3. Household Savings in Japan

Alessandro Zanello, and Daniel Citrin
Published Date:
November 2008
  • ShareShare
Show Summary Details


Japan’s household savings rate has fallen dramatically since the early 1980s (Figure 3.1). The savings rate, based on national income accounts data, declined from 18 percent in 1981 to 6 percent in 2002 with a brief pause during the bubble years. This drop has prompted discussions about the prospective evolution of Japanese savings and Japan’s long-term growth prospects in the context of a rapidly aging population.1

Figure 3.1Household Savings Rate


Source: Cabinet Office.

This chapter focuses on the implications of demographic change for household saving, and presents illustrative projections. After a review of household saving behavior in Japan from a cross-country perspective and of standard models of the impact on saving of demographics, we consider the empirical evidence. This is done using both aggregate data and more detailed information on cohort behavior. The two approaches provide different perspectives on likely savings trends, going forward.

Japanese savings in an international context

In a cross-country context, the decline in Japan’s household savings rate stands out. There are wide disparities in saving behavior across Organization for Economic Cooperation and Development (OECD) member countries (Figure 3.2). The saving rate in the United States has also fallen, but from a low level, while savings rates in some continental European countries, such as Germany, have been reasonably stable. Japan, once considered a big saver relative to other OECD countries, has seen its savings rate fall below that of many other OECD members, converging toward the OECD average.

Figure 3.2Savings Rates: Selected OECD Countries


Source: OECD.

In spite of the secular decline in their savings, the net wealth of Japanese households remains high relative to that in other industrialized countries (Figure 3.3). Japan’s large stock of wealth is the result of high savings rates during much of the postwar period and the rapid increase in land prices in the 1980s that pushed nonfinancial wealth to very elevated levels. Since the collapse in asset prices, the stock of wealth has remained flat in relation to income. By comparison, net wealth in the other G7 countries has increased since the late 1980s, although some countries have experienced some unwinding of that increase, notably the United States after the “Tech Bubble” burst in 2000.

Figure 3.3Net Wealth: Selected OECD Countries

Source: OECD.

Savings and demographics

Modern theories of consumption find a strong link between demographics and savings. The life-cycle model of household behavior regards saving as motivated by the desire of households to smooth lifetime consumption in the face of uneven income flows. Households save a portion of their income during their working years in order to finance their retirement. Thus, the higher the ratio of the retired population to the working-age population, the lower aggregate savings will be (Figure 3.4).

Figure 3.4Old-Age Dependency Ratio


Source: National Institute of Population and Social Security Research.

Accordingly, population aging has important implications for the future savings behavior. Official projections indicate that Japan’s population is expected to stop growing before the end of the current decade and then is projected to decline by 27 million over the next 50 years (Figure 3.5). 2 (Clearly, long-term projections of this kind are subject to considerable uncertainty at long horizons.) Two main factors contribute to this demographic trend: the Japanese population has the lowest birth rate and the highest life expectancy among industrial countries, while inward immigration flows—typically involving young workers—are very low by international standards. 3

Figure 3.5Japanese Population Projection


Source: National Institute of Population and Social Security Research.

Another important demographic factor is the development of the old-age dependency ratio. This ratio measures the burden an aging society places on its working-age population and is defined as the ratio of the population aged 65 and over to the population aged 20-64. In 2000, this ratio was 28 percent. With the projected aging population, this old-age dependency ratio is expected to rise to 70 percent by 2050. Overall, these demographic trends will have a significant effect on the aggregate savings rate.

There have been numerous studies of Japanese saving behavior.4 Most of the earlier studies focused on explaining Japan’s high savings rate before its rapid decline in the 1990s. Recently the focus has turned to explaining the downward trend in savings—and the role of the rapidly aging population. These studies have used aggregate household savings data from the national income accounts as well as data from household surveys. The next part of this chapter reviews the macro evidence on saving and demographics, while the following section examines household survey data, looking at saving behavior across different age cohorts. Combined with projected demographic trends, the analysis provides illustrative scenarios for household savings for the medium and longer term.

Macro evidence on the impact of demographics on savings

To quantify the interrelation between population dynamics and household savings, a standard aggregate saving equation is estimated first. As discussed above, an important prediction of the life-cycle model is that an increase in the old-age dependency ratio would reduce the aggregate savings rate. In addition, the model suggests that the larger the share of the young in the population, and the greater the stock of wealth, the lower the savings rate.5 In some econometric specifications, separate fiscal variables have been added, but this is not done in the present study.6 To conduct the empirical analysis, annual national income data from 1980 to 2003 is used.

The estimated aggregate household saving equation for Japan is consistent with the predictions of the life-cycle model (Table 3.1).7 In particular, the coefficients on the old-age dependency ratio, young-age dependency ratio (defined as ratio of minors to working-age population), and wealth as a share of disposable income are all negative, while the coefficient on inflation is positive, capturing precautionary motives.8 All of these estimates are statistically significant at the 10 percent level or above. As wealth increases, the savings rate falls. As the number of retirees relative to the working-age population increases, the savings rate declines as retirees tend to spend out of their accumulated stock of wealth. Similarly, as the number of minors relative to the working-age population increases, the savings rate declines, as those who have not yet begun to work consume and do not save.

Table 3.1Long-Run Saving Rate Model
Wealth to disposable income-0.04-2.72
Old age to working population-0.25-2.15
Young to working population-0.20-1.41
R-square0.94 Durbin-Watson1.7
Sample1980-2003 No. observations24
Source: IMF staff estimates.
Source: IMF staff estimates.

For illustrative purposes it is possible to highlight the effect demographic changes may have on aggregate savings. Based on the coefficient estimates and projections for the explanatory variables, household savings rates are projected to decline significantly. For comparison, an initial or benchmark projection for saving for 2004-10 is generated that is based on setting the path of all the variables equal to their 2004 value. This yields a constant savings rate of about 6.0 percent. An alternative path for the household savings rate, using estimates for the explanatory variables, projects savings to decline 2¼ percentage points by 2010.

The demographic effects of aging are both statistically and economically significant. The aggregate savings rate projection (Figure 3.6) can be decomposed to show the contribution of each explanatory variable to the decline in the saving ratio: the projected rise in the old-age dependency rate more than fully accounts for the projected decline in savings over the medium term (see Table 3.2). The simulation of the model further into the future indicates that the household savings rate is likely to decline steadily in the coming decades as a result of the shift toward a more elderly population.

Figure 3.6Saving Rate Projection


Source: IMF staff estimates.

Table 3.2Contribution to Change in Saving Rate, 2004–10
Source: IMF staff estimates.
Source: IMF staff estimates.

Evidence using household survey data

Further insight on the likely evolution of saving behavior can be gained by looking at household survey data.9 In particular, the data from the Family Income and Expenditure Survey (FIES) are used to examine the major features of the household savings rate, including the saving propensities across different age groups.10 The general trend of the household savings rate based on the survey data is somewhat different from the trend revealed by the aggregate national accounts data. The discrepancy between these two data sources stems from differences in the way the two savings rates are defined, in particular the treatment of rents. Indeed, several authors have argued that it is quite difficult to reconcile the two series, owing to the number of transformations that are undertaken to derive the national accounts data.11 Nevertheless, using the two data sources is helpful because each provides a different snapshot of household savings.

Household saving behavior, as revealed by the survey data for the period 1995 to 2003, differs across age cohorts (Table 3.3). The data display a humped-shape pattern consistent with the life-cycle model. The savings rates are positive and large for each age category, with the 30-39 age group reporting the highest savings rate. As expected in a life-cycle model, the savings rate of elderly households (households with a head who is 60 years old or older) is lower than that of the younger households.

Table 3.3Savings Rate of Households
29 Years or30-3940-4950-5960 or
YearAll AgesYoungerYearsYearsYearsOlder
Source: Family Income and Expenditure survey; IMF staff estimates.
Source: Family Income and Expenditure survey; IMF staff estimates.

In general, the savings rates for the different cohorts have been fairly stable over time. There is one exception to this observation. The savings rate of elderly households with salaried workers has declined rapidly since the late 1990s, dropping to 12.1 percent from 22.6 percent (Figure 3.7). This decline has been associated with a fall in disposable income for this cohort while consumption remained roughly constant. The survey data indicate that the decline in disposable income was due to a fall in both employment income and public pension income.

Figure 3.7Saving Rate: Elderly (60 and Over)


Source: Family Income and Expenditure Survey.

The micro data allows for a comparison of the saving behavior for different types of elderly households. The savings rate in Table 3.3 pertains to elderly households whose head of household is still working as a salaried worker and does not include information about the retired elderly. In 1995 the FIES started collecting data on a broader set of elderly households including retired elderly households (Figure 3.8). The savings rate of retired households (where the head of household is older than 65 and retired) is negative and large in absolute magnitude, ranging from -5.2 percent to -19.6 percent. The aggregate savings rate of both retired and working households is -2.3 percent, compared with 23 percent for the working-age households computed from the 2003 FIES.

Figure 3.8.Projected Saving Rate


Sources: Family Income and Expenditure Survey; IMF staff estimates.

Cohort saving data allow one to measure the likely impact of demographic changes on the savings rate. We assume that the savings rates across different age groups remain unchanged from their most recent (2003) levels. This is a reasonable assumption as the savings rates of each age category (except elderly households) have been relatively stable.12 Population trends can then be used to project the aggregate savings rate out to 2050. This projection shows that the age distribution of the population is shifting toward older age groups and away from the younger age categories.13 For example, in 2000, the population over 16 years of age was 108 million and is projected to drop to 88 million by the year 2050. The share of the population over the age of 60 is projected to increase from 27 percent to nearly 50 percent by 2050, with the most rapid increase in the share of the elderly occurring in the current decade (Figure 3.9).

Figure 3.9.Projected Age Distribution of Population


Source: National Institute of Population and Social Security Research.

The cohort saving behavior and the population projections illustrate once again that the savings rate is likely to fall in the medium and longer term. The extent of the decrease is likely to depend importantly on the saving behavior of the over -60 population, which in turn will depend in part on the elderly’s employment status. As a lower bound, it is assumed (although implausibly) that there is no retirement. This projection captures simply the effect aging has on savings. Given this assumption, the projected savings rate declines 3½ percentage points by 2050, with the largest ten-year decline (1 percentage point drop) occurring in the current decade. A more plausible projection, based on the assumption that anyone older than 70 is retired and the 60-69 population remains working, generates a savings rate that declines 9 percentage points by 2050.14 Once again, the largest fall (2½ percentage points) occurs in the current decade.

Summary and conclusion

This chapter has analyzed the likely impact on the household savings rate from a shift of Japan’s demographic structure toward a more elderly population. Projections using both the aggregate data and the cohort data illustrate that the household savings data are likely to decline steadily in the coming decade. The results using the macroeconomic data suggest the savings rate will fall 2¼ percentage points after six years, while the cohort data indicate a decline, ranging between 1 to 2½ percentage points, depending on the elderly’s employment status.


    BosworthBarryGaryBurtless and JohnSabelhaus1991The Decline in Saving: Evidence from Household SurveysBrookings Papers on Economic Activity Vol. 1 pp. 183241.

    De MelloL.P.M.Kongsrud and R.W.R.Price2004Saving Behavior and the Effectiveness of Fiscal PolicyOECD Economics Department Working Paper No. 397.

    De SerresA. and F.Pelgrin2003The Decline in Private Saving Rates in the 1990s in OECD Countries: How Much can be Explained by Non-Wealth Determinants?OECD Economic Studies No. 36.

    HayashiFumio1989Japan’s Saving Rate: New Data and ReflectionsNBER Working PaperDecember.

    HayashiFumio1997Understanding Saving: Evidence from the United States and Japan (Cambridge, Massachusetts: MIT Press).

    HayashiFumioAlbertAndo and RichardFerris1988Life Cycle and Bequest Savings: A Study of Japanese and U.S. Household Based on Data from the 1984 and 1983 Survey of Consumer FinanceJournal of Japanese and International Economies Vol. 2December pp. 45091.

    HoriokaCharles1984The Applicability of the Life-Cycle Hypothesis of Savings to JapanKyoto University Economic Review Vol. 54October pp. 3156.

    HoriokaCharles1997A Cointegration Analysis of the Impact of the Age Structure of the Population on the Household Saving Rate In JapanReview of Economics and Statistics Vol. 79August pp. 51116.

    HoriokaCharles2004aThe Stagnation of Household Consumption in Japanunpublished paper. Institute of Social and Economic ResearchOsaka UniversityFebruary.

    HoriokaCharles2004bDo the Elderly Dissave in Japan?Discussion Paper 605 Institute of Social and Economic ResearchOsaka UniversityJune.

    HoriokaCharles2004cAre the Japanese Unique? An Analysis of Consumption and Saving Behavior in JapanDiscussion Paper 606 Institute of Social and Economic ResearchOsaka UniversityJune.

    International Monetary Fund (IMF)2004How Will Demographic Change Affect the Global Economy?World Economic OutlookSeptember (Washington, D.C.: International Monetary Fund).

    KitamuraYukinobuNoriyukiTakayama and FumikoArita2001Household Savings in Japan RevisitedResearch in Economics Vol. 55 pp. 13553.

    KogaMaiko2005The Decline of Saving Rate and the Demographic EffectsBank of Japanunpublished paper.

    KozuTakashiYoshikoSato and MasakazuInada2003Demographic Changes in Japan and Their Macroeconomic EffectsBank of Japan Working Paper No. 04-E-6.

    MeredithGuy1995Demographic Change and Household Saving in Japan,” in Saving Behavior and the Asset Price “Bubble” in Japan: Analytical Studies” ed. UlrichBaumgartner and GuyMeredithIMF Occasional Paper 124 (Washington, D.C.: International Monetary Fund).

    Organization for Economic Cooperation and Development (OECD)2004Economic Outlook No. 76Chapter V: Saving Behavior and the Effectiveness of Fiscal Policy.

    YashiroNaohiro and Akiko SatoOishi1997Population and Savings-Investment Balance in JapaninThe Economic Effects of Aging in the United States and Japan ed. MichaelHurd and NaohiroYashiro (Chicago: University of Chicago Press).

The September 2004 World Economic Outlook (IMF, 2004) investigates how demographic change affects regional and global economies.

This projection is from Japan’s National Institute of Population and Social Security Research and represents the “medium variant.” The Institute also produces “high” and “low” variants for its projections.

The fertility rate in Japan is at present 1.2, compared with 2.1 in the United States. Life expectancy at birth is currently 81 years, the highest in the world and a full five years higher than in the United States. Since the 1950s, U.S. life expectancy has increased by 7 years, while Japanese life expectancy has risen by 17 years. Among OECD countries, Japan has one of the lowest shares of foreigners in the total population and labor force. In 1998, the population share of foreigners in Japan was 1.2 percent, one order of magnitude lower than in other advanced countries.

Important earlier studies include: Hayashi (1989, 1997), Hayashi and others (1988), Horioka (1984, 1997), and Yashiro and Oishi (1997). More recent studies explore the sharp decline in savings and the impact of demographics, including Horioka (2004a), 2004b, 2004c) and Koga (2005).

The influence of wealth on savings may be ambiguous. During the late 1980s, the decline in savings rates coincided with sharp increases in household net wealth, indicating a strong negative correlation between the two series. At other times, the correlation between savings and wealth is positive, reflecting an increase in both income and wealth.

Fiscal variables have been used to explain private savings and test for the possible presence of Ricardian equivalence, whereby forward-looking agents offset bond-financed tax reductions by higher savings. We do not follow this approach because the focus is on household rather than private savings. See, for example, OECD (2004), De Mello and others (2004), and De Serres and Pelgrin (2003). The OECD estimates the offset coefficient to be about 35 percent in the short run and 70 percent in the long run.

As pointed out by Meredith (1995), evidence based on aggregate data typically support the prediction of the life-cycle model regarding demographics and savings.

There are several ways in which inflation can affect saving independently from its influence via the real interest rate. The increase in inflation may capture the effect of uncertainty such that as inflation rises, saving would increase for precautionary motives. Also, an unexpected rise in inflation may lead households to hike savings to compensate for capital losses.

This approach has been followed, for example, in Bosworth and others (1991), Kitamura and others (2001).

The survey reports savings rates by age group for head of household for salaried workers, with two or more members. This coverage was expanded to include data on retired households (head of household over 60) after 1995.

See, for example, Hayashi (1997), Kozu and others (2003), and Koga (2005). In particular, according to Hayashi (1997) comparing the two savings rates is not straightforward because the SNA data involves many stages of imputations and transformations, while the survey data concern cash items. Specifically, he points to six major differences: (i) imputed rents; (ii) business income; (iii) employment income; (iv) medical expenditures; (v) social security benefits; and (vi) depreciation.

Pension benefits are expected to fall further, but a reasonable assumption may be that the elderly extend their working life enabling them to maintain their savings rate.

This scenario uses the medium variant population projection. The alternative high/low projections also yield similar results.

Applying the average savings rate of retired and working households yields a savings rate that declines 2¼ percentage points in the first decade and 6¼ percentage points by 2050.

    Other Resources Citing This Publication