4 Comments on “Public Debt and Economic Growth in an Aging Japan”

Keimeir Kaizuka, and Anne Krueger
Published Date:
July 2006
  • ShareShare
Show Summary Details

Comments by Robert Dekle

The chapter examines the effects of Japan’s population aging and government debt policy on its country’s growth rate and economic welfare, particularly on its future tax and social security burdens. The chapter is distinguished in its careful application of the Auerbach-Kotlikoff computational overlapping generations model to simulate the Japanese economy from 2003 to 2050.

The authors reach the following conclusions and policy recommendations. First, in the baseline case (when the surplus in the primary balance is achieved by 2010), the ratio of the sum of taxes and social security contributions to GDP (the national burden ratio) will rise from the current 30 percent to 60 percent by 2050. The ratio of outstanding government debt to GDP will rise from the current 114 percent to 175 percent by 2050.

Second, the sooner the surplus in the primary balance is achieved, the lower the national burden ratio after 2030 and the smaller the long-run debt/GDP ratio. Younger generations (those turning 20) in 2020 unambiguously prefer that the surplus be achieved sooner, while older generations prefer that the surplus be achieved later (in 2022). This is because, as the resolution of the debt is postponed, more of the taxation falls on the younger generations.

Third, the 2004 public pension reforms were a success. According to the authors, the main feature of the 2004 reforms was to keep future total public pension contributions constant, while decreasing future total benefits, in order to balance the public pension account. The authors show that the 2004 reforms will sharply lower social security benefits starting from 2015. The reforms will enhance the welfare of the younger generations (those turning 20 after 2020), while hurting the generations that are older.

Fourth, the burden of rising taxes on the younger generations will be ameliorated somewhat if consumption taxes are used to pay for the rising social welfare expenditures. An expected increase in future consumption taxes will raise the price of future goods, resulting in higher private saving. This higher saving will increase future capital stocks and GDP growth rates, making younger generations somewhat better off than with other forms of taxation.

I generally agree with these policy recommendations. In the simulation model of Dekle (2005, chapter 4), this shows that given the high current debt/GDP ratios, and the rapid aging of the population, the national burden ratio needs to reach about 50 percent by 2040 for the government to meet its (intertemporal) budget constraint. The ratio of government debt to GDP is projected to rise to a peak of 155 percent by 2020. Fiscal reform—cuts in social security and medical care benefits and, particularly, cuts in public investment—are necessary to keep tax rates and debt/GDP ratios at lower levels.

As the authors acknowledge, the dynamics of the debt/GDP ratio in their chapter and in all others depend crucially on the difference in the gap between the real interest rate and the real GDP growth rate, r - g. The larger the gap, the more rapid the accumulation of debt relative to GDP. The authors assume that, going forward, Japanese technological progress is zero. Thus, with negative population (labor supply) growth, output growth is negative, resulting in a large differential r - g, and a very rapid growth of the debt/GDP ratio. More optimistic projections of Japanese GDP growth, such as in Broda and Weinstein (2004), will result in a less rapid increase of the debt/GDP ratio—and less need for drastic fiscal reform.

While the authors cite empirical studies, showing that Japanese technological progress (total factor productivity growth) over the last two decades was essentially zero, the authors may be too pessimistic about Japanese future technological progress. For one thing, Japan has undertaken much structural reform in recent years, especially in non-tradeable sectors such as telecommunications, retail, and distribution. These reforms should result in higher productivity, as they have in other countries. For another, labor supply growth may actually turn out to be positive, particularly if greater numbers of women and the elderly participate in the labor market, and more immigrants are allowed.

Turning now to the authors’ model, their Auerbach-Kotlikoff (A-K) overlapping generations (OLG) model is quite standard, and is simpler than many recent vintages of OLG models used for simulating tax and social security policies. First, labor supply is fixed, so that potential changes in the labor supply of women and the elderly cannot be captured. The decline in the growth of lifetime wealth (from higher taxes and slower GDP growth), and the rise in wages (from fewer prime age males), should in reality both induce a higher supply of labor.

Second, government behavior in the authors’ model is too mechanical. The authors assume an exogenous path of the government debt/GDP ratio, which decreases by 0.5 percent from 2002 to 2013 and is essentially flat from then on. Given the growth in government expenditures, this debt/GDP ratio path is achieved by adjusting consumption tax rates. Where does this debt/GDP ratio path come from? If the path is the government’s stated goal, what is the relationship of this goal to the welfare of the Japanese people? For example, does this path maximize long-run per capita consumption (is the path “dynamically efficient”)?

There also may be a dynamic inconsistency problem in the authors’ specification of the debt policy. As is well known, there is no reason for the public to believe in any pre-announced path of the government debt/GDP ratio, because the government has an incentive to renege. For example, in 1997, the Japanese government carried out fiscal reform, with an announced goal of eliminating fiscal deficits by 2003. This goal was not met, as the weak economy of the late 1990s forced the government to adopt a wide variety of pump priming measures. In fact, the experience of the 1990s has made the announcement of Japanese fiscal rules not too credible.

In other words, the authors should try to better justify their debt policy rule, so that it is consistent with the welfare maximization problem of their consumers. In Dekle (2005, chapter 4), for example, I explicitly incorporate the Japanese government’s fiscal (taxation and public investment) policy into the consumers’ welfare problem, so that the path of government debt, which arises endogenously, both maximizes consumer welfare and is dynamically consistent.

Third, it is not entirely plausible to assume, as the authors’ do, that the Japanese economy is closed. Clearly, the Japanese can lend to and borrow from abroad. As the Japanese age and saving rates decline, domestic investment can be supported by foreign capital inflows. This should lead to higher future GDP levels and a lower debt/GDP path.

In sum, this is a fine chapter that takes institutional details of the Japanese fiscal system seriously. The authors’ separate modeling of the evolution of the Japanese social security, medical care, and general budget accounts is particularly noteworthy. Most previous papers simply consolidated the general government account with the social security and medical care accounts. As the authors point out, given the current institutional framework, the government cannot raid the accumulated surpluses on its social security account to finance deficits in the general account. (However, some Japanese academics such as Hatta and Oguchi (1999) have proposed that the accounts should be consolidated; since it makes no economic sense to separate them. In fact, Hatta and others have proposed that consumption taxes should be used to pay not only for general government expenditures, but also for future social security benefits. In an actuarially unfair social security system such as Japan’s, dedicated social security contributions make little sense, since large and increasing subsidies from the general government account are necessary to pay the promised benefits to the elderly.)

Comments by Toshiki Tomita

In the face of the accelerating trends towards a smaller population and an aging society, are Japan’s public finances truly sustainable? Several long-term simulations concerning this vitally important issue have already appeared in the literature. The simulation results reported by Professor Ihori and his collaborators, which are based on a general equilibrium model, serve as a significant warning that the earlier simulations, such as those by Japan’s Cabinet Office (2005) and by Broda and Weinstein (2004), may have been excessively optimistic about the sustainability of Japan’s fiscal situation.

According to the work by Ihori and his colleagues, if the social security system remains unchanged and if a positive primary balance is achieved by 2010, the debt-to-GDP ratio will still increase from the current 120 percent to 175 percent, and the so-called national burden (the sum of the ratios of taxes and social security contributions to GDP) will reach 59 percent in 2050 from the current 32 percent. If the achievement of a positive primary balance is delayed by ten years, they estimate that the debt-to-GDP ratio will increase to 450 percent and the national burden will reach 79 percent of GDP.

Moreover, in consideration of Japan’s actual budget system, Ihori and his colleagues do not consolidate in their model the public pension account from the general government account and calculate the impact of the pension reform that was adopted in 2004. While they found that this pension reform may improve the fiscal situation to some extent, their model clearly indicates that a primary surplus must be achieved within the next ten years in order to avoid fiscal disaster.

The fiscal outcome reported by Ihori contrasts sharply with the results of Broda and Weinstein (2004), which The Economist summarized as “gray hair, red ink, but blue skies.” The contrast is largely attributable to the difference in how the two studies treat future social security benefits, especially public health insurance benefits.

Ihori and his colleagues found a sharp increase in aggregate spending for public health insurance benefits, which appears obviously necessary if the existing public health benefit system is unchanged while the population continues to age. Today in Japan, the average hospital stay is 28.3 days (compared with only 6.6 days in the United States and 11.6 days in Germany) and citizens over 70 years old receive 4.9 times the public medical insurance benefits per capita of younger age groups. In contrast, Broda and Weinstein (2004) assumed either that government spending for all age categories, including seniors, grows at the same rate as per capita GDP or, in a more generous scenario, that per capita benefits to those over 65 grow at the same rate as the per capita income of the working generations.

The gap between the growth rate and the nominal interest rate affects critically the debt dynamics. In the results reported by Ihori, the endogenously determined interest rate is about 3 percentage points higher than the growth rate, while the simulations run by the Cabinet Office constrained these two rates to be equal.

I would like to elaborate on the relationship between the simulation results and interest rates on government bonds. Despite Japan’s huge amount of outstanding government bonds interest rates have persisted at historically very low levels since 1990. These rates are lower than the rates on British government bonds during the Victorian age and even lower than the rates recorded in the Republic of Genoa during the period between 1611 and 1621.

One explanation for the continued low rates on Japanese government bonds is that investors are optimistic about the future of Japan’s public finances and consider fiscal bankruptcy an impossibility. Broda and Weinstein (2004) support this view.

In contrast, the simulation results reported by Ihori suggest that Japan’s public finances may not be sustainable and that the government may go bankrupt unless the tax burden ratio is substantially increased to a level that is considered extremely difficult politically. In the case of this unsustainable scenario, theoretically, taking into account the risk of default, government bonds must require a high rate of interest.

What, then, is the actual situation in Japan’s government bond market? I would first like to discuss the background factors behind the historically low interest rates. With what is generally known as the Fisher equation, long-term interest rates can be expressed as the sum of the real interest rate, the expected inflation rate, and risk premiums derived from various uncertainties such as credit risk.

In March 2004, Japan started to issue ten-year inflation-indexed government bonds, whose principal and interest are linked with the consumer price index. Interest rates on these ten-year inflation-indexed government bonds have been hovering at the level of 0.5 percent since August of last year. We can observe the real interest rate directly from the market rates on these bonds.

Suppose risk premiums are the same for these inflation-indexed government bonds as for government bonds with fixed nominal interest rates. The credit risk on these two types of bonds is also identical because they are both issued by the Japanese government. Interest rates on the ordinary government bonds, however—unlike the rates on the indexed bonds—include an inflation risk premium reflecting expectations about future fluctuations in inflation rates. Interest rates on the inflation-indexed government bonds, on the other hand, include a liquidity risk premium reflecting the difficulty in trading bonds with limited amounts outstanding.

Thus, by comparing the market interest rates on these two types of government bonds, we can observe real interest rates and expected inflation rates. Doing so reveals that the reason why interest rates on Japanese government bonds have been so low since 1990 is that expected rates of growth and inflation were low.

The simulations of the fiscal situation produced by the Cabinet Office (2005) set the so-called Solow residual at around 1 percent and both the nominal interest rate and the nominal economic growth rate at 4 percent after 2013. When these simulation results were announced, the government bond market showed absolutely no reaction. Thus it appears that market expectations about growth and interest rates are more in line with those of Ihori’s model than with the forecast by the Cabinet Office.

At the same time, international financial markets have required default risk premiums on Japanese government bonds since the Russian financial crisis in August 1998. Figure 4.1 shows the interest rates on yen-denominated World Bank bonds and on Japanese government bonds with almost the same maturity (February or March, 2008). While interest rates for the Japanese government bonds were below those on the World Bank bonds during 1998, yield spread on Japanese government bonds has subsequently increased by about 10 basis points. Importantly, risk premiums are still being required of Japanese government bonds even though the general trend in yields on both bonds has decreased gradually as they approach maturity.

Figure 4.1(a)Japanese government-guaranteed bond (JGB) and yen-dominated International Bank for Reconstruction and Development (IBRD) bond

Source: Bloomberg.

Figure 4.1(b)Yield spreads (JGB-World Bank)

Source: Bloomberg.

The credit risk on Japanese government bonds can be recognized more clearly by comparing bonds issued in foreign currencies. Figure 4.2 compares pound-denominated Japanese government guaranteed bonds (issued by Kansai International Airport Co., Ltd., with maturity in September 2007) with British government bonds having almost the same maturity. The spread between these bonds expanded rapidly in August 1998 and although it has been narrowing as maturity approaches, in June 2005 it remains at 40 basis points, which is a wider spread than at the time the bonds were issued in August 1997.

Figure 4.2(a)Pound-denominated JGB and UK Treasury (UKT)

Source: Bloomberg.

Figure 4.2(b)Yield spreads (pound-denominated JGB-UKT)

Source: Bloomberg.

There are two other facts to consider:

  1. Interest rates on Japanese government bonds have hovered around a level higher than the rates on yen-denominated Italian government bonds issued in February 2000.

  2. Looking at sovereign credit default swap trades, Packer and Suthiphongchai reported in the BIS Quarterly Review (2003) that government bonds of Japan recorded the third largest total trading volume after those of Brazil and Mexico.

These observations make clear that the world market has been requiring credit risk premiums for Japanese government bonds. In other words, investors are not so optimistic as to rule out the possibility that Japan’s public finances might go bankrupt. From this point of view as well, Ihori’s simulation results appear to be consistent with the formation of interest rates in the market.

Simulation results are largely dependent on parameters. In closing, I would like to raise some questions about the parameters assumed by Ihori and his collaborators.

First, while these simulations set the time discount rate for the household sector at 0.01, how should we interpret the large, persistent gap between the time discount rate, (δ), and the interest rate, (r), in the simulations?

Also, taking recent trends into account, the authors set labor’s share of output (a) at 0.64 and the depreciation rate for capital stock (δk) at 0.089.1 wonder if the authors believe that these levels can be maintained throughout the twenty-first century. I think that a high depreciation rate for the capital stock is closely related to the growth of the Solow residual, i.e., an increase in total factor productivity. These assumed levels for a and δk have a major impact on the difference between the interest rate and the growth rate.

Finally, the reported simulation results do not show the estimated values for public pension benefits (B), transfers from the general government account to the pubic pension account (ήB), and government consumption (ήB) including social security benefits in kind. Because these values are extremely important for comparing the authors’ results with those obtained by Broda and Weinstein, I would encourage Professor Ihori and his colleagues to report these data as well.


    Other Resources Citing This Publication