Abstract
Changes in saving behavior brought about by financial liberalization in the Nordic countries are examined. Although this paper suggests that there have been changes in the economic relationships in Denmark and Finland, the evidence from Norway is less conclusive, and the data do not reveal any structural change in Sweden. The results also indicate that the wealth effect has played an increasing role in household saving decisions. Since the present tax rules are biased against saving, cuts in marginal tax rates and reductions in the tax value of interest payments would be expected to increase household saving.
This paper examines household saving and consumption behavior in the Nordic countries.1 The research is motivated by the sharp drop in household saving rates that took place in all Nordic countries between 1984 and 1987. This decline, which is associated with the widening of the Nordic countries’ external current account deficits, has led the national authorities to search for reasons for the weak saving behavior and has also been partly responsible for the introduction of tax reforms aimed at enhancing private saving.
Aggregate consumption and saving behavior is one of the most intensely researched areas in macroeconomics. Nevertheless, the empirical literature on the effects of a number of economic and institutional factors is inconclusive, despite several efforts to model saving behavior. In the formulation of a consumption-saving model for the Nordic countries, developments in saving behavior need to be viewed against the background of high marginal income tax rates, the tax deductibility of interest payments, and, more recently, the liberalization of financial markets. For many years, real after-tax interest rates were negative for a large majority of households and have become generally positive only recently. In the past, the effects of negative real after-tax interest rates on consumer borrowing were mitigated by financial market regulation, which limited an individual’s ability to borrow through credit rationing. This limitation implied that consumption could be shifted into the future through saving, but it could not be freely moved forward through borrowing. Therefore, consumption was more sensitive to income than it would have been under perfect capital markets. Several empirical studies have confirmed that credit rationing had a negative influence on consumer expenditure in the Nordic countries.2 The subsequent deregulation of credit markets relaxed the borrowing constraints on households, causing many to adjust their financial portfolios.
A complete empirical assessment of the effects of financial liberalization on consumption and saving decisions in the Nordic countries is precluded by the small number of observations available since the financial deregulation took place. However, it appears that many of the consumption-function relationships that are used in economic modeling have performed poorly in the post-deregulation period. This paper attempts to determine whether the structural changes brought about by the various financial reforms were so large that consumption models using data from the period prior to the liberalization cannot be used to predict accurately the developments in the post-sample period. In addition, the factors that may have led to predictive failure are analyzed and compared on a cross-country basis for the Nordic countries.
The analytical framework is based on the life-cycle hypothesis, which implies that individuals tend to save more during their years of high earnings and to dissave during their years of low earnings; that is, the working population tends to be savers, and retirees and the young are dissavers.3 The forward-looking characteristic of the life-cycle model and the considerable body of evidence supporting the hypothesis have made it a standard starting point for consumption analysis, even though the principal implication of the model—namely, the separation of income and consumption profiles—is open to some question. The paper is organized as follows. Section I reviews recent financial reforms in the Nordic countries and discusses developments in saving rates. In Section II, the basic theoretical framework is derived, which is then used empirically in Section III. Section IV presents the conclusions.
I. Financial Deregulation and Developments in the Saving Rate
The institutional characteristics of financial markets have to some extent differed among the Nordic countries. Until recently, Finland, Norway, and Sweden had broad-based credit controls combined with low interest rate policies, whereas Denmark has experienced relatively high domestic interest rates, compared with many other countries. Credit rationing was perhaps the most extensive in Finland and Norway. Nevertheless, recent developments in the household saving rate have been strikingly similar in all the Nordic countries, with saving ratios declining sharply throughout. This decline has been associated with greater access by households to credit because of financial deregulation and the removal of capital controls.
Financial Reform
Denmark has traditionally had fewer credit controls than the other Nordic countries. In the 1970s the main instruments used by the monetary authorities to manage domestic liquidity were direct controls on bank lending and regulation of the terms of bank borrowing at the Danmarks Nationalbank, In March 1979 bank lending rates, which were traditionally linked to the discount rate, were deregulated.4 In 1980 the ceilings on bank lending were abolished and replaced by a system of overall guidelines. In 1985 the guidelines were dropped when it became apparent that they were being circumvented by the banks.
In the Danish case, the removal of exchange controls was perhaps the most important institutional factor affecting financial decision making in the late 1970s and early 1980s, In 1978 Danish residents were allowed to buy exchange-listed bonds issued by international organizations. In 1983 residents were permitted to acquire freely bonds listed on foreign exchanges with a maturity exceeding two years, and nonresidents were allowed to buy all types of Danish bonds and shares.5 In addition, the restriction that linked company foreign borrowing to fixed investment was removed in 1983, and the required minimum maturity of such loans was reduced from five years to one year in 1985. Furthermore, in 1984 Danish residents were allowed to buy shares listed in foreign exchanges and Danish bonds denominated in foreign currency. In October 1988 residents were allowed freely to deposit and borrow abroad in foreign currency.
In Finland, Norway, and Sweden, deregulation advanced rapidly in domestic financial markets, while the deregulation of exchange control was relatively slow. The liberalization process had several common features in these countries, but the timing was somewhat different.
In Finland the financial markets were tightly controlled, with rigid lending rates and an extensive use of foreign exchange regulations, which resulted in credit rationing (see Abrams (1989)). The Bank of Finland began gradually opening the domestic banking system in 1982, after a gray market, which had begun to develop in the late 1970s, had reduced the effectiveness of both monetary policy and credit restrictions.6 Perhaps the most important elements in the Finnish financial reform were the elimination of interest rate ceilings and the creation of an active money market. Average bank lending rates were subject to ceilings until May 1983, when control was partially relaxed, allowing banks to pass on part of the costs of their unregulated funding to their lending rates. In 1986 the ceilings on average lending rates were abolished. At the same time, the Bank of Finland assisted in the development of the domestic money market, and in late 1986 the Bank of Finland and the commercial banks worked together to develop a domestic certificate of deposit (CD) market.
In Norway and Sweden, prior to 1984, most interest rates were regulated and credit growth was limited through lending ceilings. As in Finland, the system of direct credit controls resulted in an inefficient allocation of credit. In Norway ceilings on bank lending were repealed in 1984, and the regulation of bank interest rates was discontinued in September 1985. Since then, monetary policy has been largely implemented through market-oriented instruments that indirectly influence interest rates in the money and capital markets. In Sweden regulation of bank lending rates was repealed in May 1985, and quantitative restrictions on lending were removed in November 1985. Unlike Finland, an efficient money market had already developed in Sweden by 1982–83. Treasury discount notes for financing the government borrowing requirement were introduced in 1982, and a market for business and local government certificates developed in 1983.
Developments in the Saving Rate
Although the financial liberalization in the Nordic countries increased the efficiency of the use of credit and improved the transparency of the functioning of the markets, serious problems emerged as a result of a surge in the growth of money and credit. The deregulation of financial markets opened new opportunities for household borrowing, leading to a surge in household demand for credit, as pent-up demands were freed and households sought to rearrange their portfolios. Although the private saving rate fluctuated in all Nordic countries in the 1970s, there was no clear trend until the 1980s.7 The drop in the private saving rate was most pronounced in Denmark and Norway, declining from 13.3 percent in 1984 to 4.1 percent in 1986 in Denmark, and from 9.5 percent in 1984 to –1.8 percent in 1986 in Norway. The drop in private saving was mainly due to reduced household saving.8 In the same period, the household saving rate in Norway dropped from 5.2 percent to –6.1 percent. The household saving rate started to decline in 1984 in Finland, and in 1985 in Sweden. The fall in the household saving rate has sparked much discussion among policymakers in all Nordic countries because of its implications for current account developments and the accumulation of external debt.
Households used a part of the borrowed funds to purchase real and financial assets, which resulted in an increase in demand for equities. In Norway industrial share prices increased by 32 percent in 1985 (period average), and in Finland and Sweden share prices advanced by more than 50 percent in 1986. In addition, since deregulation reduced the saving requirements for housing purchases, demand for housing increased. This resulted in rapid increases in housing prices, particularly in Finland and Norway, where pent-up demand for housing was strongest. In Norway, for instance, urban housing prices increased by some 80 percent from the first quarter of 1984 to the first quarter of 1987.
The surge in household borrowing was facilitated by the tax deductibility of interest payments, coupled with high marginal personal tax rates.9 In Denmark, even with the 1985 tax reform, the tax value of interest deductions was only reduced to 51–57 percent.10 In Finland interest payments on housing loans are now deductible up to Fmk 25,000, and on consumer loans, up to Fmk 10,000.11 In Norway the tax reform has reduced the tax value of deductions by lowering the top marginal tax rate for interest deductions from over 66 percent in 1986 to some 48 percent in 1988, In Sweden household interest payments remain fully deductible. However, the reform that was implemented in 1983–85 reduced the tax value of interest payments to 50 percent for many households.12
In the first half of the 1980s, real after-tax interest rates were highly negative in the Nordic countries as a result of the deductibility of interest payments and high inflation. The increased access to credit thus boosted consumer debt. Furthermore, higher housing and other real asset prices appear to have increased consumer expenditure not only through increased collateral values but also through wealth effects.
A number of factors, other than the deregulation of financial markets and after-tax interest rates, also influence household saving decisions. Traditionally, the major motives for saving have been for retirement, bequests, or for some specific target, and precautionary saving.13 Rather than trying to examine all the potential determinants of saving in the Nordic countries, this paper attempts to determine whether the deregulation in the 1980s led to a fundamental “structural break,” or whether the predictive failure of traditional consumption functions was attributable to a misspeciftcation of the econometric formulation. In this way, the impact of the deregulation on saving behavior can be assessed. Common factors that contributed to the decline in the saving rate in the Nordic countries are also examined.
II. The Model
The saving rate can fluctuate considerably in the short run, while remaining relatively stable in the long run. Ideally, any consumption-saving model should capture both the short-run and long-run properties of the consumption-income relationship. If consumption (or saving) functions that include only differenced variables are used, all the information derived from economic theory on long-run properties would have to be forgone. However, Davidson and others (1978), and Hendry and von Ungern-Sternberg (1981) have introduced models that ensure that the dynamic equation has appropriate long-run, steady-state properties postulated by economic theory, while allowing for short-run divergences from these properties.
The analytical framework of this paper is based on the standard life-cycle hypothesis,14 which implies that the ratio of saving to income and the ratio of wealth to income remain constant on any given income growth path, but vary with the growth rate. The basic life-cycle theory suggests that an individual’s consumption depends on current and expected future labor income and current wealth.
Long-run properties suggested by the life-cycle theory are consistent under some parameter restrictions, with the basic functional form of the consumption function suggested by Blinder and Deaton (1985),
where c and y denote the natural logarithms of real consumption and real disposable income, respectively. The q variables denote contemporaneously dated variables and the zt-1 variables denote lagged values of q or other variables.
An attractive feature of the functional form (1) is that it accommodates specifications such as the error-correction model recommended by Davidson and others (1978) and the model proposed by Hendry and von Ungern-Sternberg (1981). These models reproduce steady-state conditions that are consistent with the life-cycle theory.
A simple-error-correction model takes the following form:15
The steady-state solution of equation (2) for Δct = Δyt = g is
and hence
Thus, the long-run consumption income ratio (saving ratio) is increasing (decreasing) with the growth rate g as long as β1 > 1, β2 ≤ 1.
Equation (2) implies that consumers adjust consumption in response to short-run changes in income, as well as to previous disequilibria (ct-1 – yt-1), which can be interpreted as a feedback response to obtain a desired long-run condition. The model also has a unitary elasticity of consumption to income at growth rate g. Moreover, if g varies, the observed C/Y ratio may have trendlike movements, but this does not rule out long-run unit elasticity for any constant growth rate.
Hendry and von Ungern-Sternberg (1981) extended the simple-error-correction model to include a so-called integral-control mechanism.16
The model takes the form17
where θ0 = β0, θ1 = β2, θ2 = β1, θ3 = β4, and a = log A, where A is real net worth. In other words, in equation (5) consumers adjust their expenditure not only because of changes in income but also by reacting to “disequilibrium” conditions to ensure constant steady-state equilibrium ratios of C to Y and A to Y.
Although the models discussed above are consistent with long-term equilibrium (steady-state) conditions implied by the life-cycle hypothesis and allow short-run changes in income to affect consumption expenditure, there are other potential determinants of saving and consumption that are short term or cyclical in nature. From the point of view of this study, the obvious shortcoming of these correction models is that they do not allow short-run changes in wealth to affect consumption expenditure. As described above, the deregulation of financial markets in the Nordic countries resulted in adjustments of households’ portfolios, which drove up relative asset prices. These adjustments are assumed to have affected consumption through the wealth effect and the easier access to credit brought about by the increased value of households’ collateral. To capture the short-run wealth effect, the change in real wealth was also included in the model of consumption in the Nordic countries.
Inflation may also play an important role in saving behavior, as suggested by Davidson and others (1978), Deaton (1977), and Koskela and Virén (1985). Inflation can influence saving-consumption behavior in a variety of ways, but perhaps the most important is its effect on financial wealth. If households have a target (constant) income-wealth ratio, unexpected and/or accelerating inflation would reduce the real value of financial assets and therefore increase households’ saving, so as to reestablish the target ratio. In addition, as Hendry and von Ungern-Sternberg (1981) have noted, inflation tends to distort the conventional measure of personal disposable income in the national accounts, since capital losses on monetary assets generated by inflation are not deducted from net interest receipts included in persona! disposable income. Therefore, in an inflationary period, the “real income” perceived by consumers may be lower than that reported in the national accounts and often used in empirical studies. For these reasons, and because the rate of inflation may also capture the effect of income uncertainty on consumption, inflation was also included in the model of consumption in the Nordic countries. These modifications do not, however, influence the steady-state properties of the model. Taking into account these factors, the modified consumption function along the lines of Davidson and others (1978) and Hendry and von Ungern-Sternberg (1981) can be written as
where θ0 = β0, θ1 = β2, θ2 = β1 and θ3 = β4 = η; p = log P, a = log A, P is the consumer price index, and A is real net wealth.
III. Empirical Results
In this section, the factors that may have contributed to the decline in private saving in the Nordic countries are analyzed empirically. The models were estimated for each country, using annual data running from 1971 to the year that the country introduced its major deregulation measures.18 This procedure left a few observations that could be used to evaluate the performance of the consumption functions outside the sample period.19
Description of Data
It was not possible to obtain comparable data for all Nordic countries. First, Danish statistics did not permit the separation of household and business saving. As a result, private disposable income was used as a proxy for personal income for Denmark.20 Second, it was not possible to construct a wealth variable that would be consistent with the life-cycle hypothesis and identical for all Nordic countries. Therefore, different proxies had to be used for household wealth in the calculations.
Although in the models, Ct stands for consumption rather than consumer expenditure, consumer expenditure was used as a proxy for consumption in all countries. For Denmark, private nominal disposable income was deflated by the consumer price index. For Finland and Norway, household disposable income was deflated by the consumer price index.21 Since Swedish statistics do not distinguish between consumption by households and nonprofit institutions, real private consumption expenditure was used as a proxy for household consumption, and the disposable income of households and nonprofit institutions, deflated by the consumer price index, was used as a proxy for real household income. For all countries, inflation was measured as the change in the consumer price index.
Due to a lack of data, the wealth variable was constructed differently for different countries. For Denmark, nominal wealth was measured as the sum of net financial assets of the private nonbanking sector including assets of pension funds and life insurance companies, the housing stock, and the stock of personal transport equipment. Real wealth was calculated by deflating nominal wealth by the consumer price index. The real capital equipment of firms was not included in the wealth variable due to a lack of data. Nominal wealth was measured at the beginning of the year and valued at the previous year’s prices (see Heinesen (1987)). Since residential housing constitutes the major part of household wealth, relative housing prices measured as the housing price index, divided by the consumer price index, were used as a proxy for household wealth for Finland and Norway. Finally, for Sweden real wealth was proxied by household net financial wealth deflated by the consumer price index (see Berg (1988)).
Estimation Results
In this section, the factors that may have contributed to the decline in private saving in the Nordic countries are analyzed empirically. Of crucial importance is whether the sharp increase in consumption can be explained through the models discussed above, or whether saving behavior changed because of the structural changes in the financial markets.
Ideally, one would like to start a sequential testing procedure from the basic equation (1), with variables q and z consisting of present and lagged values of consumer price index and real household wealth, respectively. However, the small number of observations would have resulted in an insufficient number of degrees of freedom for some desirable diagnostic tests. Nevertheless, to shed some light on the usefulness of the error correction and the Hendry and von Ungern-Sternberg model for the Nordic countries, appropriate unrestricted versions of equation (1) were estimated using ordinary least squares, and these versions were tested against the constrained versions. In the cases of Denmark, Finland, and Norway, a simple-error-correction model could not be rejected at the 5 percent significance level. The hypothesis that the Hendry and von Ungern-Sternberg model holds could not be rejected in the cases of Denmark, Finland, and Sweden at the 5 percent significance level, and at the 1 percent level in the case of Norway.
The estimated results of the Hendry and von Ungern-Sternberg model, modified to allow for short-run effects of wealth and inflation (equation (6)), are presented in Table 1. In the cases of Denmark, Finland, and Sweden, the coefficients for the rate of growth of real disposable income and the lagged consumption income ratio were significant. The low and insignificant value of the coefficient for the income variable in the case of Norway could be because of the small sample problem or overfitting. The short-run elasticity of consumption to income was highest in Denmark and Finland. In all cases, the coefficients for the lagged consumption income and lagged wealth income ratios had the correct sign. Moreover, in the case of Sweden the coefficient for the change in real wealth was clearly insignificant with the wrong sign. The coefficient for inflation, however, implied a positive effect on saving. This may be an indication of the multicollinearity problem between inflation and the change in real wealth. None of the equations in Table 1 appears to suffer from first-order autocorrelation.22
Empirical Estimates of Equation (6)
Empirical Estimates of Equation (6)
| Variable | Denmark (Sample size 1972–83) | Finland (Sample size 1971–85) | Norway (Sample size 1971–84) | Sweden (Sample size 1971–85) | |
|---|---|---|---|---|---|
| Constant | -0.1989 | 0.1172 | 0.4595 | 0.0325 | |
| (-6.36) | (1.47) | (2.23) | (1.88) | ||
| Rate of growth or real disposable | 0.8649 | 0.8387 | 0.2940 | 0.6025 | |
| income | (8.41) | (7.22) | (1.20) | (3.15) | |
| Lagged consumption income ratio | -0.6453 | -1.0689 | -0.7640 | -0.6806 | |
| (-7.52) | (-3.33) | (-2.60) | (-2.81) | ||
| Lagged wealth income ratio | 0.0384 | 0.0396 | 0.0961 | 0.0230 | |
| (1.76) | (2.12) | (2.22) | (1.41) | ||
| Change in real wealth | 0.1789 | 0.0756 | 0.3253 | -0.0259 | |
| (4.26) | (1.70) | (1.77) | (-1.13) | ||
| Rate of inflation | 0.1896 | -0.0409 | 0.2145 | 0.2007 | |
| (1.15) | (-0.49) | (0.86) | (-0.89) | ||
| Summary statistics | |||||
| R2 | 0.9774 | 0.9286 | 0.7043 | 0.7743 | |
| Standard error | 0.0066 | 0.0093 | 0.0143 | 0.0116 | |
| Sum of squared residuals | 0.0003 | 0.0008 | 0.0016 | 0.0012 | |
| F | 51.88 (4.39) | 23.43 (3.48) | 3.81 (3.69) | 6.17 (3.48) | |
| D-W | 2.14 | 1.65 | 2.14 | 1.89 | |
| Number of observations | 12.0 | 15.0 | 14.0 | 15.0 | |
| Number of observations multiplied by the sum of the square of the first five autocorrelations | 10.76 | 3.65 | 3.09 | 4.23 | |
Empirical Estimates of Equation (6)
| Variable | Denmark (Sample size 1972–83) | Finland (Sample size 1971–85) | Norway (Sample size 1971–84) | Sweden (Sample size 1971–85) | |
|---|---|---|---|---|---|
| Constant | -0.1989 | 0.1172 | 0.4595 | 0.0325 | |
| (-6.36) | (1.47) | (2.23) | (1.88) | ||
| Rate of growth or real disposable | 0.8649 | 0.8387 | 0.2940 | 0.6025 | |
| income | (8.41) | (7.22) | (1.20) | (3.15) | |
| Lagged consumption income ratio | -0.6453 | -1.0689 | -0.7640 | -0.6806 | |
| (-7.52) | (-3.33) | (-2.60) | (-2.81) | ||
| Lagged wealth income ratio | 0.0384 | 0.0396 | 0.0961 | 0.0230 | |
| (1.76) | (2.12) | (2.22) | (1.41) | ||
| Change in real wealth | 0.1789 | 0.0756 | 0.3253 | -0.0259 | |
| (4.26) | (1.70) | (1.77) | (-1.13) | ||
| Rate of inflation | 0.1896 | -0.0409 | 0.2145 | 0.2007 | |
| (1.15) | (-0.49) | (0.86) | (-0.89) | ||
| Summary statistics | |||||
| R2 | 0.9774 | 0.9286 | 0.7043 | 0.7743 | |
| Standard error | 0.0066 | 0.0093 | 0.0143 | 0.0116 | |
| Sum of squared residuals | 0.0003 | 0.0008 | 0.0016 | 0.0012 | |
| F | 51.88 (4.39) | 23.43 (3.48) | 3.81 (3.69) | 6.17 (3.48) | |
| D-W | 2.14 | 1.65 | 2.14 | 1.89 | |
| Number of observations | 12.0 | 15.0 | 14.0 | 15.0 | |
| Number of observations multiplied by the sum of the square of the first five autocorrelations | 10.76 | 3.65 | 3.09 | 4.23 | |
In order to find a parsimonious model for saving and consumption behavior, a simplification search was conducted. The results are presented in Table 2. The simple-error-correction model with the wealth variable was selected for Denmark and Norway. For Norway, this specification could not be rejected at the 1 percent significance level, and it compared well with other versions. The simple-error-correction model with the inflation variable was selected for Sweden. For Finland, the Hendry and von Ungern-Sternberg model with the wealth variable outperformed all the other models.
Out-of-sample properties of the models in Table 2 were then explored.23 The results are presented in Figure 1. In most cases, the models underestimated the actual rise in consumption. The predictive accuracy was particularly poor for Finland and Norway. In both cases, the models did not predict the direction of the subsequent changes in consumption. In the case of Denmark the model first overpredicted and then under-predicted actual developments. Only in the case of Sweden did the consumption models predict developments in consumption with a high degree of accuracy.
The predictive failure of the equations for Finland and Norway suggests that there may have been a structural change in the economy.24 The Chow test for the null hypothesis of “no structural change in any parameter between the sample and the forecast” was rejected for Finland and Denmark.
All in all, the results support the argument that it may have been difficult, a priori, to assess the effects of the financial reforms on consumption in the Nordic countries. In particular, in Denmark and Finland, and to some extent in Norway, there appear to have been changes in the economic relationships. The evidence is perhaps most clear-cut in Finland, possibly because credit rationing had traditionally been relatively stringent. Only for Sweden do the findings not support the view that structural changes have taken place since deregulation. This finding may perhaps be explained by the fact that a gray market had already developed in the second half of the 1970s in Sweden.
Empirical Estimates of Consumption Function
Durbin’s h -statistic.
Empirical Estimates of Consumption Function
| Variable | Denmark (Sample size 1972–83) | Finland (Sample size 1971–85) | Norway (Sample size 1971–84) | Sweden (Sample size 1971–85) | |
|---|---|---|---|---|---|
| Constant | -0.1643 | 0.1033 | -0.0049 | 0.0222 | |
| (-7.29) | (1.45) | (-0.33) | (1.27) | ||
| Rate of growth of real disposable | 0.7191 | 0.8562 | 0.6020 | 0.7244 | |
| income | (13.35) | (8.07) | (2.75) | (3.86) | |
| Lagged consumption income ratio | -0.6515 | -1.0981 | -0.3833 | -0.4926 | |
| (-7.09) | (-3.62) | (-1.50) | (-2.26) | ||
| Lagged wealth income ratio | … | 0.0379 | … | … | |
| (2.15) | |||||
| Change in real wealth | 0.1172 | 0.0771 | 0.1227 | … | |
| (4.53) | (1.81) | (0.76) | |||
| Rate of inflation | … | … | … | -0.3812 | |
| (-1.94) | |||||
| Summary statistics | |||||
| R2 | 0.9649 | 0.9268 | 0.5092 | 0.6849 | |
| Standard error | 0.0071 | 0.0089 | 0.0165 | 0.0124 | |
| Sum of squared residuals | 0.0004 | 0.0008 | 0.0027 | 0.0016 | |
| F | 73.29 (4.07) | 31.63 (3.48) | 3.46 (3.71) | 7.97 (3.59) | |
| LM | 0.804a | 0.39 (5.12) | 1.29 (5.12) | 0.18 (4.96) | |
| NORM (2) | 0.45 (5.99) | 0.34 (5.99) | 4.17 (5.99) | 0.17 (5.99) | |
| ARCH (1) | – | 1.93 (5.32) | – (5.32) | 1.92 (5.12) | |
| Number of observations | 12.0 | 15.0 | 14.0 | 15.0 | |
| Number of observations multiplied by the sum of the square of the first five autocorrelations | 8.03 | 3.90 | 1.25 | 6.83 | |
| CHOW | 13.37 | 6.23 | 1.37 | 0.3 | |
Durbin’s h -statistic.
Empirical Estimates of Consumption Function
| Variable | Denmark (Sample size 1972–83) | Finland (Sample size 1971–85) | Norway (Sample size 1971–84) | Sweden (Sample size 1971–85) | |
|---|---|---|---|---|---|
| Constant | -0.1643 | 0.1033 | -0.0049 | 0.0222 | |
| (-7.29) | (1.45) | (-0.33) | (1.27) | ||
| Rate of growth of real disposable | 0.7191 | 0.8562 | 0.6020 | 0.7244 | |
| income | (13.35) | (8.07) | (2.75) | (3.86) | |
| Lagged consumption income ratio | -0.6515 | -1.0981 | -0.3833 | -0.4926 | |
| (-7.09) | (-3.62) | (-1.50) | (-2.26) | ||
| Lagged wealth income ratio | … | 0.0379 | … | … | |
| (2.15) | |||||
| Change in real wealth | 0.1172 | 0.0771 | 0.1227 | … | |
| (4.53) | (1.81) | (0.76) | |||
| Rate of inflation | … | … | … | -0.3812 | |
| (-1.94) | |||||
| Summary statistics | |||||
| R2 | 0.9649 | 0.9268 | 0.5092 | 0.6849 | |
| Standard error | 0.0071 | 0.0089 | 0.0165 | 0.0124 | |
| Sum of squared residuals | 0.0004 | 0.0008 | 0.0027 | 0.0016 | |
| F | 73.29 (4.07) | 31.63 (3.48) | 3.46 (3.71) | 7.97 (3.59) | |
| LM | 0.804a | 0.39 (5.12) | 1.29 (5.12) | 0.18 (4.96) | |
| NORM (2) | 0.45 (5.99) | 0.34 (5.99) | 4.17 (5.99) | 0.17 (5.99) | |
| ARCH (1) | – | 1.93 (5.32) | – (5.32) | 1.92 (5.12) | |
| Number of observations | 12.0 | 15.0 | 14.0 | 15.0 | |
| Number of observations multiplied by the sum of the square of the first five autocorrelations | 8.03 | 3.90 | 1.25 | 6.83 | |
| CHOW | 13.37 | 6.23 | 1.37 | 0.3 | |
Durbin’s h -statistic.
The Predictive Accuracy of the Models
Citation: IMF Staff Papers 1990, 001; 10.5089/9781451956863.024.A002
Source: Staff calculations.aIn log changes.The Predictive Accuracy of the Models
Citation: IMF Staff Papers 1990, 001; 10.5089/9781451956863.024.A002
Source: Staff calculations.aIn log changes.The Predictive Accuracy of the Models
Citation: IMF Staff Papers 1990, 001; 10.5089/9781451956863.024.A002
Source: Staff calculations.aIn log changes.The Predictive Accuracy of the Models
Citation: IMF Staff Papers 1990, 001; 10.5089/9781451956863.024.A002
Source: Staff calculations.aIn log changes.The Predictive Accuracy of the Models
Citation: IMF Staff Papers 1990, 001; 10.5089/9781451956863.024.A002
Source: Staff calculations.aIn log changes.Estimation Results: Full Sample
Estimation Results: Full Sample
| Variable | Denmark (1) (1972–1987) | Finland (2) (1971–1987) | Norway (3) (1971–1987) | Sweden (4) (1971–1987) | |
|---|---|---|---|---|---|
| Constant | -0.0989 | 0.0350 | -0.0054 | 0.0209 | |
| (-2.76) | (0.38) | (-0.61) | (1.56) | ||
| Rate of growth of real disposable | 0.5967 | 0.8016 | 0.4976 | 0.7209 | |
| income | (5.22) | (5.59) | (2.51) | (4.18) | |
| Lagged consumption income ratio | -0.4026 | -0.5640 | -0.5044 | -0.4767 | |
| (-2.70) | (-1.86) | (-3.95) | (-2.56) | ||
| Lagged wealth income ratio | … | 0.042 | … | … | |
| (0.67) | |||||
| Change in real wealth | 0.1373 | 0.0863 | 0.3676 | … | |
| (2.65) | (1.49) | (4.89) | |||
| Rate of inflation | … | … | … | -0.3621 | |
| (-2.65) | |||||
| Summary statistics | |||||
| R2 | 0.7846 | 0.8419 | 0.7136 | 0.7510 | |
| Standard error | 0.0161 | 0.122 | 0.0172 | 0.0114 | |
| Sum of squared residuals | 0.0031 | 0.0018 | 0.0038 | 0.0017 | |
| F | 14.57 (3.49) | 15.98 (3.26) | 10.80 (3.41) | 13.07 (3.41) | |
| LM | 1.31 (4.84) | 2.41 (4.84) | 0.28 (4.75) | 0.13 (4.75) | |
| NORM (2) | 1.84 (5.99) | 0.43 (5.99) | 1.41 (5.99) | 0.08 (5.99) | |
| ARCH (1) | – (4.96) | 0.09 (4.96) | 0.01 (4.84) | 1.22 (4.84) | |
| Number of observations | 16.0 | 17.0 | 17.0 | 17.0 | |
| Number of observations multiplied by the sum of the square of the first five autocorrelations | 3.12 | 4.51 | 1.64 | 6.00 | |
Estimation Results: Full Sample
| Variable | Denmark (1) (1972–1987) | Finland (2) (1971–1987) | Norway (3) (1971–1987) | Sweden (4) (1971–1987) | |
|---|---|---|---|---|---|
| Constant | -0.0989 | 0.0350 | -0.0054 | 0.0209 | |
| (-2.76) | (0.38) | (-0.61) | (1.56) | ||
| Rate of growth of real disposable | 0.5967 | 0.8016 | 0.4976 | 0.7209 | |
| income | (5.22) | (5.59) | (2.51) | (4.18) | |
| Lagged consumption income ratio | -0.4026 | -0.5640 | -0.5044 | -0.4767 | |
| (-2.70) | (-1.86) | (-3.95) | (-2.56) | ||
| Lagged wealth income ratio | … | 0.042 | … | … | |
| (0.67) | |||||
| Change in real wealth | 0.1373 | 0.0863 | 0.3676 | … | |
| (2.65) | (1.49) | (4.89) | |||
| Rate of inflation | … | … | … | -0.3621 | |
| (-2.65) | |||||
| Summary statistics | |||||
| R2 | 0.7846 | 0.8419 | 0.7136 | 0.7510 | |
| Standard error | 0.0161 | 0.122 | 0.0172 | 0.0114 | |
| Sum of squared residuals | 0.0031 | 0.0018 | 0.0038 | 0.0017 | |
| F | 14.57 (3.49) | 15.98 (3.26) | 10.80 (3.41) | 13.07 (3.41) | |
| LM | 1.31 (4.84) | 2.41 (4.84) | 0.28 (4.75) | 0.13 (4.75) | |
| NORM (2) | 1.84 (5.99) | 0.43 (5.99) | 1.41 (5.99) | 0.08 (5.99) | |
| ARCH (1) | – (4.96) | 0.09 (4.96) | 0.01 (4.84) | 1.22 (4.84) | |
| Number of observations | 16.0 | 17.0 | 17.0 | 17.0 | |
| Number of observations multiplied by the sum of the square of the first five autocorrelations | 3.12 | 4.51 | 1.64 | 6.00 | |
In order to assess which factors contributed to the predictive failure of the models, the consumption functions were re-estimated using observations from the whole period. The results of the models are presented in Table 3. It is not surprising that the models used in the subsample also produced good results in the whole sample for Sweden, and the coefficients of the model were virtually the same in the subsample and in the whole period. It is also worth noting that in the case of Finland, a simple-error-correction model, modified to include inflation, fitted the data well. This suggests that changes in households’ real financial wealth, which were not included in the wealth variable because of a lack of data, may have been an important determinant of Finnish saving behavior in recent years. Finally, in the case of Norway, the elasticity of consumption to wealth increased markedly. In the longer sample period, the coefficient of the wealth variable not only became strongly significant but its value increased threefold. The elasticity of consumption to wealth also appeared to have increased somewhat for Denmark and Finland.
In summary, the estimates suggest that the deregulation had an impact on saving-consumption behavior in all the Nordic countries except Sweden. Moreover, the results indicate that the wealth effect has played an important role in household consumption-saving decisions in recent years. Although the lack of data did not permit the construction of a wealth variable that would include all elements of household real net wealth, the findings support the view that the response of consumers to changes in wealth has become stronger since the deregulation of financial markets. This is consistent with the view that freer access to credit has increased consumption expenditure not only directly, but also indirectly through higher asset values.
IV. Conclusions
This paper examined household saving behavior in the Nordic countries. A standard life-cycle model was used, based on a more general model of intertemporal consumption and saving as an analytical starting point.
The findings support the argument that household consumption and saving behavior has changed since the introduction of financial deregulation. In the cases of Finland and Denmark, and to some extent, Norway, the results indicate that earlier economic relationships appeared to have broken down. It is thus not surprising that the authorities have had difficulties anticipating the effects of the deregulation on consumption and saving behavior with any degree of accuracy. Only in the case of Sweden do the data suggest that structural changes have not taken place since deregulation. Furthermore, the results indicate that wealth effects have played an important role in determining consumption with consumers’ response to changes in real wealth apparently increasing since deregulation.
These developments in the Nordic countries should be assessed in the light of the generous tax deductions allowed on household interest payments. Such deductions were allowed not only on mortgage loans but on consumer loans as well. Prior to deregulation, low after-tax interest rates were mitigated by credit rationing. However, after deregulation, a surge in household demand for credit was not fully countered by an increase in nominal interest rates, since the fixed exchange rate policy implied that nominal domestic interest rates were largely determined by foreign interest rates. Because increases in nominal interest rates were limited, the logical alternative to depressing the demand for credit would have been to reduce the tax value of interest payments. The recent decline in the saving ratios has encouraged tax reform in the Nordic countries. The reforms have focused on the need to increase saving in the household sector, which, with unchanged public saving, would imply a stronger current account balance. At this point, it is difficult to predict whether the household saving ratio will ultimately return to its level prior to the financial deregulation after the stock adjustments of household portfolios are fully completed. Since the present tax rules are still biased against saving, cuts in marginal tax rates and reductions in the tax value of interest payments would be expected to increase private saving.
REFERENCES
Abrams, Richard K., “The Financial Reform in Finland,” Kansallis-Osake Pankki, Economic Review, No. 1 (1989).
Andersson, Krister, “Tax Reforms in Scandinavia during the 1980s,” a paper presented at the American Economic Association’s Meetings in New York (December 1988).
Ando, Albert, and Franco Modigliani, “The ‘Life Cycle’ Hypothesis of Saving: Aggregate Implications and Tests,” American Economic Review, Vol. 53 (March 1963).
Berg, Lennart, “Hushållens sparande och konsumtion,” Spardelegationen, Allmänna Förlaget (Stockholm 1988).
Blinder, Alan S., and Angus Deaton, “The Time-Series Consumption Function Revisited,” Brookings Papers on Economic Activity: 2 (Washington: The Brookings Institution, 1985).
Campos, Julia, and Neil R. Ericsson, “Economic Modeling of Consumers Expenditure in Venezuela,” International Finance Discussion Papers, Board of Governors of the Federal Reserve System, No. 325, June 1988.
Charpin, Françoise, “The Life Cycle Model with Imperfect Capital Markets,” a paper presented at the 9th World Congress of the International Economic Association, Athens, 28 August-1 September, 1989 (unpublished: International Economic Association, 1989).
Davidson, James E., and others, “Econometric Modelling of the Aggregate Time-Series Relationship Between Consumers’ Expenditure and Income in the United Kingdom,” The Economic Journal, Vol. 88 (December 1978).
Deaton, Angus, “Involuntary Saving Through Unanticipated Inflation,” American Economic Review, Vol. 67 (December 1977).
Deaton, Angus, “Saving and Liquidity Constraints,” NBER Working Paper 3196 (Cambridge, Massachusetts: National Bureau of Economic Research, December 1989).
Det ∅konomiske Råd, “Danish Monetary Policy in Transition,” (Copenhagen: Akademisk Forlag, September 1985).
Flemming, J.S., “The Consumption Function when Capital Markets are Imperfect: The Permanent Income Hypothesis Reconsidered.” Oxford Economic Papers, Vol. 25 (July 1973).
Hämalainen, Sirkka, “The Savings Behavior of Finnish Households—A Cross-Section Analysis of Factors Affecting the Rate of Saving,” Bank of Finland Publications, Series B:36 (Helsinki: Bank of Finland, 1981).
Hayashi, Fumio, “Tests for Liquidity Constraints: A Critical Survey,” NBER, Working Paper 1720 (Cambridge, Massachusetts: National Bureau of Economic Research, October 1985).
Heinesen, Eskil, “The Relationship Between Consumption, Income, and Wealth in Denmark,” a paper presented at the 19th Annual Meeting of Project Link (unpublished; Project Link, 1987).
Hendry, David F., “Predictive Failure and Econometric Modelling in Macroeconomics: The Transactions Demand for Money,” in Economic Modelling, Current Issues and Problems in Macroeconomic Modelling in the U. K. and the U.S., ed. Paul Ormerod (London: Heinemann, 1979).
Hendry, David F., “PC-Give: An Interactive Econometric Modelling System,” (Oxford: Institute of Economics and Statistics and Nuffield College, University of Oxford, January 1989).
Hendry, David F., and Thomas von Ungern-Sternberg, “Liquidity and Inflation Effects on Consumers’ Expenditure,” in Essays in the Theory and Measurement of Consumer Behavior, ed. by Angus Deaton (Cambridge: Cambridge University Press, 1981).
King, Mervyn, “The Economics of Saving, A Survey of Recent Contributions,” in Frontiers of Economics, ed. K.J. Arrow and S. Honkapohja, Basil Blackwell (1985).
Koskela, Erkki, and Matti Virén, “Inflation, Tight Money and Household Saving Behavior: Finnish Evidence,” Scandinavian Journal of Economics, Vol. 84 (1982).
Koskela, Erkki, and Matti Virén, “Credit Rationing and Consumer Intertemporal Choice,” Oxford Economic Papers, Vol. 36 (June 1984).
Koskela, Erkki, and Matti Virén, “On the Role of Inflation in Consumption Function,” Weltwirtschaftliches Archiv, Vol. 121 (1985).
Modigliani, Franco, “The Life-Cycle Hypothesis of Saving Twenty Years Later,” in The Collected Papers of Franco Modigliani, ed. by Andrew Abel (Cambridge, Massachusetts: MIT Press, 1980).
Koskela, Erkki, and Matti Virén, and Richard Brumberg, “Utility Analysis and the Consumption Function: An Interpretation of Cross-Section Data,” in Post Keynesian Economics, ed. by Kenneth Kurihara (London: Allen and Unwin, 1955).
Nickell, Stephen, “Error Correction, Partial Adjustment and All That: An Expository Note,” Oxford Bulletin of Economics and Statistics, Vol. 47 May 1985).
Palmer, Edward E., “Determination of Personal Consumption, Theoretical Foundations and Empirical Evidence from Sweden,” (Stockholm: Konjunkturinstitutet, 1981).
Palmer, Edward E., “Household Saving in Sweden and Its Composition: An Empirical Analysis,” National Institute of Economic Research (Stockholm: Almquist and Wiksell International, 1985).
Sturm, Peter H., “Determinants of Saving: Theory and Evidence,” OECD Economic Studies (Paris: Organization for Economic Cooperation and Development, 1983).
Olli-Pekka Lehmussaari was an economist in the European Department when this paper was written. He holds a degree from the University of Helsinki and is now an economist at the Bank of Finland. He thanks colleagues in the Fund for helpful comments.
In this paper “Nordic countries” refers to Denmark, Finland, Norway, and Sweden.
See, for instance, Hämalainen (1981), Koskela and Virén (1982, 1984), and Palmer (1981, 1985). See also Flemming (1973) and Hayashi (1985).
For the origins of these models, see Modigliani and Brumberg (1955), Ando and Modigliani (1963), and Modigliani (1980).
For a detailed description of Danish monetary policy, see Det ∅konomiske Råd (1985).
Nonresident purchases of exchange-listed kroner bonds and sales of exchange-listed shares abroad had already been deregulated in the early 1970s, although sales of Danish government bonds abroad were prohibited between February 1979 and May 1983.
In the gray market short-term lending and borrowing by firms took place outside the banking system. Part of this activity was intermediated by banks, but not through their balance sheets.
The saving rate is defined as the ratio of net saving to net disposable income.
Statistics do not distinguish between household and business saving in Denmark.
In 1986 top marginal income tax rates for individuals in the Nordic countries stood between 70 percent and 80 percent.
The so-called potato diet, which was introduced in 1986, further reduced the tax value of interest payments on consumer loans to 31–37 percent.
For purposes of comparison, in 1988 average household disposable income was about Fmk 65,000.
In Sweden income is divided into six sources. A deficit in one source of income can be used to offset net income from other sources. New tax rules allow a person to deduct interest payments fully against positive capital income. In this case, the tax value of interest payments is equal to an individual’s marginal tax rate. The tax value of interest payments is limited to 50 percent only if interest payments exceed capital income (Andersson (1988)).
Recent surveys of the economics of saving can be found in Blinder and Deaton (1985), King (1985), and Sturm (1983).
The principal implication of the traditional life-cycle model is that income and consumption profiles are separated (see Ando and Modigliani (1963) and Modigliani (1980)). Recent studies have suggested that the rejection of this implication in several empirical studies might he due to the existence of liquidity constraints (see Charpin (1989) and Deaton (1989)).
A straightforward way of testing the error-correction model in the context of equation (1) with no q or z variables is to test the hypothesis H1: β1 + β2 + β3 =0. If the coefficient restriction β1 + β2 + β3 =0 is imposed, then equation (1) becomes identical to equation (2).
See also Nickell (1985) and Campos and Ericsson (1988).
The Hendry and von Ungern-Sternberg model can be tested in the context of the basic equation (1) with no q variables, zt-1 = at-1 and η = β4, by testing H2: β1 + β2 + β3 + β4 = 0. If this restriction is imposed, then equation (1) becomes identical to equation (5), with θ0 = β0, θ1 = β2, θ2 = β1 and θ3 = β4.
For Denmark the estimation period started in 1972, owing to lack of data.
All results were obtained using PC-GIVE version 6.0; see Hendry (1989).
It should be pointed out that this substitution could result in a misinterpretation of empirical results. For instance, in the United Kingdom the steep fall in personal saving in the course of the 1980s was largely matched by a rise in industrial and commercial sector saving.
Norwegian national accounts have provided information on households’ disposable income only since 1975. Unpublished data provided by Norges Bank were used for 1970–76.
Since the equations have a lagged dependent variable, a Lagrange multiplier (LM) test should have been used. However, because of the small sample size, the number of degrees of freedom did not allow the computation of the LM-test statistic.
To shed more light on the parameter constancy, a recursive estimation method was also carried out. Three features were noteworthy. First, the calculated one-step residuals suggested that a shift in the economic relationships took place in Denmark and Finland. In these cases, one-step residual errors were strongly positive after the deregulation, and errors were twice the standard error region that indicates coefficient changes and that the change in consumption was larger than predicted by the model. Also, in the case of Norway, the one-step residual error was clearly positive, but the error remained within twice the standard error region. Second, the sequences of estimated coefficients showed evidence of changes in the coefficient estimates in all countries except Sweden during the years of deregulation. In the case of Sweden the coefficient estimates remained virtually the same over the whole period. Third, the elasticity of income to consumption remained relatively stable over the whole period in all countries, except Denmark, where it has declined somewhat.
For a discussion on the predictive behavior of econometric equations, see Hendry (1979).
