The Personal Savings Function of Urban Worker Households in Japan

Abstract

Listen

Translate article

SINCE the end of World War II, the Japanese economy has attained a rapid economic growth, and the necessary capital funds have been procured through personal savings rather than through forced savings via taxes or inflation. One of the remarkable phenomena is the more or less steady rise in the personal savings ratio, particularly in urban worker households. The savings ratio of urban workers rose from 2.0% in 1951 to 19% in the 1960s. Contrary to the expectation that the savings ratio in urban worker households would be stabilized, it kept increasing in the 1970s and reached 24% in 1974. This rate of 24% in 1974 is much higher than those of the other OECD countries, which range from England's 8% to West Germany's 15%. A puz'zling phenomenon in the 1970s is that the savings ratio increased as inflationary pressure mounted. After the oil crisis of 1973, for example, the savings ratio jumped by 1.7 points in 1974 under double digit inflation. Why has the savings ratio continued to increase? Does a conventional savings function explain it? Answers to these questions have serious implications for macroeconomic policies since if the high savings ratio is sustained when the Japanese economy enters a period of slower growth, one might observe excess liquidity. Furthermore, if it is found that the savings ratio is positively correlated with inflation rates, then personal income tax cuts to stimulate sagging consumption may lose effectiveness if the inflation rate is not controlled. The present paper attempts to answer the following two questions: (1) does a conventional savings function explain the urban worker household's savings behavior in the sixties and seventies? and (2) if not, what would explain the puzzling rise of personal savings ratio in the seventies? Attempts to explain Japanese savings (or consumption) have been numerous: Blumenthal (1970), Shinohara (1959, 1962, 1970a,b), Komiya (1966), Mizoguchi (1970), Noda (1970) and Yoshihara (1972), among others, have attempted analyses. Many of these studies applied the conventional consumption functions of the relative income, permanent income and life-cycle hypotheses. Among the various studies we first examine Shinohara's savings function. We apply two tests: a Bayesian test of parameter shift, and an ex post forecasting test. The Bayesian shift test was applied after the Chow test failed mainly due to the fact that one of the crucial assumptions necessary to conduct the Chow test does not seem to hold for our data. In contrast to previous studies, a recent study by Sato (1976) argues that the savings behavior of urban worker households reflects the endogenous socio-economic motivations to save. To answer question (2) above, therefore, we sharpen Sato' s formulation to explicitly introduce those variables that reflect socio-economic motivations for savings. The organization of the paper is as follows: in section II, after a brief survey of some studies of the Japanese savings function, we examine the validity of Shinohara's formulation in the sixties and in the seventies by the Chow test and a Bayesian test of parameter shift. In addition to Shinohara's formulation we shall examine the applicability of a Houthakker-Taylor (1970) type dynamic savings function and compare it to Swamy' s (1968) application to international data (one of the countries studied there is Japan). In section III, Sato's formulation (1976) is modified and estimated using time series as well as cross section data. The modified Sato formulation is compared to the Houthakker-Taylor formulation. Section IV gives concluding remarks and possible policy implications of the findings. Received for publication July 7, 1977. Revision accepted for publication March 22, 1978. * University of Pennsylvania. I am grateful to Professor Hiroki Tsurumi of Rutgers College for suggestions and encouragement, and also to anonymous referees of this REVIEW whose comments improved the paper. Responsibility for any remaining error is of course mine.