Abstract (要旨)

Motivation and background

The way the social security and tax policies are managed has received broad attention in Japan. In this country, the sustainability of social security financing is heavily dependent on the operation of those policies. In fact, the government has recently carried out several reforms of social security systems, including pension, medical, and log-term care insurances, in order to maintain their financial sustainability. Furthermore, a number of discussions have taken place about the extent to which consumption tax can be utilized to finance social security. The revision of those systems would affect the corporate and household behaviors through a change of the relationship between the social insurance benefits and burdens. Nevertheless, its behavioral impacts have not been fully elucidated in empirical researches. Unfortunately, a mistaken notion instead prevails in the public. For instance, the Nippon Keidanren expressed in their "Outline of Proposal on Pension System Reform for 2004" that "the increased burden of the social insurance contributions reduces the vitality of corporations, impedes economic activities, and discourages the corporate efforts of employment maintenance." In response, the 2004 pension reform law fixed the final rate of the insurance premium at 18.3 percentage point. However, in the economic theory, the increased employer's contribution rate does not necessarily reduce corporate profits. A lack of empirical researches is one of the possible reasons for this gap in understanding of the policy effect between economists and others.

Therefore, this thesis tries to empirically clarify, based on the framework of economics, the interaction of the social security and taxation policies with economic activities. More specifically, it examines, using micro-level data, how firms adjust wages and investments in response to policy changes, and also how household savings react to the institutional shifts. Those topics have been addressed by a number of researchers abroad; however, in Japan, only a few studies have tackled each of them yet. In addition, serious disagreement among scholars has remained unresolved in some fields. This thesis develops empirical analysis and disentangles several undecided issues.

Overview of the thesis

This thesis consists of two parts. Part I examines the behavioral effect of social security policies. This part addresses the following two topics: (1) the incidence of employer contributions to social security, and (2) the effect of pension reform on household asset accumulation. Part II examines the effect of tax policies. This part discusses the following two topics: (1) the reaction of corporate investment to the change in corporate tax rate, and (2) the effect of interest income tax on the household portfolio behavior.

I briefly summarize the abovementioned four topics of research, each of which constitutes one of the four chapters of this thesis. The first issue is the incidence of employer contributions to social security. This issue has remained controversial among empirical studies in Japan because opposite conclusions were obtained in two pioneering works: Tachibanaki and Yokoyama (2008) and Komamura and Yamada (2004). Whereas Tachibanaki and Yokoyama (2008) found "no apparent backward shifting on to employees," Komamura and Yamada (2004) concluded that a "majority of the employer's contribution rate to health insurance was shifting back onto the employees." The first chapter reconciles this confusing situation by identifying a possible bias in the estimation results of their wage equation. First, this chapter reappraises Tachibanaki and Yokoyama's (2008) findings by modifying their empirical strategy. To begin with, I control for a spurious positive correlation between the contribution rate and wages by trend variables. Second, I utilize a cross-sectional variation in the contribution rate of workers' compensation, which was not considered in Tachibanaki and Yokoyama (2008). Third, I exclude the mining and real estate industries in order to remove unusual short-run fluctuations of wages. As a result of those three modifications, the coefficients of the contribution rate are estimated to be significantly negative, suggesting that some of the employers' contribution is shifting back onto employees. Next, I consider the endogeneity problem of the contribution rate in Komamura and Yamada (2004). Our first evidence of this problem is a significant reverse causality from the monthly wage (dependent variable in the wage equation) to the employers' contribution rate (one of the independent variables in the wage equation). Further, I present more evidence of this relation by the regressions of the net wage or total labor cost on the contribution rates and the proportion of the employers' burden. The coefficients of those variables are internally inconsistent unless some endogeneity biases are assumed for the coefficients. Considering this outcome, my estimation result appears to be most consistent with the partial shifting hypothesis. Therefore, Komamura and Yamada's (2004) full shifting result would be overestimated by the endogeneity of wages. If each result of Tachibanaki and Yokoyama (2008) and Komamura and Yamada (2004) was biased in the manner suggested in this chapter, the most likely situation of Japan is the partial backward shifting. This result is also corroborated by previous studies of other countries.

In the second chapter, I analyze the effect of the 1999 pension reform on household asset accumulation in Japan. If the life-cycle hypothesis (LCH) holds true, the changes of pension wealth are offset by household assets. Aso and He (2001), which is one of the few Japanese studies examining this issue, found a significantly positive relationship between net pension benefits and household financial assets in a cross sectional setting. They then suggested that this result might be corroborated with the altruistic bequest motive hypothesis (ABMH). However, their result would reflect a spurious positive correlation between pension and private wealth, which is attributable to the positive relation between the benefits of employees' pension and before-retirement earnings. This chapter tries to avoid this problem by exploiting the exogenous reduction in pension benefits caused by the 1999 reform. This exogenous time-series variation allows me to identify the substitutability between those two assets separately from their inherent positive correlation in the cross section. In order to evaluate this substitution effect, I regress the household wealth on the net pension benefits, with other household characteristics being controlled. As a result of estimation, the substitutability is found mainly in the middle-aged households. This result implies that the saving behavior of those households is corroborated with LCH. In addition, I test whether the magnitude and significance of substitutability differ between altruistic and less altruistic households in a manner that is consistent with LCH. If LCH is more reasonable for less altruistic households as one would expect, it can be ascertained that the estimates of the offset capture the substantial effect of the pension reform. Consequently, this test indicates that the substitution effect is more significant for less altruistic households. Thus, the estimates of substitutability are probably identified by the variation of household assets caused by the 1999 pension reform. Moreover, this additional estimation suggests that the less altruistic households, for example, those with a weak motivation for leaving assets to their descendants and/or with no children, behave more consistently with LCH than other types of household.

The third chapter examines the investment responses to Japanese corporate tax reforms, based on the tax-adjusted form of Tobin's q model. Thus far, many works have faced difficulty in distinguishing the effect of tax on investment from that of other fundamental variables, because a number of variables moved together over the business cycle in aggregated time-series data. To identify the tax effect, this chapter utilizes the cross-sectional variation of the change in tax-adjusted q driven by tax reforms, instead of q itself, in estimating investment functions. As a result, I obtain significantly positive coefficients of q particularly for the tax reforms in the 1980s, suggesting that the corporate tax cut significantly impacted the corporate investment decision. Moreover, for some industries (e.g., the rubber, etc. and wood, etc. industries), the coefficients also indicate the reaction of the investment to the tax cuts in the late 1990s, whereas the firms in other industries do not appear to have responded to them at all. Those results suggest that firms' investment reacted significantly to several Japanese tax reforms in the 1980s and 1990s, though the manner of investment responses somewhat differs among industries. Also, the estimated coefficients on tax-adjusted q are considerably larger than those obtained by conventional estimation methods. This is probably because the measurement error problem is reduced by using the change in tax-adjusted q driven by tax change, which is correctly observable, as an independent variable. These estimates are more valid as a parameter of the adjustment cost function in Tobin's q model as compared with those obtained previously.

Finally, this thesis treats the interest tax effect on the household portfolio behavior. Thus far, a number of empirical studies have also addressed this issue, but the identification of the tax effect has been a difficult problem. In the countries other than Japan, many previous studies tried to avoid a simultaneity problem of the marginal tax rate under aggregate income taxation. Specifically, the marginal tax rate is a function of household income in this system, and therefore, it is affected by portfolio choice itself. Some of the studies replaced the actual marginal tax rate with a hypothetical one to deal with this problem. Even after removing this simultaneity, however, marginal tax rates still appeared to suffer from another endogeneity problem of their direct dependence on taxable income; in other words, marginal tax rates are a non-decreasing function of taxable income. Therefore, in the usual situation that high-income households possess more wealth, the estimated tax effect would reflect a wealth or income effect. In fact, positive tax effects have been broadly reported in the previous foreign studies. In Japanese empirical works, the identification has also been a problem due to a data limitation and the spurious correlation between the tax rate and household wealth. Therefore, this chapter attempts to identify the tax effect on the asset choice and allocation using an exogenous tax variation derived from availability of the small-savings tax-exemption scheme ("Maruyu" system). This variation depends neither on the amounts of household wealth nor income. Contrary to the results of previous studies abroad, I found a significantly negative tax effect on the asset ownership of risky assets, such as bonds and mutual funds, in the probit estimation. Meanwhile, taxation hardly has any effect on the ownership probability of other safety assets, stocks, and life insurance. A likely explanation for this contrast may be a difference in level of the asset return. Since bonds and investment funds yielded higher return rate than safety assets did, the tax exemption led to a larger tax saving for those assets. On the other hand, the asset demand equation, which is estimated by Heckman's two-step method, reveals that taxation did not have any significant effect on the asset allocation. This result appears to suggest that tax advantage was not an important determinant in choosing the portfolio composition.

当論文は,社会保障政策と租税政策が家計と企業の行動に与える影響を実証分析によって明らかにしようとする4本の論文から構成されている。

まず,2章からなる第1部は,社会保障政策を対象にしている。

第1章「On the incidence of social insurance contributions in Japan」は,社会保険の事業主負担がどのように転嫁されるかを分析している。先行研究では結論が割れており,明確な結果が得られていないが,この章では先行研究で見過ごされていた点の改良をおこない,大部分が労働者に帰着するという結果を導いている。ここで新たに考慮されているのは,変数がもつトレンドの適切な処理,雇用保険料の横断面での変化の情報の活用,保険料の内生性の処理等,非常に本質的な課題であり,計量経済学的に正当な処理を丁寧に積み重ねることで説得的な結論を得るに至っている。なお,この章のもととなった研究は,主査である岩本康志との共同研究である。これらの研究での濱秋氏の貢献は,実証研究の部分を全面的に担当した他,含意の解釈に対する議論に対等に貢献し,論文を分担執筆している。

第2章「The effects of the 1999 pension reform on household asset accumulation in Japan: a test of the life-cycle hypothesis」は,公的年金給付の将来の価値が家計の貯蓄残高に与える影響を分析している。家計の貯蓄が利他的遺産動機で形成されているならば,年金給付価値が増えた額だけ貯蓄が増えるという関係が成立する。そのような関係を確認した先行研究が存在するが,現役世代の所得は年金給付と貯蓄の双方に正の影響を与えるため,年金と貯蓄との因果関係が正しく検証されているわけではないという問題を残していた。そこで,この論文では,1999年の年金改正で将来の公的年金給付の減額が起きたことを外生的な政策変更として,因果関係の検証に取り組み,中高年では公的年金資産の増加がそれ以外の金融資産を有意に減少させているという結果を得た。また,利他的動機の強い家計とそうでない家計を対比させ,後者で負の影響がより強く生じていることも示される。両者の結果は,ともにライフサイクル仮説が当てはまるものと解釈できる。

つぎに,2章からなる第2部は,租税政策に関する研究がまとめられている。

第3章「Investment responses to Japanese tax reforms: a cross-industry comparison」は,法人課税が設備投資に与える影響を個別企業のデータをもとに検証したものである。税制の投資への影響の研究は長い歴史をもつが,税制の影響が強く検出されないことが長らくの謎であった。しかし,1990年代に研究の進展があり,税制の変数の内生性の問題から生じる推定値の下方バイアスの問題に対処する推定方法が採用されることにより,法人税が大きな影響をもつことが確認できてきた。この論文では,こうした研究の展開を踏まえ,わが国の80,90年代の税制改革を外生的な影響として識別し,それに対する反応を計測することを試みている。この結果,わが国の研究者による先行研究よりも1桁大きな反応度(理論的予想とより整合的である)を観測している。

第4章「The effects on household portfolio behavior: evidence from the Japanese "Maruyu" system」は,金融所得課税が資産選択に与える影響に関する実証分析である。常識的には,ある資産に対する税率の上昇はその資産保有を減少させると予想させるが,これまでの内外の実証研究はこの関係が明確には検出されていない。それは,制度の性質から高資産者ほど税率が高くなる傾向があり,高資産者が資産選択を多様化させる現象と税率の影響が識別できなかったためであると思われる。この問題点に関して,資産残高をコントロールして税率の違いが反映するような独自の推定手法を工夫することで,税率が債券,投資信託の選択に有意な負の影響をもっていることが検出されている。

これら4論文に共通する研究手法の特徴は,政策の外生性にとくに注意を払って,推定方法に精緻な工夫を施すことにある。経済政策の実験がおこなわれることはまれであるために,政策変数が影響を考える変数に対して外生的であることが,使用するデータでかならずしも保証されないことが多い。そのため,内生性の問題が政策効果の推定にバイアスをもたらす可能性がある。このような問題に対する配慮が十分ではない研究が,残念ながらわが国では散見される。そうした状況のなかで,この論文の研究手法は非常に貴重な存在であり,丁寧な分析によって説得的な結論が得られていることは,経済政策の実証分析への多大な貢献であるといえる。また,いずれの研究もマイクロデータを使用し,複雑な政策変数の構成をおこなうなど,多大な労力を投入した努力も高く評価される。

一方で,選択された研究テーマがやや古いきらいがあり,より先端の研究テーマを追及することで,多くの研究者にインパクトを与える研究ができるのではないか,という意見が審査委員からあった。また,政策効果の結論がかならずしも明確とはならず,不確定な要素を残すものがあり,結果のインパクトが弱いという指摘があった。さらに個別の論点としては。第1章では,変数の内生性の処理について適当な操作変数は得られないか,という課題がある。また,第4章では,使用しているデータの性質から,フローの貯蓄額に対する影響を見ることができない,という指摘があった。しかし,これら個別の論点は,使用しているデータの制約から現状の分析ではただちに解明できず,あらためて別種のデータが利用可能なときに取り組むべき課題であるといえる。

以上のような課題も指摘されたが,それらは,この論文がなした貢献の価値を減じるものではなく,各章の研究は,その関係する研究領域で,確固たる地位を築いたといえる。なお,第1章の一部はJapanese Economic Review誌に掲載予定であり,第3章はJapan and the World Economy誌に掲載されている。他に,第1章のもとになった研究が紀要に1本,書籍に1本掲載されており,当論文がすでに一定の評価を得ていることを物語っている。これらの点を総合的に判断して,審査委員の全会一致で本論文が博士論文にふさわしいとの結論に至った。