Forced Saving and Repressed Inflation in the Soviet Union, 1986–90: Some Empirical Results

In countries like the former Soviet Union, where wealth is mainly stored in monetary assets, the behavior of the money–income ratio is a poor indicator of the growth of undesired monetary balances (monetary overhang). In those countries a monetary overhang is primarily a wealth overhang, which has to be analyzed by evaluating deviations of actual from desired wealth holdings; this requires an empirical analysis of consumption and saving decisions. In this paper, a consumption function for the Soviet Union is estimated, from which an evaluation of the monetary overhang existing at end–1990 is derived. [JEL D12, E21, E4I, P22]


In countries like the former Soviet Union, where wealth is mainly stored in monetary assets, the behavior of the money–income ratio is a poor indicator of the growth of undesired monetary balances (monetary overhang). In those countries a monetary overhang is primarily a wealth overhang, which has to be analyzed by evaluating deviations of actual from desired wealth holdings; this requires an empirical analysis of consumption and saving decisions. In this paper, a consumption function for the Soviet Union is estimated, from which an evaluation of the monetary overhang existing at end–1990 is derived. [JEL D12, E21, E4I, P22]

In countries like the former Soviet Union, where wealth is mainly stored in monetary assets, the behavior of the money–income ratio is a poor indicator of the growth of undesired monetary balances (monetary overhang). In those countries a monetary overhang is primarily a wealth overhang, which has to be analyzed by evaluating deviations of actual from desired wealth holdings; this requires an empirical analysis of consumption and saving decisions. In this paper, a consumption function for the Soviet Union is estimated, from which an evaluation of the monetary overhang existing at end–1990 is derived. [JEL D12, E21, E4I, P22]

Before 1991, the recorded rate of inflation in the former Soviet Union, as measured by official price indices, was extremely low. In the period 1960–80, the retail price index (which is largely a measure of the prices prevailing in official markets) remained basically unchanged; during the 1980s it increased by slightly more than 1 percent a year. The stability of the official price index is widely believed to have reflected the pervasiveness of price controls, which prevented markets from reaching their equilibrium levels, resulting in repressed inflation and constraints on the availability of consumption goods in the official market. These constraints became more pervasive in the second half of the 1980s.

The concept of repressed inflation, as defined by, among others, Barro and Grossman (1974) and Portes (1989), refers to a situation in which aggregate demand exceeds supply; therefore, the elimination of price controls and rationing would lead to an increase in the average price level.1 If all excess demand is for consumer goods, rationing implies forced saving and an increase in nominal wealth above the desired or equilibrium level. The difference between the nominal stock of wealth actually held and the amount desired in the absence of current and past rationing could be defined as wealth overhang. When the constraints on the availability of financial instruments limit the portfolio choices of households to very liquid or monetary assets, as has been the case in the Soviet Union and in most centrally planned economies (CPEs), it is likely that practically all the involuntary increases in the stock of wealth would take the form of higher holdings of monetary balances, which are defined as excess liquidity or monetary overhang.2

In this paper we argue that if the monetary overhang in the former Soviet Union is primarily a wealth overhang arising from forced saving, then its measurement will clearly have to be based on an analysis of desired against actual wealth accumulation—that is, household saving and consumption behavior (Pickersgill (1980a)). Moreover, since money is primarily a store of value in the Soviet Union, it also follows that standard indicators of monetary disequilibrium, such as a decline in money velocity, which are useful for countries in which money is held primarily for transaction purposes, may be misleading for the Soviet Union. If money is the main store of wealth, the money-to-income ratio should be interpreted as a wealth–to–income ratio. The extent to which an increase in this ratio signals a disequilibrium will therefore have to be assessed against the factors influencing the desired wealth-to-income ratio.

Following a review in Section I of the debate on the existence of forced saving in the Soviet Union and in other CPEs, Sections II and III present a quantitative measure of the wealth and monetary overhang accumulated until the end of 1990, based on an econometric analysis of consumption behavior and of forced saving. Section IV summarizes the main conclusions.3

I. Repressed Saving in the Soviet Union

It is widely accepted that over the second half of the 1980s household consumption in the Soviet Union was below the desired level, and, therefore, savings were being involuntarily accumulated, mostly in the form of monetary balances. There is much less agreement, however, on the nature of consumer behavior and the extent of suppressed inflation before the mid-1980s.

A large body of literature holds the view that forced saving and repressed inflation have been a permanent characteristic of all CPEs, including the Soviet economy.4 According to this view, the existence of possibly mild, but chronic, macroeconomic rationing is an essential component of macroeconomic management in CPEs. In the first place, because of the soft budget constraint of enterprises, the wage bill would always tend to exceed targeted figures (see Winiecki (1985) and Kemme (1989)). Moreover, the wage targets themselves would allow some degree of excess demand, because the distribution of purchasing power in excess of what is required to absorb the supply of consumer goods (at desired saving rates) would allow policy planners to avoid the risk of insufficient demand for consumer goods (Ofer (1990)). Finally, the persistence of shortages would enhance social discipline: social tensions, for example, could be eased by allowing temporary increases of the supply of consumer goods at the most appropriate moment.5

It is also frequently argued that, in an economy characterized by chronic shortages, not only is forced saving present, but voluntary saving tends to be higher because buyers maintain high reserves of purchasing power in order to be able to acquire goods that appear in the market in a random and unpredictable fashion (Kornai (1980, pp. 457, 458), Schroeder and Severin (1976), and Grossman (1990)). Such a precautionary increase in purchasing power tends to inflate the observed saving rate. This point is important because it implies that the desired level of wealth is not independent of either the current or the expected state of shortages affecting the system. Should the shortages be relieved, due, for example, to a price liberalization policy, the portion of wealth voluntarily maintained for precautionary purposes would become part of the overhang.6

Given its nature, chronic excess saving would be difficult to detect even through sophisticated analysis of macroeconomic consumption behavior, because it would result in a permanent increase in the average observed saving rate.7 In order to prove the existence of chronic shortages in the Soviet Union, some authors (for example, Birman (1980), Birman and Clarke (1985), Pindak (1984), Winiecki (1985), and Nove (1986)) have therefore relied mainly on an indicator of stock disequilibrium, namely the ratio between household financial holdings (most of which are in monetary form) and consumption or disposable income. These ratios have indeed exhibited rapid growth throughout the last 30 years (Figure 1), which has been interpreted as indicating an increasing undesired accumulation of monetary balances.

Figure 1.
Figure 1.

Household Wealth, 1955–90

Citation: IMF Staff Papers 1992, 002; 10.5089/9781451947106.024.A002

Source: See Appendix I.

Many authors have, however, challenged this evidence as well as the view that repressed inflation and involuntary saving have always been characteristics of the Soviet economy. The main critique is that, in countries with limited capital markets money-to-income ratios, such as those reported in Figure 1, have a different meaning than they do in economies with well-developed financial markets. As mentioned earlier, in the former Soviet Union, money is primarily a store of wealth, and its behavior in relation to gross domestic product (GDP) should therefore be interpreted as a wealth-to-income ratio. Viewed in this way, several factors could explain a voluntary increase in the aggregate wealth-to-income ratio. Ofer (1990) has explained the increase in the wealth-to-income ratio as arising from (1) the expansion of expenditure in durable goods, given the virtual absence of consumer credit; and (2) the deterioration in the quality and availability of public services and social security payments. Moreover, since the level of household wealth in relation to income or consumption was extremely low at the beginning of the 1960s (compared, for example, with wealth-to-consumption ratios in Western countries), a steady rise in the desired wealth-to-income ratio should not be surprising (Asselain (1981), Portes (1989)).

Finally, a well-known implication of the life cycle approach to consumption behavior is that, due to aggregation, the equilibrium aggregate wealth-to-income ratio is a negative function of the growth rate of income (see Modigliani (1986), for example). Thus, the increase in the wealth-toincome ratio in the Soviet Union may be explained by the deceleration of disposable income growth between the 1960s and the 1980s (Figure 2, upper panel). Note also that if we took the money-to-income ratio as an indicator of a monetary overhang, we would have to conclude not only that the overhang accumulated throughout the last 30 years, but also that its rate of growth did not accelerate in the second half of the 1980s, which is hardly credible in light of the increased rationing characterizing the most recent years.

Figure 2.
Figure 2.

Disposable Income and Household Saving, 1955–90

Citation: IMF Staff Papers 1992, 002; 10.5089/9781451947106.024.A002

Source: See Appendix I.

These criticisms strongly suggest that an analysis of repressed inflation in the Soviet Union should start from an evaluation of forced saving and of undesired wealth accumulation. In this respect, Figure 2 (lower panel) shows that the saving rate remained approximately stable, at a level close to 6 percent for about two decades before 1985 and that only in the second half of the 1980s did it start to rise rapidly. Although this evidence suggests that forced saving became more intense in the last few years, the existence of repressed inflation in the previous period cannot be ruled out a priori. Indeed, the stability of the saving rate in 1965–85 may hide a constant degree of forced saving (giving rise to what may be called a “chronic component” of this overhang). The increase of the saving rate observable between 1955–66 and the following period supports this possibility. Consequently, in the economic analysis in Section II we take an “agnostic view” and we do not rule out the existence of repressed saving before the mid–1980s.8

II. Actual and Desired Consumption in the Soviet Union

In this section a quantitative measure of the monetary overhang that accumulated until the end of 1990 is derived from an analysis of consumption behavior and forced saving.

Analytical Framework

Standard consumption theory based on the life cycle hypothesis (for example, Blanchard (1985)) stresses the existence of a relation of proportionality between (desired) household consumption and household human and nonhuman wealth:


where Cd is household consumption, W is nonhuman wealth, and H is human wealth (see Appendix I for definitions). We assume that A can be expressed as

A=bo  exp(Xb),(2)

where b is a vector of coefficients, and X stands for a set of variables affecting the ratio of equilibrium consumption to wealth. Equation (1) assumes a unitary wealth elasticity of consumption; and it is based on the hypothesis of equal elasticity of consumption to human and nonhuman wealth.9

In order to allow for more flexibility in the empirical specification of consumption behavior, we assume, instead, that


where a1 and a2 are parameters. Of course, if a1 = 1 and a2 = 0, equation (2) collapses to equation (1). In addition, in order to reduce the collinearity between regressors, it is convenient to write (3) as


Consider now the possibility that observed consumption differs from desired consumption; without loss of generality, the relation between actual and desired consumption can be written as


where C is actual consumption, and R is unity minus the amount for which consumption demand is rationed (as a percentage of desired consumption). Note that R is, of course, equal to unity in the absence of disequilibrium and, as long as C is measured exactly, that it reflects all possible forms of forced substitution, including purchases in the black market.10 For the sake of generality, we express R in the following way:


where c0 and c1 are, respectively, positive and negative constants, and IR is an observable variable that is an increasing function of the degree of rationing. Equation (6) allows for the presence of macroeconomic rationing embodied in an unobservable component (that is, c0≠ 1) that remains approximately constant in relation to consumption.11 We assume that co incorporates possible voluntary shifts of consumption because of the existence of rationing at the micro level (due, for example, to increasing precautionary saving, as argued by the chronic rationing school).12IR (the rationing indicator) is, in our case, the ratio between the price level in the nonofficial market (specifically in the kolkhoz market) and in the official market (weighted or unweighted for the size of the kolkhoz market with respect to official markets), which is assumed to increase with the level of rationing. It is, therefore, implicitly assumed that the kolkhoz market is extended enough to provide a reliable indicator of excess demand, but is not sufficient (especially because of the limited number of products supplied) to eliminate excess demand at the macro level.13

By substituting equations (2), (4), and (6) into (5), we finally obtain (in log form)


It remains to specify the variables included in X. Four variables have been considered to be relevant in the present context: (1) the real interest rate on deposits; given the low level of private wealth in the Soviet Union, the expected sign of this variable is negative as the substitution effect prevails on the income effect; (2) the “dependency ratio,” defined as the ratio between the nonworking population (children below 16 years of age and pensioners) and the remaining population with an expected positive sign, since a large number of children and old people imply a high share of people with relatively low saving rates; (3) the “benefit ratio,” defined as the ratio between the Social Consumption Fund benefits (mainly pensions and other grants provided by the Government) received by each nonworking member of the population and the wage rate, with an expected positive sign, since high nonlabor incomes reduce the need to accumulate savings; and (4) the inflation rate, with an expected negative sign, because high inflation reduces the real value of accumulated wealth, which has to be restored by increased saving.14

A dynamic specification is postulated, assuming a slow adjustment of consumption to changes in wealth; in particular, the following “quasi-error-correction mechanism” was adopted:


where Y is disposable income, and D is the first-difference operator.15

In the above framework the difference, CdC=SSd, represents forced saving, which, short of the unobservable chronic component of rationing, co, can be measured from the estimates of equation (8). To move from the definition of flow disequilibrium (forced saving) to that of stock disequilibrium (the wealth overhang), note, first, that the choice on desired consumption involves a choice of (end-of-period) desired wealth; calling Wd desired end-of-period wealth, we have, in the absence of capital gains


where Sd=YCd.16 Thus, if desired saving is smaller than actual saving, actual wealth will be higher than desired wealth; that is, there will be a wealth overhang at the end of the period.17 We now define the overhang at time t as the difference between actual wealth holdings and the amount of wealth that would have been held in the absence of current and past forced saving. This is given by


where W0 is the initial value of wealth (that is, the value of wealth before the first period of rationing).

Assume now that we had an estimate of all parameters of equations (7) and (8), including a judgmental estimate of the breakdown of the constant allowing for an identification of co (the chronic overhang); desired saving could be obtained through a dynamic simulation of the equation over the complete period, starting at a point in which the overhang is considered to be zero or small, and the parameters reflecting the amount of rationing, co and c1, are set to zero. The cumulative sum of desired saving (plus the initial value of wealth) represents the desired level of wealth. Comparing this level with actual wealth, as in equation (10), it is possible to evaluate the stock of the wealth overhang.

In order to evaluate the monetary overhang, the first step is, therefore, to estimate equation (8). Since that equation takes explicitly into account the possibility of rationing, in principle, there is no need to exclude from the estimation period years in which rationing was considered to be particularly strong (such as the late 1980s). It was found, however, that estimating (8) over the entire data range available yielded poor results. More specifically, it was clear that the parameter reflecting the degree of observable rationing, d3, was not stable. This instability was an indication of the existence of a structural break, probably due to an increasing degree of rationing.18 The above procedure for the measurement of the overhang was therefore modified as follows: first, equation (8) was estimated in a period in which behavioral parameters appeared to be stable. As a first approximation, the sample period included 1964–85, but, as reported in Section III below, statistical tests were then used to identify more precisely the year of the break. The estimated equation was then used to simulate over the entire data sample (including the late 1980s) the behavior of desired consumption (after setting the rationing coefficients to zero, if significant, in the estimates); forced saving was finally measured as the difference between actual and desired saving.

Estimates of the Consumption Function

Table 1 reports the OLS estimates19 of equation (8) on annual data for the period 1964–85.20 Together with coefficient estimates, t-statistics, and the usual goodness-of-fit indicators, the table shows a number of statistics relative to “diagnostic tests” on the normality of residuals, on the absence of autocorrelation and of heteroscedasticity, and on the within-sample stability of the equation.21

Table 1.

Estimates of Household Consumption

(OLS; annual data: 1964-85; dependent-variable D log C)

article image
article image
Note:t-statistics are reported in parentheses. For all the above tests, with the exception of the Durbin-Watson (DW) and Harvey’s PSI tests, the table reports the percentage of the appropriate test distribution lying on the right of the computed test statistic under the null hypothesis of absence of misspecification. The lower bound of the Durbin-Watson statistic (at the 5 percent level) is 0.637 in the equation with eight regressors and 1.026 in the equations with four regressors (including the constant). Harvey’s PSI test has a t distribution with T - K – 1 degrees of freedom. Y = households’ disposable income (per capita in real terms); W – total nonhuman wealth (per capita in real terms); C = households’ consumption (per capita in real terms); H = human wealth (per capita in real terms); IR1 = Holzman’s disequilibrium indicator (see Holzman (1960)); and IR2 = relative kolkhoz prices.

Lagrange multiplier test of normality (Jarque and Bera (1980)).

Lagrange multiplier test of autocorrelation of residuals modified for small samples (Harvey (1981)).

Portmanteau test of autocorrelation of residuals (Ljung and Box (1978)).

Test of autoregressive conditional heteroscedasticity (Engle (1982)).

Test of linear dependence between residuals and regressors (Breusch and Pagan (1979)).

Test of stability of equation parameters against the hypothesis of structural break in 1975.

Test of stability of residual variance against the hypothesis of structural break in 1975 (Phillips and McCabe (1983)).

Wald test of stability of equation parameters against the hypothesis of structural break in 1975 in the presence of instability of residual variance (Honda and Ohtani (1986)).

Stability test based on recursive estimates (starting from the beginning of the sample) (Harvey (1981)).

Stability test based on recursive estimates (starting from the end of the sample) (Harvey (1981)).

In the table, the most general specification (equation (A)), which includes as an indicator of rationing the relative kolkhoz prices weighted by the share of expenditures on the kolkhoz market, as suggested by Holzman (1960)), shows a remarkably good fit, with an adjusted R2 = 0.87 and a standard error of 0.84 percent. The normality and residual autocorrelation tests are easily passed, but one of the two heteroscedasticity tests shows unsatisfactory results; traces of instability are also shown by the Chow tests, which fall well beyond the 5 percent critical value. Of the coefficient estimates, the human and nonhuman wealth effects and the impact effect of disposable income have the correct sign, and the corresponding t-statistics are close or higher than 2. All other variables are statistically nonsignificant and, with the exception of the proxy for rationing (IR1), also have the wrong sign. In summary, while the overall performance of the equation appears to be adequate, there are signs of mispecification. More satisfactory results are obtained in equation (B) by removing all variables that were not significant at the 10 percent level in equation (A): all test statistics remain approximately stable or improve (with the single exception of the variance ratio test), and the standard error drops to 0.76 percent. Note also that all t-statistics increase substantially. Finally, the coefficients on human and nonhuman wealth, even if unrestricted, are extremely close. The equality restriction is imposed in specification (C) and is easily accepted (see last row of the table), which implies, in terms of the parameters of equation (7), that a1 = 1, and a2 = 0. The equation’s standard error is further lowered and the variance ratio test improves. In summary, specification (C) seems to be entirely adequate in terms of goodness of fit, diagnostic tests, and consistency of parameter values with what could be expected from economic theory. Actual and fitted values for this equation are shown in Figure 3.22

Figure 3.
Figure 3.

Consumption Equation

(Equation (C) of Table 1)

Citation: IMF Staff Papers 1992, 002; 10.5089/9781451947106.024.A002

In equation (D), the unweighted ratio between kolkhoz and state prices replaces Holzman’s index as an indicator of rationing (IR2 in the table). Specification (D) has a slightly better fit, but shows signs of instability; and although the rationing index and the benefit ratio have the expected sign and are significant, the interest rate and the inflation variables, together with the dependency ratio, still have the wrong sign. What is worse, the term including human wealth is not significant, implying, in principle, the absence of a direct effect of labor income on consumption in the long run. Because of this surprising feature, the equation was re-estimated, dropping all variables with the wrong sign. In the re-estimated equation (equation (E)), the significance of both the benefit ratio and the rationing proxy drops below the 10 percent level, while the coefficient on human wealth increases in value and significance. When the (now insignificant) benefit ratio is removed, the t-statistic of the rationing proxy falls to – 0.91, while that on human wealth rises above 3. Thus, regardless of the rationing proxy used in the general specification, the specification search leads to the simple representation of consumption behavior reflected in equation (C).23

Some of the properties of this equation deserve comment. The specification implies a stable long-run relation between consumption behavior and total (that is, human and nonhuman) wealth of the household sector. Since the elasticity of consumption with respect to wealth is unity, the long-run marginal and average propensities to consume are equal. Moreover, it has been shown that (as d1 = d2) changes in human and nonhuman wealth have the same effect on consumption.

Regarding the dynamic properties of the equation, the fact that the coefficient on the change of disposable income is 0.85 (that is, it is less than 1) implies that the short-run effect of income changes, possibly because they are not perceived as permanent, is lower than the long-run effect, which can explain why in years of strong decline in real disposable income (such as 1981–82) the saving propensity tended to be low (Figure 2). Moreover, since the nonhuman wealth variable enters only in lagged form, changes in human wealth tend to affect consumption faster than changes in nonhuman wealth.

The preferred specification is relatively parsimonious, since it explains the movements in consumption simply in terms of wealth and disposable income movements, without making use of additional variables. The failure to identify additional effects may of course be due to the absence of sufficient variability of some of the other included regressors or to measurement problems. However, the fact that the estimated model produces a remarkable fit and generates small i.i.d. (identically and independently distributed) residuals indicates that in the sample period the main movements of consumption expenditure in the Soviet Union can be explained without the use of additional variables. Note, however, that the failure to identify any rationing effect does not rule out the possibility that rationing, while present, was equivalent to a constant proportion of actual consumption and, being part of the constant of the equation (that is, co≠ 0), could not be identified. Indeed, as already mentioned, the possible existence of some chronic rationing will have to be taken into account judgmentally.

III. Estimates of Household Wealth Overhang

An estimate of the wealth overhang can be derived by cumulating excess saving, which in turn can be obtained as a difference between desired and actual consumption. Since the “preferred equation” does not include any rationing proxy, we can conclude that desired and actual consumption were equal during the sample period—that is, there was no overhang (except, possibly, for the chronic overhang component). We also show that the estimated consumption function presents a high degree of parameter stability during the sample period. However, as we move toward the second half of the 1980s, it becomes clear that the relation between observed consumption and the regressors included in the preferred equation progressively breaks down.

Figure 4 reports the one-step-ahead prediction errors of equation (C) of Table 1: these errors are consistently negative (actual consumption is lower-than-projected consumption) outside the sample period (and exceed the two-standard-errors band as of 1987). To test formally for the existence of instability, the equation was re-estimated over 1964–90, and the hypothesis of parameter stability (against that of a structural break in 1985) was evaluated by a Chow test. The value of the test statistics (7.64) falls beyond the critical value even at the 0.5 percent level. This instability may be due to three reasons. First, the structural relation determining desired saving may have changed; more specifically, Soviet households substantially increased their propensity to save. Second, because unrecorded consumption transactions in parallel markets increased, measurement errors may account for the apparent increase in saving rates.24 Third, actual consumption differed from desired consumption because of rationing.

Figure 4.
Figure 4.

One-Step-Ahead Prediction Errors

(Equation (C) of Table 1)

Citation: IMF Staff Papers 1992, 002; 10.5089/9781451947106.024.A002

Note: The dashed lines mark the two-sided standard error band.

Clearly, there is no statistical basis to attribute the behavior of projection errors in the second half of 1980 to one of the above specific reasons. However, we find the third reason plausible. Consumption behavior had remained particularly stable in the Soviet Union over two decades, and it is hard to explain why, in the absence of rationing, it should have changed in such a drastic manner. Although the expansion of consumption on secondary markets may have been relevant to the second reason, it has been taken into consideration, at least partially, by the use of adjusted figures for income (see Appendix I). Moreover, since consumption is derived residually from income,25 any underestimate of consumption must be due to an underestimate of disposable income. But in this case not only was actual consumption underestimated, but also desired consumption (albeit to a lower extent), thus moderating the effect of possible underreporting of consumption in parallel markets (Appendix II).

In conclusion, the strengthening of rationing appears the most likely reason for the “oversaving” that characterized the late 1980s. We will therefore assume, as a first approximation, that all the projection errors observed outside the sample period are due to forced saving.

It should be added that the stability tests performed do not tell us very precisely the point at which the break between actual and desired consumption took place. Table 2 shows the results of Chow tests on the parameter stability against that of the structural break between 1975 and 1986. On the one hand, the hypothesis of stability can be rejected at the 1 percent level for all break years after 1980 (and at the 5 percent level for all breaks after 1978).26 On the other hand, a recursive Chow test based on one-step-ahead predictions allows the rejection of parameter stability at the 5 percent level only after 1987; loosely speaking, this means that while the break may have occurred before, only after 1987 is the behavior of the residuals such that the hypothesis of stability can be rejected. This leaves a rather large interval (1981–87 or, possibly, even 1979–87) in which the break may have occurred.

Table 2.

Chow Tests on the Existence of a Structural Break in Consumption Behavior

article image

Area of the F-distribution on the right of the reported F-value.

Some additional indications can be derived by applying the maximum- likelihood procedure suggested in Quandt (1958) for identifying points of structural break. The value of the log-likelihood function for breaks occurring in each year between 1976 and 1987 reaches an absolute maximum for the event of a break in 1985, but it also reaches a high local maximum in 1982. All values of the likelihood function between 1982 and 1986 are fairly close. In light of the limited number of observations on which the likelihood function was computed, it would be unwise to draw precise conclusions from relatively small differences in the value of the likelihood function.

In the absence of precise indications about the moment at which the break occurred, it seems reasonable to consider different possible starting points for the beginning of rationing. Column (2) of Table 3 reports the cumulated value of forced saving at the end of 1990, derived as the difference between consumption as projected by equation (C) of Table 1 and actual consumption; each row refers to a simulation starting in the reported year and ending in 1990.27 The value of the cumulated sum at the end of 1985 is also reported in column (1). The reported figures are interpreted as estimates of the overhang at the end of 1990 and of 1985, respectively, with the exclusion of the possible chronic overhang (that is, for c0 = 1), which is not revealed by the econometric estimates and will be evaluated judgmentally (see below). The cumulative sum of forced saving ranges from 89 billion rubles (rub) for simulations starting in 1987 to rub 143 billion for the simulation starting in 1982, with an average of rub 116 billion. Figure 5 (top panel) shows the evolution of the estimated overhang for each of the possible starting years of rationing. It is clear that, regardless of the starting year, the overhang accumulated faster as we approach the end of the decade.

Table 3.

Estimates of the Monetary Overhang

(Billions of rubles)

article image

Excluding the chronic component.

Column (6): under the assumption that the chronic overhang component increases after 1985; column (7): under the assumption that the chronic overhang component remains constant after 1985.

Figure 5.
Figure 5.

Estimates of the Wealth Overhang

Citation: IMF Staff Papers 1992, 002; 10.5089/9781451947106.024.A002

a Under the hypothesis that the chronic component of the overhang increases after 1985.b Under the hypothesis that the chronic component of the overhang does not increase after 1985.

Consideration of a possible chronic overhang (implying c0≠ 1) has to he entirely judgmental. The only indication on the existence of a chronic overhang is derived from the results of surveys of household saving and wealth conducted by the Soviet authorities; according to these surveys, the stock of involuntary monetary saving at the end of 1985 was equal to rub 59 billion. This figure exceeds all estimates of the 1985 overhang derived from the equations (Table 3, column (1)); the difference between the survey estimate (rub 59 billion) and the estimates reported in column (1) would give estimates of the chronic overhang existing at the end of 1985; these estimates, reported in column (3) of the table range from rub 38 million to rub 59 billion. Note that from these estimates, and with the additional assumption that the total overhang (including its chronic component) was zero until the mid-1960s, it is also possible to derive estimates of co.28 These estimates, reported in column (5) of the table, are close to 1 percent; the interpretation is that during the 1960s and 1970s consumption fell short of what it would have been in the absence of price controls by about 1 percent; since the size of total consumption is close to that of disposable income, this component of the overhang is also close to about 1 percent of disposable income. Thus, the increase in the saving rate observed around the mid-1960s could be explained, at least partially, by the beginning of the accumulation of a chronic overhang.

If we accept the hypothesis that c0≠ 1, then the estimates of forced saving reported in column (2) of the table should be adjusted, since they underestimate the involuntary wealth by the chronic overhang already existing in 1985 (column (3)), and by the increase, after 1985, in this component. Based on the estimate of co, it is possible to assess the value of the chronic overhang component at the end of 1990; these estimates, reported in column (4), range from rub 54 billion to rub 83 billion. Finally, by adding column (2) to column (4), we find estimates of the total overhang at the end of 1990 ranging from rub 173 billion and rub 197 billion, with an average of rub 189 billion (column (6)). It could, however, be argued that making the chronic overhang increase in the second half of the 1980s is unrealistic. Indeed, if we interpret deviations of c0 from unity not as “chronic forced saving,” but as increases in the precautionary reserves due to the existence of micro shortages, then the need to accumulate additional balances disappears when wealth was already increasing at rates much higher than desired, due to macroeconomic shortages. If we maintain the value of the chronic overhang constant at its 1985 level, we obtain a range for the total overhang from rub 148 billion to rub 181 billion, with an average of rub 168 billion (column (6)).

Note that the average estimates (that is, the estimates obtained by averaging simulation results starting in different years) are also very close to the results obtained from the simulation starting in 1985, the year in which the break is “most likely,” according to the behavior of the likelihood function discussed above. Thus, whether we follow the maximum- likelihood criterion or we take the “average” of all possible cases, we reach estimates of the overhang of close to rub 170 or to rub 190, according to the hypothesis made on the growth of the chronic overhang component. These estimates would be equivalent, respectively, to 30 percent and 34 percent of the actual stock of financial wealth of households at the end of 1990, and to 44 percent and 52 percent of the desired stock (Table 3).29 Note that it is convenient to compare the wealth overhang with the financial component of wealth (rather than with total wealth), because this is the component subject to erosion in the case of price liberalization. In particular, the ratio between the overhang and desired financial wealth represents the increase in the price level that would have been necessary at end-1990 to wipe out the monetary overhang (under the hypothesis that the desired amount of wealth in real terms is independent from the price level30).

IV. Conclusions

In CPEs since monetary balances are often the main form in which private agents can accumulate wealth, the evaluation of excess money holdings (that is, of a monetary overhang) in the former Soviet Union requires, in the first place, a comparison between desired and actual wealth, or, in terms of flows, between desired and actual saving. For this purpose, a consumption function was estimated on annual data between the mid-1960s and the mid-1980s. It was shown that a dynamic error- correction model, involving long-run stability between consumption and total (human and nonhuman) wealth of Soviet households, produced reasonable results in terms of fit and diagnostic checks.

It was also shown that approximately at the middle of the 1980s, the stable relation between consumption and wealth broke down as actual consumption consistently fell short of projected consumption. This development is attributed primarily to macroeconomic rationing in the consumer goods market. The amount of undesired wealth holdings at the end of 1990 was then estimated by cumulating the difference between desired and actual consumption, including an estimate of the chronic overhang accumulated before the mid-1980s. It was concluded that the undesired holdings of wealth in the Soviet Union at the end of 1990 amounted to about rub 170–190 billion, close to 20 percent of GDP and about one third of existing financial assets.

It must be added that given the limited share taken by the real component of saving in the last 20–25 years (Cottarelli and Blejer (1991)), the involuntary accumulation of saving has almost certainly been in the form of financial, and specifically, monetary, assets. It has been estimated that a price increase of 45–50 percent would have been necessary at the end of 1990 to wipe out these excessive monetary holdings.

APPENDIX I Statistical Sources

This Appendix provides a summary of the sources for the statistics used in this paper. For more detailed information, see Cottarelli and Blejer (1991).

Financial Assets and Saving

Financial wealth is the sum of cash, bank deposits, government bonds, and insurance policies, net of household borrowing. All data on financial assets between 1964 and 1989 were provided by the Gosbank; data on bank deposits come from official publications, while data on currency and other financial assets of households were estimated by imposing a constant ratio with respect to bank deposits (at the 1964 level). This may underestimate the actual amount of cash before 1964 if the declining trend in the ratio between cash and deposits observable after 1964 started before that date. In the absence of capital gains on financial assets, net financial saving was equated to the change in the nominal stock of financial assets net of the change in household credit.

Household Disposable Income

Data on disposable income for the 1980s were provided by Goskomstat, and adjusted by adding to the official series the income from the private sale of agricultural products as estimated in U.S. Central Intelligence Agency (CIA) (1989); information on the previous period was derived from CIA (1989) and Pickersgill (1983).

Real Household Investment

Real household investment is made up of two components: houses, and other real investment (both considered net of amortization). For the first component, Goskomstat provided data from 1970 to 1989; data for the previous period were derived from Smith (1973). Goskomstat also provided data for the second component (including mainly livestock and other property of the rural population) for 1970 to 1989; for the previous period, this series, which is small, was kept constant with respect to financial saving.

Household Wealth

Nonhuman wealth was derived as the sum of three components: net financial wealth (see Financial Assets and Saving), houses, and the stock of other real wealth. The value of houses owned by the population, net of depreciation and at current (official) prices from 1965 to 1989 was provided by Goskomstat; the housing investment series was used to derive the stock of houses for the previous period. Goskomstat also provided the value of the third component, but only for 1970–89 and as a cumulative sum of previous investments, not at current prices. For the pre-1970 period the corresponding investment series was also used to derive the stock of this wealth component.

Human wealth is defined as the present discounted value of disposable labor income; this was computed by adding to a three-period centered moving average of current disposable labor income the discounted expected stream of income in the next 27 years; it was assumed that per capita real income was projected to grow at a constant annual rate of 2.5 percent (close to the 1965–85 average), using the average interest rate on bank deposits as a discount factor. The 27-year interval was an estimate of the average number of years that a middle-aged worker expects to receive labor income (including pension payments) for the rest of his or her life.

Household Saving and Consumption and Main Other Series

Household saving was derived as the sum of the three saving components (net financial saving, investment in houses, and investment in other real assets). Consumption was derived residually from disposable income. The inflation series was derived from CIA (1989) and from Pickersgill (1983). Nominal interest rates were provided by the Gosbank; real interest rates were derived by using the above-mentioned inflation series.

APPENDIX II Parallel Markets and the Measurement of the Overhang

The underreporting of transactions on parallel markets can affect the procedure suggested in the text for measuring the overhang for two reasons: first, because the parameters of the consumption function may be affected by measurement errors; and second, because, even if the true parameters were known, the difference between desired and actual consumption may be overestimated if actual consumption and inflation are underestimated.

To appreciate the extent of these problems, consider the following linear version of the model presented in the text. Let Y¯ be disposable income, equal to the sum of YS (disposable income in the socialized sector plus all measured income generated on parallel markets) and of YN (unreported income). Let C and C¯ denote, respectively, measured and actual consumption, and S, saving, which is assumed to be measured with accuracy from financial statistics. The following relation holds:


So, actual consumption is underestimated by an amount equal to YN, and the actual saving rate (S/Y¯) will be overestimated by an amount equal to s(YN/Y¯), where s is the average propensity to save.

A second variable affected by inadequate measurements of activity in parallel markets is the inflation rate. Express the relation between true and measured inflation rate (respectively, p¯ and p) as p¯=p+Φ, and assume now that the true consumption function is


where W is nonhuman wealth; u is an i.i.d. error term; and b1, b2, and f are parameters, the latter summarizing the relation between income and human wealth. By substituting the previous definitions into equation (12), we get


where η=(b1f1)YN+b2Φ+u,which means that in a regression of measured consumption on measured income, the error term of the equation will include terms that are functions of YN and of Φ. Of course, the properties of the OLS estimates of equation (13) will depend on the size of the measurement errors, on the magnitude of the coefficients, b1, f, and b2, and on the stochastic properties of the measurement errors, YN and Φ. We argue that, during the estimation period 1964–85, the effect of the measurement errors is moderate.

Consider first the inflation rate. As described in Appendix I, CIA inflation estimates were used in the regressions, rather than official Soviet figures. We believe that this substantially reduced the systematic component of the measurement error (that is, the systematic underestimation of inflation), possibly reducing the measurement error, Φ, to i.i.d. with no effect on the estimates. Three points can be raised concerning the measurement error of disposable income. First, again, official figures were partially adjusted to account for activity on parallel markets. Second, the relevance of the error is inversely related to the propensity to consume out of disposable income—as b1f approaches unity the effect of the underestimation of YN becomes less important.31 Since we know that the saving rate is quite low in the Soviet Union, the effect of the omission of YN is reduced. Finally, the fact that in the estimates the error appears to be indicates, again, that, if present, YN was small.

Consider now the effect of the omission of YN on the estimates of forced saving. As argued, the omission of YN involves an underestimation of actual consumption equal to YN; desired consumption is, however, also underestimated, for an amount equal to b1 fYN, So the error on the computation of C¯Cd (that is, of forced saving) is only (1b1f)YN. Again, the error is smaller if b1 f is close to unity. Assume that, despite the corrections made to official figures, disposable income is underreported by 2.5 percent in 1988, by 5 percent in 1989, and by 10 percent in 1990;32 by simulating equation (C) of Table 1 with these revised data, the estimate of the overhang at the end of 1990 would be smaller by about rub 10 billion, still a relatively contained percentage of the overhang estimates reported in the text. This amount should be incremented by the increase in desired saving possibly connected to higher inflation in the second half of the 1980s, since part of the increase in savings observed in that period may have been due to the attempt to restore the real value of the stock of wealth eroded by price increases (in both official and parallel markets). However, desired saving may have been depressed by the increasingly negative yield of financial wealth, equally due to higher inflation. Consideration of those effects is, unfortunately, impossible, given the failure to identify the elasticity of consumption with respect to inflation and the interest rate in the estimation period.


  • Asselain, Jean-Charles, “Mythe ou Realite de l’Epargne Force dans les Pays Socialistes,” in Travail et Monnaie en Systême Socialiste, ed. by M. Lavigne (Paris: Economica, 1981).

    • Search Google Scholar
    • Export Citation
  • Barro, Robert J., and Herschel I. Grossman, “Suppressed Inflation and the Supply Multiplier,” Review of Economic Studies, Vol. 41 (1974), pp. 97104.

    • Search Google Scholar
    • Export Citation
  • Birman, Igor, “The Financial Crisis in the USSR,” Soviet Studies, Vol. 32 (January 1980), pp. 84105.

  • Birman, Igor, and Roger A. Clarke, “Inflation and the Money Supply in the Soviet Economy,” Soviet Studies, Vol. 34 (October 1985), pp. 494504.

    • Search Google Scholar
    • Export Citation
  • Blanchard, Olivier, “Debt, Deficits, and Finite Horizons,” Journal of Political Economy, Vol. 93 (April 1985), pp. 22347.

  • Breusch, Trevor S., and A. R. Pagan, “A Simple Test for Heteroschedasticity and Random Coefficient Variation,” Econometrica, Vol. 47 (September 1979), pp. 128794.

    • Search Google Scholar
    • Export Citation
  • Burkett, John, “Slack, Shortage and Discouraged Consumers in Eastern Europe; Estimates Based on Smoothing by Aggregation,” Review of Economic Studies, Vol. 55 (July 1988), pp. 493505.

    • Search Google Scholar
    • Export Citation
  • Cottarelli, Carlo, and Mario I. Blejer, “Forced Savings and Repressed Inflation in the Soviet Union: Some Empirical Results,” IMF Working Paper 91/55 (Washington: International Monetary Fund, June 1991).

    • Search Google Scholar
    • Export Citation
  • Davis, Christopher, and Wojciech Charemza, eds. (1989a), Models of Disequilibrium and Shortage in Centrally Planned Economies (London; New York: Chapman and Hall).

    • Search Google Scholar
    • Export Citation
  • Davis, Christopher, and Wojciech Charemza, eds. (1989b), “Introduction to Models of Disequilibrium and Shortage in Centrally Planned Economies,” in Models of Disequilibrium and Shortage in Centrally Planned Economies, ed. by Christopher Davis and Wojciech Charemza (London; New York: Chapman and Hall).

    • Search Google Scholar
    • Export Citation
  • Dornbusch, Rudiger, and Holger Wolf, “Monetary Overhang and Reforms in the 1940s,” NBER Working Paper No. 3456 (Cambridge, Massachusetts: National Bureau of Economic Research, October 1990).

    • Search Google Scholar
    • Export Citation
  • Engle, Robert F., “Autoregressive Conditional Heteroschedasticity with Estimates of the Variance of the U.K. Inflation,” Econometrica, Vol. 50 (July 1982), pp. 9871007.

    • Search Google Scholar
    • Export Citation
  • Fair, Ray C. and Dwight M. Jaffee, “Methods of Estimation for Markets in Disequilibrium,” Econometrica, Vol. 40 (May 1972), pp. 497514.

    • Search Google Scholar
    • Export Citation
  • Grossman, Gregory, “Monetary and Financial Aspects of Gorbachev’s Reform,” in Financial Reform in Socialist Economies, ed. by Christine Kessides and others, EDI Seminar Series (Washington: World Bank, 1990).

    • Search Google Scholar
    • Export Citation
  • Harvey, Andrew C., Econometric Analysis of Time Series (Oxford: Philip Allan, 1981).

  • Holzman, Franklyn D., “Soviet Inflationary Pressures, 1928–1957: Causes and Cures,” The Quarterly Journal of Economics, Vol. 74, No. 2 (May 1960), pp. 16188.

    • Search Google Scholar
    • Export Citation
  • Honda, Yuzo, and Kazujiro Ohtani, “Modified Wald Tests in Tests of Equality between Sets of Coefficients in Two Linear Regressions Under Heteroschedasticity,” The Manchester School of Economic and Social Studies, No. 56 (June 1984), pp. 20818.

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund, World Bank, Organization for Economic Cooperation and Development, and European Bank for Reconstruction and Development, A Study of the Soviet Economy, 3 vols. (Paris: IMF, World Bank, OECD, and EBRD, 1991).

    • Search Google Scholar
    • Export Citation
  • Jarque, Carlos M., and Anil K. Bera, “Efficient Tests for Normality, Homoschedasticity and Serial Independence of Regression Residuals,” Economics Letters, Vol. 6 (June 1980), pp. 25559.

    • Search Google Scholar
    • Export Citation
  • Kemme, David M., “The Chronic Excess Demand Hypothesis,” in Models of Disequilibrium and Shortage in Centrally Planned Economies, ed. by Christopher Davis and Wojciech Charemza (London; New York: Chapman and Hall, 1989).

    • Search Google Scholar
    • Export Citation
  • Kornai, Janos, Economics of Shortage, 2 vols. (Amsterdam: North-Holland, 1980).

  • Kornai, Janos, Growth, Shortage, and Efficiency (Amsterdam: North-Holland, 1982).

  • Ljung, Greta M., and G. E. P. Box, “On a Measure of Lack of Fit in Time Series Models,” Biometrika, Vol. 66 (1978), pp. 297303.

  • Modigliani, Franco, “Life Cycle, Individual Thrift, and the Wealth of Nations,” American Economic Review, Vol. 76 (June 1986), pp. 297313.

    • Search Google Scholar
    • Export Citation
  • Nove, Alec, The Soviet Economic System (Boston: Allen & Unwin, 1986).

  • Ofer, Gur, “Macroeconomic Issues of Soviet Reforms,” Working Paper 222 (Jerusalem: Hebrew University, April 1990).

  • Ofer, Gur, and Joyce Pickersgill, “Soviet Household Saving: A Cross-Section Study of Soviet Emigrant Families,” Quarterly Journal of Economics, Vol. 95 (August 1980), pp. 12044.

    • Search Google Scholar
    • Export Citation
  • Phillips, G. D. A., and B. P. M. McCabe, “The Independence of Tests for Structural Change in Regression Models,” Economics Letters, Vol. 12 (December 1983), pp. 28387.

    • Search Google Scholar
    • Export Citation
  • Pickersgill, Joyce, “Soviet Household Saving Behavior,” Review of Economics and Statistics, Vol. 58 (May 1976), pp. 13947.

  • Pickersgill, Joyce, (1980a), “The Financial Crisis in the USSR: A Comment,” Soviet Studies, Vol. 58 (October), pp. 58385.

  • Pickersgill, Joyce, (1980b), “Recent Evidence on Soviet Households Saving Behavior,” Review of Economics and Statistics, Vol. 62 (November), pp. 62833.

    • Search Google Scholar
    • Export Citation
  • Pickersgill, Joyce, “Household Saving in the USSR,” in The Determinants of National Savings and Wealth, ed. by Franco Modigliani and R. H. Hemmings (New York: St. Martin’s Press, 1983).

    • Search Google Scholar
    • Export Citation
  • Pindak, Frantisek, “Inflation under Central Planning,” Societe Universitaire Europeene de Recherches, Financieres Reprint Series No. 29 (Tilburg: SUERF, 1984).

    • Search Google Scholar
    • Export Citation
  • Portes, Richard, “The Theory and Measurement of Macroeconomic Disequilibrium in Centrally Planned Economies,” in Models of Disequilibrium and Shortage in Centrally Planned Economies, ed. by Christopher Davis and Wojciech Charemza (London; New York: Chapman and Hall, 1989).

    • Search Google Scholar
    • Export Citation
  • Portes, Richard, and David Winter, “Disequilibrium Estimates for Consumption Goods Markets in Centrally Planned Economies,” Review of Economic Studies, Vol. 47 (January 1980), pp. 13749.

    • Search Google Scholar
    • Export Citation
  • Portes, Richard, Richard E. Quandt, and Stuart Yeo, “Tests of the Chronic Shortage Hypothesis: The Case of Poland,” Review of Economics and Statistics, Vol. 70 (May 1988), pp. 28895.

    • Search Google Scholar
    • Export Citation
  • Quandt, Richard E., “The Estimation of the Parameters of a Linear Regression System Obeying Two Separate Regimes,” American Statistical Association Journal, Vol. 53 (December 1958), pp. 87380.

    • Search Google Scholar
    • Export Citation
  • Schroeder, Gertrude E., and Barbara S. Severin, “Soviet Consumption and Income Policies in Perspective,” in Soviet Economy in a New Perspective, Joint Economic Committee, U.S. Congress (Washington: U.S. Congress, October 1976).

    • Search Google Scholar
    • Export Citation
  • Smith, Williard S., “Housing in the Soviet Union—Big Plans, Little Action,” in Soviet Economic Prospects for the Seventies, Joint Economic Committee, U.S. Congress (Washington: U.S. Congress, June 1973).

    • Search Google Scholar
    • Export Citation
  • U.S. Central Intelligence Agency (CIA), “USSR: Estimates of Personal Incomes and Savings,” SOV 89–10035 (Langley, Virginia: CIA, April 1989).

    • Search Google Scholar
    • Export Citation
  • van Brabant, Jozef M., “Socialist Economics: The Disequilibrium School and the Shortage Economy,” Journal of Economic Perspectives, Vol. 4 (Spring 1990), pp. 15775.

    • Search Google Scholar
    • Export Citation
  • Winiecki, Jan, “Portes ante Portas: A Critique of the Revisionist Interpretation of Inflation under Central Planning,” Comparative Economic Studies, Vol. 27 (Summer 1985), pp. 2551.

    • Search Google Scholar
    • Export Citation

Carlo Cottarelli is a Senior Economist in the Fiscal Operations Division I of the Fiscal Affairs Department. He is a graduate of the University of Siena and holds a graduate degree from the London School of Economics and Political Science.

Mario I. Blejer, Assistant Director in the European Department, is currently on assignment as Chief of the Macroeconomic and Adjustment and Growth Division at the World Bank. He holds a Ph.D. from the University of Chicago and has taught at the Hebrew University of Jerusalem, Boston University, and New York University.

The authors thank Hugh Bredenkamp, Gregory Grossman, Luigi Guiso, Piroska Nagy, Kent Osband, Teresa Ter-Minassian, Allan Whittome, and Thomas Wolf for comments and suggestions. Thanks are also due to the Bank of Italy for allowing the use of the ESTITEST Speakeasy routine used for the econometric results presented in Sections III and IV.


Of course, the existence of aggregate excess supply has to be assessed, taking into account the role of parallel (that is, unofficial or black) markets in supplying additional goods. The empirical relevance of parallel markets for our estimates is discussed in Appendix II, and a more extensive discussion of the theoretical problems involved can be found in Cottarelli and Blejer (1991).


Of course, a monetary overhang can also exist in the absence of a wealth overhang, as in the case in which actual and desired wealth are equal but wealth composition is suboptimal (and biased towards money).


A similar econometric approach to the one adopted here, as well as a discussion on the monetary overhang of Soviet enterprises, is presented in International Monetary Fund and others (1991).


For an extensive discussion of the debate on the existence and extent of repressed inflation in the CPEs, see the surveys by Davis and Charemza (1989a), Fortes (1989), and van Brabant (1990).


Note that, according to this view, households are not able to react to excess saving by reducing their labor supply, given the structural rigidities characterizing labor markets in CPEs.


It could be argued, however, that precautionary saving may actually rise in the wake of systemic reforms if these imply increasing uncertainty regarding employment or social services.


This point is similar to that raised by Kornai (1982, p. 35).


Previous econometric analysis of consumption behavior in the Soviet Union reached ambiguous conclusions on the state of the consumption goods market before the 1980s. Pickersgill (1976) found that consumption behavior between 1955 and 1971 could be explained mainly by movements of disposable income without the use of proxies for rationing effects. However, Pickersgill (1980b, 1983) identified the existence of a structural break in consumption that occurred in the middle of the 1960s (reflected in the increase of the saving rate; see Figure 2); she suggested that this break was due to rationing effects.


The most obvious reason why this may not have been the case in the Soviet Union is that, in the absence of (efficient) capital markets, households could not borrow against human wealth. It must be recalled, however, that this aspect becomes relevant only in the presence of liquidity constraints to consumption. Therefore, we cannot a priori rule out the possibility that, even in the absence of efficient capital markets, desired consumption was equally influenced by both wealth components. Uncertainty on future labor incomes may also imply that the discounted income flow has a lower “weight”; this aspect may, however, not be very relevant in CPEs, given the high degree of “certainty” attached to future labor incomes (see, on this point, Ofer and Pickersgill (1980)).


Thus, if forced substitution (between different goods on the official market or between goods on the government and on the black market) is large, total expenditure (albeit not its composition) approaches desired expenditure, and R is close to unity.


This corresponds to the existence of a “normal” degree of rationing, as suggested by the “chronic rationing” school.


Note that, strictly speaking, this component does not reflect actual rationing, since consumption is voluntarily reduced to allow a buildup of precautionary reserves; this is the reason why this component may not be related to observable demand pressure indicators. However, in the case of price liberalization, this component plays the same role of forced saving; as the reason for the accumulation of precautionary reserves disappears with price liberalization, the accumulated balances become part of the overhang that people desire to spend.


The use of rationing proxies for the estimation of demand curves under rationing, based on Fair and Jaffee (1972), is simpler than the one based explicitly on disequilibrium econometrics (see, with reference to CPEs, Portes and Winter (1980), Portes, Quandt, and Yeo (1988) and Burkett (1988), for example). First, estimates can be performed with ordinary least squares (OLS). Second, there is no need to specify a supply function, nor to separate the sample between points on the supply and points on the demand function; this procedure seems inappropriate given the limited number of observations available. Proxies for rationing in the estimation of consumption functions for the U.S.S.R. have also been used by Pickersgill (1980b).


14 Lack of adequate information on the stock of consumer durables prevented its use as an additional explanatory variable, despite its potential relevance in explaining saving movements in the Soviet Union (as argued in Section I).


Note that if d1 = d2, the long-run elasticity of C with respect to H + W is unity, and there are no composition effects, which corresponds to the case in which, in equation (7), a2 = 0, and a1 = 1.


Note that desired saving has to be evaluated, for all periods, along the equilibrium path; that is, it has to be derived by using in the consumption function desired wealth and not actual wealth.


Even if households desire to spend this wealth overhang, they may not necessarily want to spend it entirely in one period. The share of the overhang that households want to spend in the current period will depend on several factors, including the types of consumer goods and services that were rationed in the past and all factors affecting the intertemporal distribution of households’ resources. In the extreme case in which forgone consumption has not created an overhang of “unsatisfied needs,” the overhang has the same effect on expenditure as a “windfall gain.” which, in a life cycle perspective, will be spent only gradually throughout life.


If the included proxy for rationing were very good, an increased degree of rationing would not necessarily induce instability in the equation. But, in practice, it is possible that the proxies used are inadequate to describe fully the extent of the increase in rationing that occurred during the late 1980s.


One known problem in estimating consumption functions with OLS is the endogeneity of income with respect to consumption demand; in the context of CPEs, however, this endogeneity should not be taken for granted, as total income may be entirely supply determined (so that random shocks in consumption demand would affect mainly stockbuilding rather than total output). More serious may be the consequences of measurement errors affecting disposable income, especially if related to the existence of black markets; however, as argued in Appendix II, the effect of measurement errors in the sample period is probably limited.


By limiting the estimation period to the post-Krushchev era, we are forgoing the possibility of explaining the “jump” in the propensity to save that occurred between the first and second half of the 1960s. However, official data on financial saving (particularly cash holdings) are available only as of 1964, and it cannot be ruled out that the jump in the savings ratio observed in the middle of the 1960s reflects in part differences in data quality (see Appendix I).


Note that for some of the tests reported in the table (specifically for the normality test, the Ljung-Box tests, and the two heteroscedasticity tests) only the asymptotic distribution is known. For those tests, in light of the limited number of observations available, it seems safe to accept the null hypothesis only when the value reported in the table is “substantially” higher than 5 percent. It must also be recalled that in dynamic regressions with only 22 observations, the results may be affected to some extent by bias of the Hurwicz type.


Further attempts to reintroduce in this equation, individually or in combination, the variables previously excluded failed, as these variables remained insignificant.


As detailed in Cottarelli and Blejer (1991), similar conclusions were reached in a specification search using net financial wealth, rather than total wealth. Poorer results were obtained when consumption was related to disposable income, rather than wealth.


Also, it is possible that actual inflation (that is, inflation taking into account price behavior on parallel markets) exceeded measured inflation. While, possibly due to insufficient inflation variability before 1985, inflation did not appear to have influences on consumption in the Soviet Union, it stands to reason that, in the presence of strong inflationary pressures, desired saving may have increased in the late 1980s to restore the value of financial wealth eroded by the higher price level.


Remember that saving is measured reasonably well by financial statistics. Therefore, errors in the measurement of consumption must be due to errors in income measurement.


Similar results are obtained with the generalized Chow test of Honda and Ohtani (1984), also reported in the table.


In each case, equation (C) has been re-estimated ending the sample period in the last year of equilibrium.


Although the choice of an initial year when the overhang was zero is necessarily arbitrary, the results are not very sensitive to the specific year considered; choosing 1970, instead of 1965, as the initial year would alter the results only marginally.


Lack of consideration of the effect of past inflation on the value of real wealth may imply that the above figures overestimate the size of the overhang. However, the overhang may be underestimated if expectations of future price increases actually led to an increase in the current propensity to consume, an effect not measured by the estimated equation.


This assumption is appropriate as long as households perceive that the price increase will not permanently erode their real labor income (that is, as long as wages are moved in line with prices). Note, however, that an equal increase in prices and nominal incomes, while it can eliminate the initial stock of the overhang, would not eliminate the source of the overhang accumulation—that is, the “excessive” real wages. Thus, equilibrium of both stock and flows may require some decline in real income.


Intuitively, this occurs because the implication of having b1f close to unity is that most increases in income are consumed. But the effect of the underestimation of Y is precisely that of making the correlation of Y and C closer to unity (because Cis obtained as a residual, given S). So the problem is less relevant when the correlation between Y and C is actually high.


These percentages correspond to an underestimate of disposable income, by rub 11, 25, and 57 billion, respectively, for the three years. These values are fairly large with respect to the estimated size of parallel markets. For example, according to CIA (1989) the net income of households from sales of agricultural products (including sales on official markets) totaled rub 19 billion in 1987.

IMF Staff papers: Volume 39 No. 2
Author: International Monetary Fund. Research Dept.