiscal policy through government debt. Particularly, when voters observe only noisy signals of aggregate government spending, the myopic policy that imposes a burden on the future would be approved by the majority of voters. The authors show that the size of a government deficit decreases with the precision of the signal Milesi-Ferretti (2004), Shi and Svensson (2006), and Alt and Lassen (2006) find that higher budgetary transparency reduces the incentive to accumulate debt. 22

24 The important point for our study is whether we can explain the intergenerational differences in fiscal consolidation policy preferences by the extent of citizens knowledge. Even if citizens information on the fiscal situation and current politics matters in their opinion toward consolidation, it cannot explain the intergenerational differences in consolidation policy preferences unless there is significant difference in knowledge across generations. If voter information is an important determinant of policy preferences and, moreover, there is a strong tendency for older citizens to have greater economic and political knowledge than younger citizens, the older citizen is more likely to support fiscal consolidation because they can evaluate the cost of myopic policy more precisely. 6 Empirical Analysis of the Mechanisms In the previous section, we stated three theoretical hypotheses to explain why older generations are more likely to support a consumption tax hike. Although all three explanations are theoretically possible, their implications for the actual politics of fiscal policies significantly differ. If older individuals prefer a consumption tax hike because they demand a large amount of the public good, increasing the political power of older individuals should prevent fiscal consolidation. However, if older individuals support a consumption tax hike because they dislike the devaluation of government debt, or if they have more information about fiscal conditions, then the opposite should be true. In these cases, increasing the political power of older individuals is desirable, at least to consolidate the fiscal deficit. 6.1 Preference for Government Spending Proposition 2 shows that the elderly prefer a higher consumption tax rate when the elderly s demand for public goods is much higher than that of the young. Intuitively, if older individuals receive substantial benefit from public goods or government services then they would support a consumption tax hike to finance the high levels of government spending. We examine whether this is the case in two steps. First, we inspect whether the elderly have a significant demand for government spending in the first place. Second, we test the claim that citizens with a high demand for public goods or services tend to support consumption tax hikes. To measure the individual evaluations of public goods, we exploit several questions on the attitudes toward the role of government for social security and health care for the elderly. The JGSS contains several questions on the government s responsibility to provide life security or health care for the elderly. The question is, Is the life security of the elderly the responsibility of individuals or families? Or is it the responsibility of the state or local government? Similarly, the survey includes the following question on health care, Is health care for the elderly the responsibility of individuals or families? Or is it the responsibility of the state or local government? The respondents provided responses from 1 to 5, where 5 corresponds to responsibility of the state or local government and 1 corresponds to responsibility of individuals or families. These questions do not directly ask about the demand for public 23

25 goods and services that the governments provide. However, if older individuals place greater weight on government provision of goods or services in their utility, there would be a tendency for older individuals to allocate responsibility to the government for policies that particularly benefit the elderly. Here, we use citizens attitudes toward the opinion that the government should ensure welfare or provide health care services for the elderly as a proxy of demand for government spending. Additionally, we also use the question on the general attitude toward redistribution policies since income transfer accounts for a large part of government spending. The JGSS asks for opinions on redistributive policies in the following statement, Are you in favor of the statement that the government should adopt policies to reduce the income gap between the rich and the poor?. We examine the type of citizens who are in favor of high government spending using these questions. [Table 8 here] Table 8 shows the determinants of welfare policy and redistribution preferences. Although there is a significant tendency for the elderly to support redistributive policies, we cannot say that older individuals are more likely to prefer substantial government spending. Rather, older individuals are less likely to think that the government should ensure the welfare of the elderly. 45 [Table 9 here] Table 9 shows that our finding that age is positively correlated with support for a consumption tax hike does not change even if we control for attitude toward welfare policies and redistribution. Columns (1) to (3) show that those who think that the government should be responsible for welfare policies or income redistribution are less likely to support a consumption tax hike. Therefore, there is no evidence that the elderly demand high government spending. Additionally, the demand for welfare policies or redistribution does not positively correlate with support for a consumption tax hike. Consequently, the differences in preferences for public policies cannot explain the positive relationship between age and support for a consumption tax hike. Column (4) shows the results when we include political affiliation as an additional control variable. The results show that the Democratic Party of Japan (DPJ) supporters are significantly more likely to support a consumption tax hike, whereas supporters of the Japanese Communist Party (JCP) are significantly less likely to support a consumption tax hike. 46 The coefficient of age is almost unchanged even if we control for political affiliation. These results imply that it is difficult to explain why older individuals are likely to support a consumption tax hike considering the intergenerational differences in preference on other political dimensions Household income and education are significantly negatively correlated with preferences for welfare policies and redistribution. This is consistent with previous research examining the determinants of preferences for redistributive policies (e.g., Alesina and La Ferrara 2005, Corneo and Gruner 2002). 46 This is not surprising because the DPJ was the ruling party during the survey period and pushed through a consumption tax hike. 47 The relatively positive attitude of the elderly toward a tax hike might come from their fear of the collapse 24

26 6.2 Financial Asset Holdings and Information In Section 5.3, we consider the possibility that the real value of debt would be devaluated by fiscal consolidation. In this case, older individuals would support a consumption tax hike if they have a large amount of domestic assets. If this explanation is plausible, there should be some relationship between financial asset holdings and political preference for a consumption tax. Fortunately, there is a question on financial asset holdings in JGSS The respondents were asked whether they hold bonds, stocks, an investment trust, and foreign deposits. Using this information, we examine whether asset holdings are related to consumption tax preferences. If the concerns for the value of savings creates support for a consumption tax hike, holding domestic assets is positively related with approval for a consumption tax hike. [Table 10 here] Columns (1)to(4) in Table 10 show the determinants of financial asset holdings. Naturally, the share of those who hold each type of asset significantly increases with age with the exception of foreign deposits. [Table 11 here] Columns (2) to (3) in Table 11 report the empirical results when we include financial asset holdings as an additional explanatory variable. The sample size significantly decreases since only half of the respondents in JGSS 2012 were asked about financial asset holdings. Columns (2) and (3) show that bond holding is significantly related to the approval of a consumption tax hike. The probability that a citizen who holds bonds supports a consumption tax hike is approximately 10% higher than for citizens who do not hold bonds. The magnitude of the coefficient of financial asset holdings is large compared to that of other variables. Interestingly, foreign deposit holdings do not have a significant relationship with consumption tax policy preferences. Another hypothesis that explains the intergenerational differences in consumption tax preferences is that the elderly are likely to support a consumption tax hike because they have more information on the current fiscal situation. According to this hypothesis, the young are more optimistic about the government s current financial situation and more likely to disapprove of a consumption tax hike because they do not recognize the need for fiscal consolidation policies. Unlike previous studies such as Blinder and Krueger (2004) and Hayo and Neumeirer (2017), we have no data that directly measure respondents knowledge. As proxies of information held by each respondent on the current fiscal situation, we use the frequency with of the pension system. If older individuals expect that a large general government deficit will reduce pension benefits in the near future, they would approve of a tax hike to enhance the sustainability of the pension system. The JGSS asks participants for their expectations concerning pension benefits from those who have not yet received any pension benefits. Although the sample is restricted to respondents less than 65 years of age, we do not find a significant relationship between a negative view of the future pension system and opinions on a consumption tax hike. 25

27 which each citizen reads newspapers and the political concerns of respondents. Both the 2010 and 2012 JGSS contain a question on the frequency of reading newspapers. The respondents of JGSS are asked how frequently they read newspapers. They choose their answer from five alternatives, almost every day, several times a week, about once a week, less than once a week,, and never. For the political concerns of respondents, the respondents are asked their opinion of the statement I am personally interested in politics. The respondents were asked to choose an answer from the alternatives, which are agree, somewhat agree, neither agree nor disagree, somewhat disagree, and disagree. We use these data as a proxy of respondents political knowledge. 48 Columns (5) and (6) in Table 10 show the determinants of the frequency of reading newspapers and political interest, respectively. These illustrate the clear tendency that both the frequency of reading newspapers and political interest increases with age. While only approximately 20% of respondents in their 20s read a newspaper every day, more than 80 % of citizen over the age of 60 do so. Column (4) in Table 11 shows that the frequency of newspaper reading and political interest are positively correlated with support for a consumption tax hike and that these relationships are statistically significant. Moreover, the estimated impact of age on support for a consumption tax hike slightly decreases when we control these factors. Finally, column (5) in Table 11 provides the estimation results when we control both financial asset holdings and the measure of political knowledge simultaneously. When we control the intergenerational differences in financial asset holdings and political knowledge, the estimated impact of age on support for a consumption tax hike decreases by nearly half although it remains significant. Therefore, at least part of intergenerational conflict around consumption tax policy can be explained by the differences in financial asset holdings and political knowledge. 7 Discussion and Conclusion In the previous section, we show that bond holding and the measure of political knowledge are positively related with support for a consumption tax hike. Moreover, we find that the correlation between age and support for a consumption tax hike weakens when we control these variables. Therefore, intergenerational differences in financial asset holdings and political knowledge can explain at least part of the positive correlation between age and support for a tax hike. While previous studies have found that political knowledge affects preferences on consol- 48 We define the frequency of reading newspapers by the ordered variable, which takes 5 if the respondent chooses almost every day as the answer to the above question, 4 if the respondent chooses several times a week, 3 if the respondent chooses about once a week, 2 if the respondent chooses less than once a week, and 1 if the respondent chooses never. Similarly, we define the political concern by the ordered variable, which takes 5 if the respondent chooses agree as the answer to the above question, 4 if the respondent chooses somewhat agree, 3 if the respondent chooses neither agree nor disagree, 2 if the respondent chooses somewhat disagree, and 1 if the respondent chooses disagree. 26

28 idation policy (Blinder and Krueger 2004, Hayo and 2017), the finding that financial asset holdings are significantly related to preferences on fiscal policy is novel and worth discussing. In this section, after we argue some limitations of our analysis, we discuss the validity of the argument that the concern for asset value of debt holders contributes support for consolidation. There are some points that must be considered when interpreting the results in the previous section. First, although it is theoretically important to distinguish between domestic bonds and foreign assets, we cannot distinguish domestic assets from foreign assets appropriately because the bond category in the JGSS includes both government and municipal bonds, corporate bonds, and even foreign bonds. Similarly, the stock category includes foreign stocks. Moreover, the JGSS asks respondents whether they hold each type of asset but not the value of these assets. Therefore, we cannot precisely measure the extent of the motive to preserve the value of the assets. Additionally, even if bond holdings are positively correlated with support for a consumption tax hike, this might not be caused by the citizen s desire to preserve the real value of assets. There may simply be a tendency for wealthy persons to support tax increases for some other reasons. However, there is no significant relationship between foreign deposit holdings and support for a consumption tax hike. If wealth increases the support for tax increases, a positive correlation should exist between wealth and tax increases. While deriving a conclusion requires further investigation and more sophisticated estimation strategies, there is a positive correlation between bondholding and support for consumption tax. Additionally, our evidence does not reject concerns for the risk of a reduction in asset values as one possible explanation for the observed relationship between age and support for a consumption tax hike. 49 Although there is insufficient support, the hypothesis that the motive to preserve the value of assets affects preferences for fiscal consolidation policy provides new insights into the argument on the political economy of fiscal consolidation. In the theoretical literature of the political economy of debt, debt holdings affect incentives to vote for the repayment of public debt (Tabellini 1991, Dixit and Londregan 2000, Guembel and Sussman 2009). According to this argument, older citizens are more likely to support consolidation policy if the older citizens have more financial assets. Thus far, insufficient attention has been given to this theory and the argument concerning the politics of fiscal consolidation policy. However, we consider that it is not irrelevant to the politics concerning consolidation policies. In the Greek crisis, some citizens voted for austerity because they feared that their assets would be devalued. Vicky Pryce, the chief economic adviser at the Centre for Economics and Business Research, explains the reason why she supported the yes vote as follows: It [no vote] would almost certainly mean banks becoming insolvent, an exit from the euro and 49 There is little research on the relationship between asset holdings and political attitude. Jha (2015) exploits novel data on economic endowments of members of parliament during the English Civil War and shows that overseas shareholding of the elite influenced their decision to support the reform. Using demutualizations as a source of exogenous variation in the stockholdings, Kaustia et al. (2016) show that ownership of stocks has a positive impact on right-of-center voting. 27

29 a much faster decline in economic activity with hyperinflation following as the drachma that is introduced instantly devalues. 50 There is also an article that expresses a similar view to that in this section in EUROPP, an academic blog by the London School of Economics and Political Science. It isn t certain whether giving more voting power to younger people would lead to more sustainable policies. Thus in Europe, electorates even in the crisis countries have, broadly speaking, proven relatively conservative and particularly attached to the common currency. Older voters appear terrified at the thought of losing their pensions and savings to a devalued currency. Young people are traditionally more open to European cosmopolitanism and express relatively higher attachment to the euro in polls. 51 If there is considerable risk of devaluation of domestic assets when standard consolidation policy fails, the prevailing view of intergenerational conflict over fiscal policy would not hold. Finally, we briefly discuss the direction of future research. In our analysis, over half of the correlation between age and support for a consumption tax still requires explanation. Although this might be because we have only limited information on asset holding, there could be another mechanism by which age affects attitude toward fiscal consolidation. Empirically, we need richer data to provide definitive evidence concerning the relationship between financial asset holdings and preferences on fiscal policy. The empirical research that identifies the causal effect of financial asset holdings on fiscal policy preference is left for future study. 50 Greek referendum: how would top economists vote? The Guardian, July 3, Craig J. Willy, In order to avoid a demographic death trap, Western Europe must implement new and fair policies for both present and future generations EUROPP 28

30 References Alesina, A., and Fuchs-Schündeln, N. (2007). Good-Bye Lenin (or Not?): The Effect of Communism on People s Preferences. American Economic Review, 97(4), Alesina, A., and Giuliano, P. (2011). Preferences for Redistribution. In Handbook of Social Economics, Vol. 1, pp Alesina, A., Barbiero, O., Favero, C., Giavazzi, F., and Paradisi, M. (2015). Austerity in Economic Policy, 30(83), Alesina, A., and Passalacqua, A. (2016). The Political Economy of Government Debt. In Handbook of Macroeconomics, Vol. 2, pp Alt, J. E., and Lassen, D. D. (2006). Fiscal transparency, political parties, and debt in OECD countries. European Economic Review, 50(6), Arai, R., and Ueda, J. (2013). A numerical evaluation of the sustainable size of the primary deficit in Japan. Journal of the Japanese and International Economies, 30, Barro, R. J. (1979). On the determination of the public debt. Journal of Political Economy, 87(5), Blinder, A. S., and Krueger, A. B. (2004). What Does the Public Know about Economic Policy, and How Does It Know It? Brookings Papers on Economic Activity, (1), 327. Boeri, T., and Tabellini, G. (2012). Does information increase political support for pension reform? Public Choice, 150(1-2), Brunner, E. J., and Balsdon, E. (2004). Intergenerational conflict and the political economy of school spending. Journal of Urban Economics, 56(2), Brunner, E. J., and Johnson, E. B. (2016). Intergenerational conflict and the political economy of higher education funding. Journal of Urban Economics, 91, Buchanan, J. M., and Wagner, R. E. (1977). Democracy in Deficit. Acad. Press. Campanella (2016) Cattaneo, M. A., and Wolter, S. C. (2009). Are the elderly a threat to educational expenditures? European Journal of Political Economy, 25(2), Corneo, G., and Gruner, H. P. (2002). Individual preferences for political redistribution. Journal of Public Economics, 83(1), Dixit, A., and Londregan, J. (2000). Political power and the credibility of government debt. Journal of Economic Theory, 94(1), Doi, T. and Ihori, T. (1998) Economic Analysis of Japanese Politics, (Nihon Seizi no Keizai Bunseki), Bokutakusya (in Japanese). 29

31 Giuliano, P., and Spilimbergo, A. (2013). Growing up in a Recession. Review of Economic Studies, 81(2), Guembel, A., and Sussman, O. (2009). Sovereign debt without default penalties. The Review of Economic Studies, 76(4), Hayo, B., and Ono, H. (2010). Comparing public attitudes toward providing for the livelihood of the elderly in two aging societies: Germany and Japan. The Journal of Socio-Economics, 39(1), Hayo, B., and Neumeier, F. (2017). Public Attitudes toward Fiscal Consolidation: Evidence from a Representative German Population Survey. Kyklos, 70(1), Hansen, G., and Imrohoroglu, S. (2016). Fiscal Reform and Government Debt in Japan: A Neoclassical Perspective, Review of Economic Dynamics, 21, Heinemann, F., and Hennighausen, T. (2012). Understanding public debt preferences. FinanzArchiv: Public Finance Analysis, 68(4), Ihori et al., 2005 Ihori (2016), Imrohoroglu, S., Kitao, S., and Yamada, T. (2016). Achieving Fiscal Balance In Japan, International Economic Review, 57, Imrohoroglu and Sudo, N. (2011). Will a Growth Miracle Reduce Debt in Japan? Economic Review, 62(1), Jha, S. (2015). Financial asset holdings and political attitudes: evidence from revolutionary England. The Quarterly Journal of Economics, 130(3), Kato, S. and Kobayashi, K. (2017) Public Finance and Democracy, (Zaisei to Minsyu-syugi), Nihonkeizaishinbunsyuppansya (in Japanese). Kaustia, M., Knupfer, S., and Torstila, S. (2016). Stock Ownership and Political Behavior: Evidence from Demutualizations. Management Science, 62(4), Lucas, R. E., and Stokey, N. L. (1983). Optimal fiscal and monetary policy in an economy without capital. Journal of monetary Economics, 12(1), Milesi-Ferretti, G. M. (2004). Good, bad or ugly? On the effects of fiscal rules with creative accounting. Journal of Public Economics, 88(1), Ohtake, F., and Tomioka, J. (2004). Who supports redistribution?. The Japanese Economic Review, 55(4), Rattsø,J. and Sørensen, R.J. (2010). Grey power and public budgets: Family altruism helps children, but not the elderly. European Journal of Political Economy, 26(2),

32 Reinhart, C. M., and Rogoff, K. S. (2010). Growth in a Time of Debt. American Economic Review, 100(2), Shi, M.,and Svensson, J. (2006). Political budget cycles: Do they differ across countries and why?. Journal of public economics, 90(8), Song, Z., Storesletten, K., and Zilibotti, F. (2012). Rotten Parents and Disciplined Children: A Politico-Economic Theory of Public Expenditure and Debt. Econometrica, 80(6), Sørensen, R. J. (2013). Does aging affect preferences for welfare spending? A study of peoples spending preferences in 22 countries, European Journal of Political Economy, 29(1), Stadelmann, D., and Eichenberger, R. (2012). Consequences of Debt Capitalization: Property Ownership and Debt versus Tax Choice. Southern Economic Journal, 78(3), Stadelmann, D., and Eichenberger, R. (2014). Public debts capitalize into property prices: empirical evidence for a new perspective on debt incidence. International Tax and Public Finance, 21(3), Stix, H. (2013). Does the Broad Public Want to Consolidate Public Debt? - The Role of Fairness and of Policy Credibility. Kyklos, 66(1), Tabellini, G. (1991). The Politics of Intergenerational Redistribution. Journal of Political Economy, 99(2), Walter, S., Dinas, E., Jurado, I., and Konstantinidis, N. (forthcomin). Non-cooperation by popular vote. expectations, foreign intervention and the vote in the 2015 Greek bailout referendum. forthcoming in International Organization. Yamamura, E. (2012). Social capital, household income, and preferences for income redistribution. European Journal of Political Economy, 28(4), Yamamura, E. (2014). Trust in government and its effect on preferences for income redistribution and perceived tax burden. Economics of Governance, 15(1),

33 Table 1: Summary statistics :JGSS2010,2012 Variable Mean Std. Dev. obs. support for tax hike age female married number of children high school some college university income regular employment part-time employment self-employed live in a large city house ownership subjective health status (five grades) trust in politicians experience of charity will to volunteer

34 Table 2: Summary statistics : UTAS Variable Mean Std. Dev. obs. prefcon prefctaxlong female age 30s age 40s age 50s age 60s age 70 over high school some college university regular employment part time employment self-employed public servant student

35 Table 3: Preferece on consumption tax by age group :JGSS2010,2012 tax cut status quo tax increase Total age group obs. fraction obs fraction obs fraction obs A. all sample Total B. male only Total C. female only Total

36 Table 4: Estimation results of benchmark model: JGSS2010, 2012 (1) (2) (3) (4) age (0.0006) (0.0006) (0.0008) (0.0012) female (0.0157) (0.0158) (0.0212) (0.0318) married (0.0194) (0.0195) (0.0268) (0.0390) have children (0.0224) (0.0225) (0.0304) (0.0453) high school (0.0225) (0.0227) (0.0316) (0.0474) some college (0.0293) (0.0295) (0.0402) (0.0584) university (0.0262) (0.0265) (0.0357) (0.0534) income (0.0021) (0.0021) (0.0028) (0.0043) regular employment (0.0209) (0.0210) (0.0277) (0.0417) part time employment (0.0210) (0.0211) (0.0293) (0.0434) self-employed (0.0243) (0.0244) (0.0323) (0.0479) live in a large city (0.0142) (0.0193) (0.0289) house ownership (0.0185) (0.0251) (0.0373) health status (0.0202) (0.0268) (0.0403) trust in politician (0.0203) (0.0313) experience of charity (0.0440) will to volunteer (0.0324) N pseudo R Marginal effects; Robust standard errors in parentheses p < 0.1, p < 0.05, p <

37 Table 5: Nonlinear specidication about age and number of children: JGSS2010, 2012 (1) (2) age 30s (0.0346) (0.0347) age 40s (0.0345) (0.0347) age 50s (0.0355) (0.0357) age 60s (0.0349) (0.0353) age 70 over (0.0374) (0.0378) female (0.0158) (0.0158) married (0.0196) (0.0196) havechildren (0.0223) one child (0.0266) two children (0.0237) three and more children (0.0264) N pseudo R Marginal effects; Robust standard errors in parentheses. The education, employment status, and household income of respondents are also controlled. p < 0.1, p < 0.05, p <

38 Table 6: Determinants of preference on educational policy :JGSS2010 (1) (2) age (0.0023) (0.0024) female (0.0633) (0.0637) married (0.0783) (0.0783) have children (0.0832) (0.0838) high school (0.0934) (0.0937) some college (0.1123) (0.1125) university (0.1056) (0.1068) income (0.0086) (0.0088) regular employed (0.0774) (0.0776) part-time employed (0.0826) (0.0831) self-employed (0.1061) (0.1066) live in a large city (0.0561) house ownership (0.0719) health status (0.0839) N pseudo R Ordered probit regression. Robust standard errors in parentheses p < 0.1, p < 0.05, p <

39 Table 7: Estimation results of benchmark model: UTAS (1) (2) Fiscal Consolidation 10% Consumption Tax age 30s (0.1151) (0.1145) age 40s (0.1114) (0.1111) age 50s (0.1096) (0.1091) age 60s (0.1149) (0.1141) age 70 over (0.1199) (0.1190) female (0.0569) (0.0562) high school (0.0809) (0.0798) some college (0.0947) (0.0939) university (0.0931) (0.0920) regular employed (0.0770) (0.0764) part-time employed (0.0843) (0.0849) self-employed (0.0844) (0.0823) public servant (0.1279) (0.1270) student (0.2246) (0.2240) N pseudo R Ordered probit regression. Robust standard errors in parentheses. p < 0.1, p < 0.05, p <

40 Table 8: Determinants of preference on government spendings (1) (2) (3) Life Security for the Old Health Care for the Old Redisribution age (0.0012) (0.0013) (0.0013) female (0.0340) (0.0345) (0.0343) married (0.0425) (0.0436) (0.0430) have children (0.0472) (0.0482) (0.0486) high school (0.0515) (0.0524) (0.0489) some college (0.0619) (0.0632) (0.0588) university (0.0591) (0.0595) (0.0582) income (0.0045) (0.0046) (0.0048) regular employed (0.0443) (0.0444) (0.0448) part-time employed (0.0439) (0.0452) (0.0439) self-employed (0.0550) (0.0564) (0.0554) N pseudo R Ordered probit regression. Robust standard errors in parentheses. p < 0.1, p < 0.05, p <

41 Table 9: Control Preference on Government Spendins (1) (2) (3) (4) age (0.0006) (0.0006) (0.0006) (0.0006) life security for the old (0.0061) (0.0082) health care for the old (0.0069) (0.0092) redistribution (0.0069) (0.0072) LDP support (0.0200) DPJ support (0.0228) JCP support (0.0491) N pseudo R Marginal effects; Robust standard errors in parentheses We control sex, marital status, family structure, employment status and household income. p < 0.1, p < 0.05, p <

42 Table 10: Determinants of asset holdings and political concern (1) (2) (3) (4) (5) (6) Bond Stock Investmen Foreign Reading Political Trust Deposit Newspaper Interest age (0.0041) (0.0037) (0.0035) (0.0041) (0.0028) (0.0022) female (0.1077) (0.1026) (0.0968) (0.1205) (0.0736) (0.0599) marry (0.1292) (0.1291) (0.1230) (0.1537) (0.0857) (0.0728) havechildren (0.1537) (0.1380) (0.1373) (0.1651) (0.0935) (0.0806) highschool (0.1528) (0.1740) (0.1643) (0.2404) (0.1088) (0.0917) techcollege (0.1875) (0.1992) (0.1898) (0.2681) (0.1251) (0.1066) university (0.1750) (0.1826) (0.1802) (0.2538) (0.1264) (0.1045) income (0.0103) (0.0105) (0.0102) (0.0119) (0.0115) (0.0076) regular (0.1478) (0.1318) (0.1246) (0.1534) (0.0908) (0.0807) parttime (0.1484) (0.1534) (0.1296) (0.1612) (0.0908) (0.0779) selfemployed (0.1664) (0.1523) (0.1542) (0.1907) (0.1283) (0.0982) N pseudo R Robust standard errors in parentheses p < 0.1, p < 0.05, p <

43 Table 11: Control Asset Holdings and Frequency of Reading Newspaper (1) (2) (3) (4) (5) age (0.0010) (0.0010) (0.0010) (0.0011) (0.0011) bond (0.0445) (0.0464) (0.0465) stock (0.0436) (0.0437) investmen ttrust (0.0433) (0.0435) foreign deposit (0.0563) (0.0575) frequency of reading newspaper (0.0128) (0.0129) interest in politics (0.0101) (0.0102) N pseudo R Marginal effects; Robust standard errors in parentheses We control sex, marital status, family structure, employment status and household income. p < 0.1, p < 0.05, p <

44 Figure 1 43