Excess sensitivity of consumption, liquidity constraints, and mandatory saving

Abstract

Using Singapore mandatory saving system, it is examined whether liquidity constraint is a major reason for the excess-sensitivity of consumption to predictable income growth. Although the mandatory saving rate for employees could be a good measure for the financial condition of a liquidity-constrained consumer, it is found, through the nonlinear instrumental variable estimation, that consumption growth is not sensitive to changes in the mandatory saving rate for employees. This finding suggests that liquidity constraints would not be a major reason for the excess-sensitivity puzzle.

Acknowledgements

We would like to thank econometric studies unit at National University of Singapore and the enquiry units in various Ministries including Monetary Authority of Singapore, Ministry of Manpower, Ministry of Trade and Industry and Central Provident Fund Board.

Notes

Campbell and Mankiw (1991) also imply that consumers’ rule-of-thumb behaviour has a relation with liquidity constraints through an international comparison.

The results of Granger causality tests are available upon the request.

Currently, the mandatory saving can be used not only after retirement but also for medical cost, the purchase of house, education and insurance.

The remaining 33% of working force who does not participate in the CPF scheme includes foreign workers, casual and part-time workers and certain categories of the contract workers.

The quarterly consumption data in Singapore has not been decomposed into durables and non-durables and services. Hence, the consumption data inevitably include consumption on durable goods. However, the inclusion of durable goods would not result in a large distortion in the following analysis because the fraction of consumption on durable goods (clothing and footwear, furniture and household equipments, personal transport equipment, etc.) in total consumption from annual data has been fairly stable at 0.19.

Since

in the regression of income on instruments dated t−2 is far lower than that in the regression on instruments dated t−1 in Singapore, the result suggests that instruments dated t−1 have superior power to instruments dated t−2 or before.

Other sets of instruments have been tried such as lags 1–4 of consumption growth only or additional inclusion of change in nominal interest rates to each set. The results are similar to those reported in this paper.

The insignificant coefficient on CPF rate for employees is consistent with the results of Abeysinghe and Choy (2003), although they did not relate this to the permanent income hypothesis.