Information about Asia and the Pacific Asia y el Pacífico
Chapter

VII Determinants of Private Saving in Singapore

Author(s):
Kenneth Bercuson
Published Date:
December 1995
Share
  • ShareShare
Information about Asia and the Pacific Asia y el Pacífico
Show Summary Details
Author(s)
Aasim M. Husain

A dissaver in the early 1960s, Singapore over the past three decades has become the thriftiest country in the world, saving on average 41 percent of its GNP in the 1980s. This section presents an econometric analysis that attempts to identify the factors behind Singapore’s remarkable saving performance.

Three factors appear to be responsible for the increase in Singapore’s saving rate by 24 percentage points over 1970-92. First, the sizable increase in the ratio of working age population to total population—from about 55 percent in the 1960s to about 70 percent during the 1980s—affected private saving significantly. Indeed, the econometric analysis reported below indicates that this demographic change raised the gross national saving rate by 20 percentage points.

Second, rapid growth of private disposable incomes, averaging about 9 percent in real terms and 6 percent in real per capita terms during 1970-92, also played an important role in accounting for the increase in the saving rate, raising it an estimated 2 percentage points.

Third, the high level of public saving, which increased from 5 percent of GNP in the early 1970s to over 10 percent of GNP in recent years, strongly influenced gross national saving, although part of the effect was offset by private dissaving. The estimates reported below suggest that if public saving had not increased, but, instead, public sector consumption had risen by 5 percent of GNP, private consumption would have been 3 percent of GNP lower. This suggests that the increase in public saving raised the gross national saving rate by 2 percentage points.

Finally, mandatory saving with the CPF, which increased relative to GNP from about 3 percent in 1970 to over 13 percent in recent years, was not a significant determinant of overall private saving. The empirical estimates indicate that increases in compulsory saving were offset by decreases in voluntary private saving.

Historical Trends

Singapore’s gross national saving rate, which averaged about 16 percent of GNP during 1965-69, grew steadily to an average of 28 percent in the 1970s and 40 percent during 1980-84. After declining somewhat in 1985-87, the saving rate recovered to 41 percent in 1988 and reached a record level of 45 percent in 1992 (Chart 7-1 and Table 7-1). Domestic saving financed an increasing share of gross capital formation, rising from about 50 percent in the early 1970s to almost 100 percent in 1985. Singapore became a net capital exporter in 1986, and saving has exceeded investment since then.

Chart 7-1Gross National Saving Rate

(In percent of GNP)

Sources: Singapore authorities; and IMF staff estimates.

Table 7-1Gross National Saving Rate(In percent of GNP)
1970-741975-791980-841985-891990-92
Gross national saving rate24.431.242.239.743.6
Private saving17.923.432.230.333.0
Central Provident Fund (CPF)4.88.914.214.413.8
Other13.214.517.915.919.1
Public saving6.57.810.09.410.7
Source: IMF staff estimates.
Source: IMF staff estimates.

Although a breakdown of saving into public and private sectors is not published officially, public saving can be approximated from the Government’s fiscal accounts, with private saving the residual (Chart 7-2). Calculated in this way, government saving has been more stable than private saving since 1970, the first year for which fiscal accounts data are available with the required classification. The government saving rate averaged about 7 percent of GNP in the 1970s, and rose rapidly in the early 1980s to 12 percent of GNP in 1983, before declining to an average of less than 9 percent of GNP during 1984-87. Most recently, the rate was in excess of 10 percent of GNP each year during 1988-92.

Chart 7-2Public and Private Saving Rates

(In percent of GNP)

Source: IMF staff estimates.

Both the growth and the volatility of the private saving rate have been far more pronounced. Starting from under 15 percent of GNP in 1970, the rate rose steadily until the early 1980s. After a surge to more than 33 percent of GNP in 1984, the private saving rate declined during 1985-88, before gradually rising back to 34 percent of GNP in 1992.

These estimates, however, must be interpreted with caution for three reasons. First, saving by statutory boards is not included in the central government’s current budget surplus. While large operating surpluses of the boards in recent years indicate that saving by the central government understates total public sector saving, the operation of most boards along commercial principles suggests that their saving should be counted as part of private saving. Second, central government data are recorded by fiscal years (April-March), whereas national saving is measured on an accrual basis for calendar years. For the purposes of this study, fiscal year budgetary data are converted to calendar years under the assumption that flows of fiscal revenue and expenditure are constant throughout the fiscal year. Third, distinctions between current and capital expenditures of the Government can be arbitrary and may bias the estimate of the current account surplus. The resulting bias induced by such classification, however, is unlikely to have changed significantly over time.

Central Provident Fund

An important component of private saving is the CPF. Initiated in 1955, the CPF is a compulsory saving scheme under which employees and employers are required to contribute a portion of labor income to the CPF. CPF accounts are specific to individuals. Contribution rates were originally set at levels at which individuals could accumulate enough savings to finance a basic retirement income. Over the years, the CPF has been broadened as a saving vehicle to help finance individuals’ housing, health, investment, and other outlays.

Starting from mandatory contribution rates of 5 percent of wages by employees and 5 percent by employers in 1955, the rates of contribution remained unchanged until 1968 but were then gradually increased to a peak in 1984 of 25 percent each for employees and employers. In 1986, the employers’ contribution rate to the CPF was reduced to 10 percent as a measure to counter the recession, but since then it has been increased in steps to 18 ½ percent as of July 1993. The employee contribution rate, in contrast, was reduced only slightly, to its present level of 21 ½ percent (Table A10).1 The intention is to equalize the employer and employee rates at 20 percent over the medium term.

The CPF scheme extends to all employees and self-employed persons, in both the private and the public sectors. Membership—the number of persons with accounts in the CPF—has grown at a rate similar to that of the labor force. As a fraction of the total population, membership has climbed steadily, from about half in 1976 to over three fourths in 1991.

CPF saving has increased relative to GNP since the 1960s, but the CPF’s share of total private saving has declined in recent years.2 Starting from about 2 percent in the late 1960s, the ratio of CPF saving to GNP increased steadily to average about 15 percent in the early 1980s, before declining during the recession in the mid-1980s and then rebounding to almost 13 percent in recent years. The ratio of CPF saving to total private saving rose steadily from about 25 percent in the early 1970s to about 50 percent in the mid-1980s, before declining to about 40 percent during 1988-92.

The main purpose of the CPF is to provide benefits for members upon retirement. Accordingly, members may withdraw their CPF savings, after setting aside a fixed amount, upon reaching the age of 55. The fixed amount is used to pay monthly returns to members after they reach the age of 60. In 1993, the fixed amount that had to be left with the CPF was S$34,600 per individual with an associated monthly income of S$266. In addition to retirement benefits, withdrawals from the CPF are also permitted in the event of permanent disability or upon permanent emigration from Singapore. Other uses of CPF savings have been liberalized over time. Under the liberalization policy, CPF savings could initially be used to purchase publicly built housing and, later, private residential property. As a result, home ownership is now nearly universal.3 More recently, withdrawal facilities have been expanded to include an investment scheme, under which members may invest part of their saving in trustee stocks, unit trusts, convertible loan stocks, and gold;4 a Singapore Bus Service (SBS) scheme, which allows members to purchase SBS shares from their CPF savings; a nonresidential properties’ scheme, under which CPF savings may be invested in nonresidential property; an education scheme, which enables members to borrow part of their CPF savings to pay for their own or their children’s tertiary education; Medisave, under which withdrawals to meet hospitalization expenses may be made; and CPF-administered life, health, and home insurance schemes.

Withdrawals of CPF savings have increased over time, from an annual average of about S$550 million during 1976-79 to S$4,340 million during 1987-92, reflecting the liberalization of approved withdrawal schemes as well as the increased stock of the total CPF balance (Table A11). As a ratio to the outstanding CPF balance, the increase in CPF withdrawals has been less pronounced, rising from an average of about 9 percent during 1976-83 to 13 ½ percent during 1984-88, then declining somewhat in recent years to about 10 percent.5 More than half of withdrawals in recent years have been for housing.

The sizable mandatory CPF contribution rates have attracted attention. While these rates appear high in comparison with social security schemes in other countries, the fully funded and individual-specific nature of CPF saving does not afford a direct comparison with pay-as-you-go social security programs of other countries. Also, expanding withdrawal facilities for home ownership and investment purposes has enlarged the role of CPF saving considerably beyond that of a social security scheme.

Assets of the CPF at end-1992 totaled S$52.5 billion (69 percent of GNP). Of this, SS45.6 billion was invested in government bonds, and S$6.1 billion was placed as advanced deposits with the MAS, pending the issue of such bonds. While the CPF is the main lender to the Government, the funds are not needed for budget financing and are invested (mainly as overseas portfolio investment) by the Government of Singapore Investment Corporation.

Interest earned on CPF saving balances is deposited into members’ CPF accounts. The implicit annual rate of return on CPF savings can be calculated as the ratio of interest paid to the average of CPF balances at the beginning and the end of the period. This calculation indicates that the implicit annual rate of return was about 6 percent for most of the period between 1976 and 1985. The average interest rate on bank deposits during this period was significantly higher. In recent years, the interest rate earned on CPF deposits has been based on the average deposit rates offered by the four major local banks, subject to a statutory minimum of 2.5 percent. Interest is paid twice annually and averaged 4 percent in 1992 and 2.6 percent in 1993, Employees’ contributions to the CPF are excluded from their taxable earnings both when contributed and withdrawn. In addition, earnings on members’ balances are tax exempt.

International Comparisons

Singapore’s gross national saving rate is high in comparison with those of other countries (Table 7-2).6 Although most developing countries in Asia have experienced a steady increase in saving rates since the early 1970s, the growth rate of saving achieved in Singapore as well as the level, as a ratio to GNP, were significantly higher than in all other countries in the sample. Singapore’s gross national saving rate rose from about 5 percentage points of GNP above the average for developing countries in the sample during 1970-74 to more than 15 percentage points of GNP higher than the average in 1990–91. Saving rates in recent years were significantly higher than those achieved even in Japan during the early 1970s.

Table 7-2Gross National Saving Rates in Selected Countries(Period overage; in percent of GNP)
1970-741975-791980-841985-891990-91
Group of Three27.424.323.223.824.6
Germany, Republic of127.322.721.924.726.8
Japan38.232.430.932.634.2
United States16.917.716.914.112.8
Newly industrializing economies25.229.632.435.035.5
Hong Kong226.329.429.332.432.5
Korea, Republic of19.024.624.234.335.9
Singapore24.431.242.239.743.6
Taiwan Province of China30.933.133.833.529.83
Developing countries18.123.426.326.226.7
India16.226.232.030.5
Indonesia14.121.226.929.233.4
Malaysia20.728.327.830.129.23
Pakistan14.217.924.623.221.6
Philippines22.225.025.819.818.8
Thailand21.321.720.424.630.33
Source: IMF, International Financial Statistics, various issues.

Before October 1990, data are for the Federal Republic of Germany only

In percent of GDP.

Data cover 1990 only.

Source: IMF, International Financial Statistics, various issues.

Before October 1990, data are for the Federal Republic of Germany only

In percent of GDP.

Data cover 1990 only.

An important factor contributing to the rise in national saving is the rate of public saving. Lahiri (1989) notes that in Indonesia, Malaysia, and Singapore, government saving as a proportion of GNP rose by 4 percentage points between the early 1970s and early 1980s. In Korea, the rise was 1.5 percentage points over the same period. However, differences in fiscal policies can explain only part of the variation in national saving rates across countries, and there is also substantial variation in private saving rates. In Singapore, increases in public saving contributed to the rise in national saving, but the bulk of the increase in the national saving rate was due to the increase in private saving.

Empirical Analysis of Private Saving

This section analyzes saving behavior in Singapore and examines the factors responsible for the high saving rate and for changes in the pattern of saving. The roles of economic growth and changing demographic characteristics are studied. In addition, the impact of public saving and mandatory saving through the CPF on voluntary (non-CPF) private saving is analyzed. Since a breakdown of private saving into household and corporate saving is not available, the empirical analysis focuses on total private saving. The results indicate that high growth and the changing age structure of the population were major factors responsible for Singapore’s high saving rate. In addition, the results suggest that high levels of public saving raised the overall saving rate, but their effect was partly absorbed by private saving.

Other factors have also contributed to the high rate of saving, but their effect is difficult to quantify. Some aspects of the tax system, for example, encourage saving. Capital gains are not taxed, and interest on deposits with the Post Office Saving Bank are tax exempt. Food and energy products are not subsidized, and a certificate of entitlement requirement on automobiles amounts to a substantial tax on consumption. In addition, the absence of a welfare system and unemployment benefits contributes to a higher level of saving. Finally, a low initial stock of wealth, relative to other newly industrializing economies, may also have accounted for a high saving rate.7

Economic Growth and Demographic Changes

Life-cycle theory suggests that economic growth and changing age structure are important determinants of the saving rate. The experience in Singapore appears to confirm this hypothesis. The ratio of working age population (defined as the population between the ages of 15 and 64) to total population increased steadily from 56 percent in 1968 to 70 percent in 1983, but subsequently stabilized at that level. Coinciding with the changes in age structure, the rapid increase in the private saving rate appears to have abated in the mid-1980s. An important qualification to the apparently close relationship, however, is that census information on the age structure of the population is available only every ten years or so. Data for intermediate years are constructed by demographers using interpolation techniques, and the correlation between age composition and the time trend can be expected to be high. Any apparent relationship between age structure and saving, then, could be due to the effect of a time trend or some other time-varying factor.

Fluctuations in the growth rate of real private disposable income, as well as its high trend level, also appear to have affected private saving. In particular, the slowdown in growth in 1982 and the recession in 1985 were associated with sharp falls in the saving rate, and the recovery in the late 1980s was accompanied by an increase in private saving, though the effect was more protracted.

The effects of demographics and growth on saving have been quantified in a number of studies. Lahiri (1989), in particular, estimated consumption functions derived from a simple life-cycle model for eight Asian countries, including Singapore. His estimates indicate that both the changing age structure of the population and output growth were important determinants of private consumption in almost all of the countries in his sample. In addition, he tested for the importance of both anticipated and unanticipated inflation in determining private consumption. Although such inflation effects have been suggested in the theoretical literature and were found to be present in some of the countries in Lahiri’s sample, they did not have a significant impact on saving in Singapore, possibly because Singapore experienced low inflation for most of the sample period.

In the present study, a consumption function incorporating the effects of growth and the changing age structure of the population was estimated for Singapore over the period 1970-92 (Appendix 7-1). Consistent with Lahiri’s findings for other Asian countries, the estimates obtained indicate that growth and demographics had an important effect on private saving, while inflation and terms of trade changes did not. The effects of two other potentially important determinants of private saving—contractual saving with the CPF and the saving of the public sector—were also estimated.

The estimated consumption function can be used to calculate the steady state average propensity to save (APS) of the private sector. Using annual averages in recent years for the steady state values of the main determinants of private saving—5 percent annual real per capita income growth and a working age population ratio of 70 percent—the APS in Singapore is estimated at 45 percent of private disposable income, or 34 percent of GNP. This contrasts with Lahiri’s estimates of the steady state zero-growth APS for other countries, which range from about 20 percent of private disposable income for Indonesia, Korea, the Philippines, Sri Lanka, and Thailand, to 30 percent for India and 40 percent for Malaysia. These steady state values would, of course, be higher if positive growth rates were assumed in the steady state.

The results indicate that, on average, a 1 percentage point increase in the rate of growth of private disposable income resulted in a 0.3 percentage point increase in the private saving rate in the long run. The changing age structure of the population also was found to affect the private saving rate significantly. An increase in the ratio of working age population to total population by 1 percentage point increased, on average, the saving rate by 1.7 percentage points.

Substitutability of Public and Private Saving

The extent to which public saving may have contributed to Singapore’s high saving rate is of particular importance, given the high rate of government saving. Analytical models such as that of Barro (1974) argue that, with Ricardian equivalence, any change in government saving, because it must be accompanied by a change in future taxation, is exactly offset by private sector dissaving. Hence, government saving is completely substitutable with private saving and has no impact on national saving. In practice, however, many of the assumptions required for Ricardian equivalence need not be valid, and a number of studies have focused on explanations for possible deviations from Ricardian equivalence and on empirical evidence on the impact of government budget variables on private saving.8 More recently, Asilis and Ghosh (1992) have noted that fiscal policy choices can shift an economy to a higher saving and growth path. In their model, the rate of return on an individual’s saving depends upon the aggregate level of saving, and a commitment by the Government to a high saving rate is likely to give rise to private sector expectations that overall saving will be high, thereby inducing private individuals to save more. Such a commitment could be made credible, they argue, by a sustained period of fiscal restraint. This effect may well have occurred in Singapore, but its importance, especially with respect to private sector expectations, is difficult to quantify.

The effect of government saving on the private saving rate in Singapore can be analyzed by examining the relationship between government spending and private consumption expenditure. As a first pass at quantifying this relationship, the fiscal current account surplus to GDP ratio was included as an explanatory variable in the estimation of the consumption function outlined above. These estimates indicate a statistically insignificant relationship between government saving and private consumption. While this implies that there is no private sector dissaving in response to an increase in the government saving rate, the result must be interpreted with caution. The Ricardian equivalence proposition indicates that expected future levels of government spending and taxation influence current private consumption. Current or lagged values of the fiscal surplus, then, are important only to the extent that they serve as useful predictors for future spending.

A more accurate measure of the substitutability of public and private consumption was estimated by adapting a framework developed by Aschauer (1985). Using restrictions placed on the data by the first-order necessary conditions for intertemporal optimization in consumption, a private sector consumption function that incorporates the effects of expected future government spending on private consumption decisions was estimated for Singapore (Appendix 7-2). The empirical evidence indicates a high degree of substitutability between government spending and private consumption. A dollar’s increase in government consumption was offset, on average, by a reduction in private consumption of 0.6 dollar.9 Consequently, holding other factors fixed, if the entire amount of public saving, averaging over 10 percent of GNP in recent years, were instead to be channeled into public consumption, the overall saving rate would decline by 4 percent of GNP.

CPF Saving

The high level of mandatory saving through the CPF may also have been a factor responsible for Singapore’s high overall private saving rate. Because withdrawals have been liberalized over time, however, saving with the CPF has likely become more substitutable with voluntary (non-CPF) private saving. The degree of substitutability, then, between CPF and voluntary private saving is important in determining how much the CPF saving scheme affects total saving. Since the degree of liberalization of withdrawal policies from the CPF is difficult to quantify, it is not possible to measure the change in the substitutability of CPF and non-CPF saving over time. Instead, the impact of CPF saving on total private saving over 1970-92 was estimated. The estimates probably understate the effect of CPF saving in the early part of the sample period and overstate it for the later part.

Whether participation in a compulsory saving scheme encourages or discourages other forms of saving has been much debated. Several effects that together determine the impact of compulsory saving on voluntary private saving have been recognized. These include a substitution effect, by which a saver regards his compulsory contributions as partly or fully substitutable for other forms of saving, and a forced saving effect, which would force an increase in saving of those who would otherwise not save. The net effect, however, can be positive or negative.

In a study of the effects of social security saving on noncompulsory private saving, Datta and Shome (1981) examined empirical evidence from five Asian countries—India, Malaysia, the Philippines, Singapore, and Sri Lanka—over the period 1960-74. They found no statistically significant effect of compulsory saving on noncompulsory saving, suggesting that increases in mandatory saving were fully reflected in increases in total private saving. In a study covering 54 industrial and developing countries, Kopits and Gotur (1980) examined the role of social security schemes in determining intercountry differences in saving propensities. For developing countries, Kopits and Gotur found the net effect of social security schemes was a “draw”—there was neither a positive nor an adverse impact on voluntary saving.

The estimates obtained in this study contrast with the earlier studies. The results indicate that changes in the ratio of CPF saving to private disposable income do not have a statistically significant impact on total private saving, suggesting that there is a full offset in voluntary (non-CPF) private saving resulting from increases in CPF saving. This result is likely due to the liberal guidelines on the use of CPF balances, including the purchase of residential property and stock market investments. These guidelines have made CPF saving highly substitutable with voluntary private saving. While changes in CPF saving may have affected overall private saving in earlier years when withdrawal facilities were more limited, such an effect is difficult to measure owing to the small sample period.

Appendix 7-1: Modeling of Private Consumption and Saving Behavior

This appendix discusses the methodology used in formulating and estimating a model of private consumption behavior. The main determinants of private saving are identified and their estimated impact is reported.

Model Specification

The precise functional form and dynamic structure of consumption functions has been much debated in the literature. Two general types of models stand out—models incorporating the life-cycle/ permanent income hypothesis and error correction models. The general specification adopted here, similar to the one proposed by Blinder and Deaton (1985), incorporates both types of models. The underlying long-run relationship reflects the life-cycle/ permanent income hypothesis. The estimation approach is in the tradition of error correction models that have proven particularly fruitful in generating estimates consistent with observed data patterns in industrial countries and, at the same time, embody the long-run relationships implied by economic theory.

Private consumer spending is modeled here as an aggregate, without separate specifications for services, nondurable goods, and durables. The basic form of the estimated consumption function is:

where c and y denote the natural logarithms of real per capita private consumption and private disposable income, respectively, W is the percentage of the population aged 15-64, π is the rate of inflation, π is the terms of trade, CPF is the ratio of CPF saving to private disposable income, and GS is public saving, expressed as a ratio to GDP. A Δ in front of a variable denotes a first difference. The posited linear combination of the percentage of the population that is of working age and the change in that percentage allows for an adjustment period of two years.10

The error correction model proposed by Davidson and others (1978) has the dynamic property that the ratio of consumption to income is constant in the long run. Any short-term divergence of this ratio from its long-run value is corrected over time. This amounts to the restriction:

and the functional form of consumption simplifies to:

In a noninflationary steady state with constant terms of trade, Wt = W, CPFt = CPF, GSt = G, Δct = Δyt = g, and ignoring the error term, the ratio of consumption to income, or the average propensity to consume (APC), may be expressed as:

As long as β1, < 0, 0 < β2 < 1, and <4 < 0, the long-run average propensity to save (APS), which is simply (1 - APC), rises with an acceleration in growth or with an increase in the working age population ratio. If, in addition, <8 < 0, an increase in the CPF saving to income ratio is associated with an increase in the overall APS. Finally, if β9 > 0, an increase in the steady state level of the public saving to GDP ratio lowers the long-run APS as consumers offset the additional public sector saving with private sector dissaving.

Data Issues and Estimation

As noted above, a breakdown of gross national saving into public and private saving is not available. Furthermore, a breakdown of private saving into saving of the household and corporate sectors is also unavailable. Hence, public saving is calculated as the central government’s consolidated current account surplus, and the remaining gross national saving is attributed to the private sector. CPF saving is measured as total contributions to the CPF during each calendar year plus accumulated interest on CPF balances minus withdrawals on account of retirement, death, or permanent emigration from Singapore. Voluntary (non-CPF) private saving, then, is just the difference between total private saving and CPF saving.

National income is derived by adding net transfers to GNP. Private disposable income may then be calculated as the sum of private consumption, which is obtained directly from the national accounts, and private saving. The government consumption series is also taken from national accounts data. All income and consumption variables are deflated by the GDP deflator. Inflation is measured as the percentage change in the consumer price index.

Equation (3) was estimated over the period 1970–92 using two-stage least squares to account for simultaneity amongst the variables. A dummy variable to capture the effect of the recession during 1985-86 on consumption was included in the estimated equation. A conventional F-test indicated that inflation and changes in the terms of trade were not important in determining changes in consumption patterns. This may be explained by the fact that except for the oil shock in 1973-74, Singapore experienced low inflation and stable terms of trade throughout the sample period.

In addition, the impact of government saving and CPF saving on private consumption was found to be statistically insignificant. The latter result indicates that consumers fully offset any changes in CPF saving by changes in voluntary (non-CPF) saving, probably because of the variety of purposes for which CPF saving may be used, particularly in recent years, which make it highly substitutable with voluntary saving. The insignificance of the government saving variable in explaining changes in consumption patterns suggests that the private sector does not adjust consumption in response to changes in public saving, and any increase in public saving is fully reflected by an increase in gross national saving. However, as noted above, this result must be interpreted with caution. According to the Ricardian equivalence proposition, private consumption expenditure is influenced by expected future levels of government spending. If current or lagged values of the fiscal surplus are not good predictors of future spending, the insignificance of contemporaneous public saving in explaining changes in private spending may not be meaningful.11

The results from estimating the consumption function, sequentially omitting all variables that were statistically insignificant, are presented in Table 7-3. It may be noted that in addition to inflation, the terms of trade, CPF saving, and public saving variables, the change in the working age population was also omitted.12

Table 7-3Two-Stage Least-Square Estimates of Consumption Function
Constant0.3579
(2.09)
Rate of growth of private disposable income0.8766
(4.66)
Lagged consumption to income ratio–0.2160
(1.80)
Percentage of population aged 15-64–0.0069
(2.10)
Dummy for 1985-86 recession0.0548
(2.49)
R20.6182
Standard error0.0195
Sum of squared residuals0.0069
Note: The dependent variable for each equation is the first difference of the natural logarithm of real per capita private consumption expenditure. The instruments used in the first stage were: contemporaneous working age population ratio, level, and change: lagged natural logarithms of real per capita private consumption and disposable income; lagged natural logarithm of real money balances; the first two lags of the CPF saving to income ratio; lagged per capita real overall fiscal balance of the consolidated central government; the first two lags of the public saving to GDP ratio; and a dummy variable for 1985-86. The sample period is 1970-92. The values in parentheses are t-statistics.
Note: The dependent variable for each equation is the first difference of the natural logarithm of real per capita private consumption expenditure. The instruments used in the first stage were: contemporaneous working age population ratio, level, and change: lagged natural logarithms of real per capita private consumption and disposable income; lagged natural logarithm of real money balances; the first two lags of the CPF saving to income ratio; lagged per capita real overall fiscal balance of the consolidated central government; the first two lags of the public saving to GDP ratio; and a dummy variable for 1985-86. The sample period is 1970-92. The values in parentheses are t-statistics.

The estimates obtained from the final version of the consumption function, when substituted into the long-run equation, indicate that the APS in Singapore was about 45 percent of disposable income.13 This calculation is based on assumed steady state values of per capita income growth and the working age population ratio equal to the average that prevailed during 1987-92.

An important determinant of Singapore’s saving rate was the demographic structure of the population. The results indicate that a 1 percentage point increase in the working age population ratio was associated with a 1.7 percentage point rise in the saving rate. Thus, the increase in the working age population ratio to 71 percent during 1987-92 from 60 percent during 1970-74 accounted for an increase of about 22 percentage points in the APS.

Growth was also found to be an important determinant of private saving. For every percentage point increase in real per capita income growth, the saving rate grew by 0.2 percentage point. Had there been no growth in real per capita terms instead of the annual average of 6.1 percent that was actually achieved during 1970-92, the APS would have been 2 percentage points lower.

Finally, public saving was found not to be important in determining the level of private saving. This result, however, is subject to qualification. First, expectations of future levels of public saving and consumption are likely to matter in current saving decisions. Contemporaneous public saving, therefore, is not the appropriate variable to include on the right-hand side. Second, changes in government saving reflect not only changes in public consumption but also changes in government tax revenue. The latter affects private disposable income and will affect both private consumption and income. Its impact on the ratio of consumption to income, however, is unclear a priori. For these reasons, an alternate method of estimating the effect of changes in government consumption on the pattern of private consumption expenditure is taken up in Appendix 7-2.

Appendix 7-2: Substitutability of Public and Private Consumption

The effects of fiscal policy on private consumption and aggregate demand have aroused a great deal of interest since Barro’s (1974) seminal paper on Ricardian equivalence. The basic proposition, that the choice of tax versus debt financing of government expenditure is irrelevant to the determination of the level of aggregate demand, has been much investigated and debated. In this appendix, a framework for testing this proposition using the experience in Singapore is developed and estimated.

Analytical Framework

The framework employed here is an extension of the model developed by Aschauer (1985) to estimate the substitutability of public and private consumption expenditure in the United States. Assuming a quadratic utility function for private agents, defined over current and future consumption and ignoring the effects of government spending, the first-order condition for expected utility maximization implies:

where C is real per capita consumption. If, in addition, government spending substitutes for private consumption in utility, then equation (1) may be rewritten as:

where G is real per capita current expenditure of the government. The parameter Ά measures the substitutability of public and private consumption or, alternatively, the degree to which government spending substitutes for private spending.

In estimating the substitutability of public and private consumption, equation (2) may be transformed to the following estimable equation:

where D is the real per capita central government overall fiscal balance and u is the stochastic disturbance term. Since contemporaneous government consumption is determined simultaneously with contemporaneous private spending, instrumental variables must be employed in the estimation. From equation (2), the parameter restrictions

can be imposed. Moreover, the joint hypothesis of rational expectations and Ricardian equivalence implies that past values of government deficits should not have explanatory power for private consumption apart from their role in forecasting contemporaneous public spending. This yields the restriction:

In order to avoid problems of nonstationarity, equation (3) was estimated in first differences.

Estimation Results

Two versions of the model contained in equation (3) were estimated. The first version of the model, consistent with the null hypothesis of rational expectations/Ricardian equivalence, was estimated with the constraints in equations (4) and (5) using an iterative nonlinear two-stage least-squares procedure. The results of this constrained estimation, reported in Table 7-4, indicate that a 1.0 dollar rise in public spending is associated with a reduction in private consumption of 0.6 dollar. While this suggests that increases in government consumption could have an expansionary effect on aggregate demand, the point estimate for the consumption substitutability parameter is not significantly different from unity.14 Consequently, the hypothesis of fiscal neutrality cannot be rejected.

Table 7-4Iterative Nonlinear Two-Stage Least-Square Estimates of Consumption Function
ConstrainedUnconstrained
α185.7948232.0711
(2.60)(3.01)
β0.3857· · ·
(1.95)
θ0.6128· · ·
(2.46)
μ10.385710.3398
(1.70)
μ10.23631-0.1321
(0.42)
μ3-0.61281-0.7801
(2.67)
μ4· · ·0.0208
(0.20)
μ5· · ·0.0460
(0.48)
R20.330.38
Standard error156.31164.30
Log likelihood-140.74-139,95
Note: The dependent variable is the first difference of real per capita private consumption expenditure. The instruments used were; first and second lags of the first differences of real per capita private consumption, government consumption, government savings, and overall fiscal balance of the consolidated central government. The equations were estimated over the period 1971-92. The values in parentheses are t-statistics.

Coefficients obtained by substitution of constrained coefficient estimates into the set of restrictions in expression (4).

Note: The dependent variable is the first difference of real per capita private consumption expenditure. The instruments used were; first and second lags of the first differences of real per capita private consumption, government consumption, government savings, and overall fiscal balance of the consolidated central government. The equations were estimated over the period 1971-92. The values in parentheses are t-statistics.

Coefficients obtained by substitution of constrained coefficient estimates into the set of restrictions in expression (4).

In addition, a second unconstrained version of equation (3) was estimated, the results of which are also presented in Table 7-4. The table indicates that the estimated coefficients from the two versions are quite similar. A log-likelihood ratio test, used to test the null hypothesis, yields a test statistic of 1.58, which may be compared with its critical value of 6.25. Consequently, the joint null hypothesis of rational expectations and Ricardian equivalence cannot be rejected.15

The estimated coefficients may be used to calculate the effect of public saving on private saving. Over the period 1987-92, government saving averaged about 10 percent of GNP, while government income (current revenue net of subsidies to the private sector) averaged almost 21 percent of GNP. If, instead of saving almost half of its net revenue, the Government had consumed the entire amount, private consumption would have been 6 percent of GNP lower. Consequently, private saving would have been higher by the same amount, and gross national saving would have been 4 percentage points of GNP lower than was actually achieved.

Lower contribution rates apply to pensionable public sector employees and, since 1988, to employees aged over 55. Contributions are capped at a monthly salary level of S$6,000.

CPF savings are calculated as contributions from employees and employers plus accrued interest minus withdrawals on account of retirement, death, medical and educational expenses, or permanent emigration from Singapore.

The overall home ownership rate in 1990 was 90 percent—92 percent in publicly built housing and 87 percent in private housing.

In 1993, the basic investment scheme was revised to permit larger withdrawals and to permit 10 percent of eligible funds to be invested in nontrustee stocks. In addition, an enhanced investment scheme was created, allowing withdrawal of 80 percent of balances in excess of S$50,000 for investment in a wider range of instruments.

Preliminary estimates are that withdrawals rose to S$11 billion in 1993 (including S$4 billion for the Telecom float), representing 17.5 percent of members’ balances.

International comparisons of saving rates have to be interpreted with caution, because national income statistics are not directly comparable across countries owing to differences in accounting practices. As noted by Hayashi (1989), a significant portion of the difference between the saving rates of the United States and Japan is removed once common accounting practices are applied to the national accounts of both countries.

Hayashi (1989) associates Japan’s low stock of wealth after World War II with high saving rates and a rapid accumulation of wealth during the 1970s and early 1980s.

Leiderman and Blejer (1988) provide a comprehensive survey of these studies.

In his study for the United States, using quarterly data from the period 1948:Q1 to 1981:Q4, Aschauer (1985) estimated the substitutability parameter at 0.2–0.4.

The Blinder and Deaton specification also includes wealth variables. However, since measures of wealth are not available for Singapore, they are not included here. The resulting specification is similar to that used in earlier studies, such as Lahiri (1989) and Bovenberg and Evans (1990).

An alternative and more accurate measure of public and private consumption was also estimated and is described in the following appendix.

Diagnostic checks of the estimated equations showed no evidence of serial correlation of the errors.

Based on the average ratio of private disposable income to GNP in recent years, this estimate of the APS implies steady state private saving of about 34 percent of GNP

The estimated substitutability is somewhat higher than the results obtained by Aschauer (1985), who finds that the substitutability of public and private consumption in the United States over the period 1948–81 was in the range of 0.2–0.4.

Both versions of the estimated equation were checked for serial correlation. No evidence of first order autocorrelation was found in either version.

    Other Resources Citing This Publication