Journal Issue

Household Demand for Money in Poland Theory and Evidence

International Monetary Fund. Research Dept.
Published Date:
January 1992
  • ShareShare
Show Summary Details

Money is often accorded an unimportant place in a socialist economy. According to the traditional view, monetary circulation is segmented into household and enterprise circuits (Birman and Clarke (1985), Hartwig (1985), Székely (1990), and Abel and Székely (1988)). In each circuit, flows of money play only a passive, record-keeping role, adapting themselves to the real allocations prescribed by the central plan; enterprise revenues and expenditures are constrained by planned input allocations and output targets, while household expenditures are constrained by the availability of consumer goods. Money flows between the two sectors are associated with wage payments and consumers’ payments for goods and services, but these are also specified in the plan. This traditional view implies that in a socialist economy, demand for money, whether by the household or the enterprise sector, has little empirical relevance: money holdings are not themselves an object of choice, but a by-product of the planning process. As a result, it is not surprising that there have been relatively few studies of money demand in socialist economies (Hartwig (1987)).

A corollary of this traditional view is that household money holdings are in some part involuntary, which is associated with the notion of a liquidity overhang. If households’ incomes exceed the amount they are able to spend on goods, there is forced saving, which, in the absence of an extensive array of alternative financial assets, takes the form of an unwanted accumulation of money (Birman (1980) and Cottarelli and Blejer (1992)). If prices were freed and such an accumulation has taken place, the general price level would have to rise substantially to bring about market equilibrium—implying that before price liberalization takes place there is “repressed inflation” (Nuti (1986, 1990) and Pindák (1983)).

This traditional view has been qualified or challenged by a number of observers, especially as it pertains to the money holding behavior of the household sector. The main criticism is that, in practice, households in socialist economies have many more alternatives than the overhang view would imply (Olés and others (1987)). In particular, even if goods are in short supply in the official shops, there are often large unofficial markets, including free agricultural markets in several countries, as well as black markets, flea markets, and other manifestations of the “shadow economy,” where a variety of commodities can be purchased (Cassei and others (1989)). Moreover, shortages may lead not only to forced saving, but also to “forced spending” on other goods (Kornai (1980, chap. 18) and Podkaminer (1988, 1989)); goods may also play a role as a store of value (Weitzman (1991); see also Portes (1989) and Portes and others (1987)).

A further point is that the range of financial assets is broader than the traditional view assumes, since in many socialist countries household holdings of foreign currency are large. In some countries these holdings consist largely of “mattress money” and deposits in banks abroad, whereas in some countries—notably, Poland and Yugoslavia—there have also been large legal holdings of foreign currency deposits. In Poland in 1988, these foreign currency deposits, when evaluated at the black market exchange rate, accounted for about 40 percent of household money and quasi-money; unofficial holdings were, of course, unknown, but were believed to be quite large. Not only have households held foreign currency, but they have been able to buy and sell it relatively easily; in Poland, in particular, the black market in foreign exchange was very active, even before it was legalized in March 1989.

These considerations suggest that it would be useful to examine the household demand for money in a socialist economy, taking account of the different constraints to which households in this environment are subject. In particular, it is important to consider the shortages of goods in the official shops, the availability of goods in the black market, and the possibility of holding foreign currency as an alternative to domestic money. Households should be viewed as choosing optimal money holdings subject to these constraints—in contrast to the traditional view, according to which households are largely passive in adapting their spending to the availability of goods.

This paper develops this approach to money demand in a socialist economy, first theoretically and then empirically with reference to Poland in the 1980s. The theoretical model is a variant of the cash-in-advance model associated with Clower (1967) and Lucas (1980, 1984).1 This approach synthesizes the various motives for holding domestic and foreign money in this environment (Hartwig (1987)). Under the transactions motive, it is assumed that domestic money is needed for purchases in the official shops, but either domestic or foreign money can be used in the black market. The model also incorporates the precautionary motive, or what Kornai (1980, pp. 457–58) characterized as “purchaser alertness”: under conditions of shortages, consumers can never tell when, and in what quantities, desired goods may be available in the shops, and must have money ready to pay for them. Domestic and foreign money are also stores of value, and holdings of the two moneys will be shown to depend on their rates of return in relation to goods, as well as on the variances and covariances of these returns.2

This theoretical framework is then used as the basis of empirical work, using Polish data on household incomes and money holdings. The empirical method used is the two-stage error-correction model proposed by Hendry (1985, 1986).3 In previous literature, various indicators have been used to represent the influence of shortages on demand for money, but some of these seem rather circular; for instance, Hartwig (1987) used k, the reciprocal of velocity, and Payne (1990) used real consumer credit, as explanatory variables in money demand equations; Feltenstein, Lebow, and van Wijnbergen (1990) constructed “virtual prices” as a combination of official prices and real money, and used these to explain savings—which in China consist mainly of changes in real money.4 This paper takes a different approach: it is assumed that the black market exchange rate adjusts to establish equilibrium in the money market, given the controlled official prices; official prices are hypothesized to adjust toward money market equilibrium—a hypothesis that is scrutinized by testing for cointegration of money, real income, and official prices. This approach implies that the black market exchange rate must be treated as an endogenous variable (as in Charemza and Ghatak (1990)); for this reason, and because expectational variables are hypothesized to affect money demand, an instrumental variables approach is used in estimation.

The paper proceeds as follows. Section I presents the theoretical model and derives its implications for money demand. Section II discusses the data and the underlying economic conditions in Poland over the period of study, December 1979 to October 1988. Section III presents the empirical results, and Section IV offers conclusions.

I. The Model

The model is a variant of the cash-in-advance model. A representative household maximizes expected utility, which is a function of consumption, over an infinite horizon. The household holds two kinds of money, domestic (zloty) and foreign (dollar). Purchases of goods must be backed with money; zloty money must be used in the official shops, while either zlotys or dollars can be used in the parallel (black or free) market. For simplicity, it is assumed that there is only one consumption good; the price is lower in the official shops, but the quantity available there is limited and uncertain, whereas in the parallel market, a consumer may buy as much as desired, but at a higher price. Therefore, zlotys are held in preparation for purchasing goods at a favorable price in the official shops. Whether all these zlotys are used up on the official shops, or some are spent in the black market, or some are held over until the next period depends on what quantity of goods turns out to be available in the official stores, and at what price, as well as on the realization of the black market price.

The timing of transactions in the model is as follows. In the morning, the consumer purchases goods in two markets, constrained by money held over from the previous day. In the afternoon, income is received, and the (parallel) foreign exchange market opens, enabling one currency to be exchanged for another. There are no other financial assets. The price and available quantity in the official shops, as well as the black market price and exchange rate, are treated as random variables, whose realization is not known until the beginning of the day.

The household maximizes expected utility, which is a function of consumption, ct, over an infinite horizon:

where Et denotes the expectations operator conditional on information available at time t. Consumption consists of amounts purchased in official and black markets:5

The household’s budget constraint, expressed in zlotys, for each period requires that purchases of goods during the period plus end-of-period money holdings are financed by money carried over from the previous period and by income:

where pto and ptb are prices of the good in the two markets, expressed in zlotys; mtd and mtf are holdings of domestic and foreign money; et is the exchange rate (the price of dollars in terms of zlotys); and yt is nominal income (expressed in zlotys).

In addition to the budget constraint, the household’s purchases are subject to two cash-in-advance constraints, one for each market. Purchases in the official market are limited by holdings of zlotys carried over from the previous period:

and total purchases on the two markets are constrained by holdings of the two moneys:

The amount that the household can buy in the official market may be limited by the availability of supply:

where the quantity available is ¯xt. If pto<ptb, and if anything is bought in the black market, either the constraint on the household’s purchases in the controlled market implied by the availability of supply (equation (6)) or the cash-in-advance constraint (equation (4)) is binding.

There are also nonnegativity conditions on holdings of the two monies, and on purchases of the goods in the two markets:


The household therefore maximizes the expected discounted value of utility (equation (1)) with respect to xto,xtb,mtd, and mtf in each period, subject to constraints (3), (4), (5), (6), (7), and (8). The solution may be characterized using standard recursive methods. Defining a value function, J(mtd,mtf), as the expected maximized present value of utility from period t + 1 onward as a function of money carried over from period t, the Lagrangean for each period f can be written as follows:

Using this device, one obtains the following first-order conditions for purchases in the two markets:

where utu/ct. The usual Kuhn-Tucker conditions apply to the inequality constraints (7) and (8). Assuming, for simplicity, that something is bought in each of the two markets in each period (so that λ7,t = λ8,t = 0), conditions (10a) and (10b) can be solved to yield

where θt, is the price premium on the free market relative to the official market—that is, θt, = pb/po - 1. Equation (11) can be interpreted as showing that the wedge between the prices in the official and the black market is associated either with supply shortages in the official market as reflected in λ4,t, or with a lack of money to pay for goods there, as reflected in λ2,t. In other words, given the low prices but limited quantities in the official shops, the consumer either buys all that is available at the low prices, or runs out of zloty money in the attempt.

Next, consider the conditions for optimal holdings of domestic and foreign money:

where Ji,t(mtd,mtf)/mi,t,i=d,f. Using the definition of the value function

that is, zloty money is expected to contribute to future utility by providing more wealth and thus easing the budget constraint (reflected in λ1,t) as well as to the extent that it may ease the two cash-in-advance constraints (reflected in λ2,t, and λ3,t). Holding dollar money is expected to ease the budget constraint and the free market cash-in-advance constraint.

For simplicity, let us focus on solutions in which foreign currency is used as a long-term store of value, and the household thus does not exhaust its dollar money this period. In this case, neither the overall cash-in-advance constraint (equation (5)) nor the nonnegativity constraint on dollar balances is binding in period t, so λ3,t = λ6,t = 0. Using conditions (12) and (13) and substituting from (10), overall savings are determined by

where at, = e is the rate of appreciation of the dollar against the zloty from period t to t + 1, and πtb is the percentage increase in prices on the free market from t to t + 1. Next, the condition for optimal holdings of zlotys can be written as

The left-hand side of equation (15) can be interpreted as the expected utility-weighted excess return to holding dollar rather than zloty money. When this expression is positive, the right-hand side is positive, implying that λ5,t+1 > 0, sud that Etλ4,t+1 is small; that is, no zlotys are held, and as a result, the quantity constraints in the controlled market are not binding unless the quantity available, ¯xt, is zero. When the expected utility-weighted excess return to foreign currency is negative, the right-hand side is also negative, implying λ5,t = 0 and Etλ6,t+1 > 0; that is, some zlotys are held, so there are some eventualities that would lead the consumer to have zlotys ready to spend in the controlled market but be unable to buy the desired quantity. If the left-hand side of equation (14) is just zero, this implies that mtz=inf{pt+1ox¯t+1}; that is, just enough zlotys are held that there is no possibility that shortages are a binding constraint on purchases, so that no zlotys are ever left over after the allowed amount has been purchased on the controlled market. A lower expected utility-weighted excess return leads the household to hold larger zloty balances, incurring a higher probability of hitting the quantity constraint before the zlotys-in-advance constraint is reached, as well as implying that the former constraint binds more tightly; this is reflected in a higher Eλ4,t+1.

In order to examine further the implications for the demand for zloty money, take a second-order Taylor approximation of equation (15):

Equation (16) can be used to provide a portfolio-theoretic rationale for the behavior of the demand for domestic currency money. First, the demand for zloty money depends negatively on the expected excess return of dollars over zlotys as a transactions medium, which is the expected rate of appreciation of the dollar vis-à-vis the zloty, net of the expected premium of prices in the free market over prices in the controlled market. This difference is divided by the expected (black market) inflation factor (1+Eπtb), which, for a given dollar appreciation and given intermarket price premium deflates the returns to both dollars and zlotys. Money demand also depends on the variances and covariances of consumption and inflation and the return differential, Eat - Eθt+1, where the signs of the relationships depend on the return differential.

Consider, first, the case in which Eθt+1 > Eat, so the benefit of buying at controlled prices rather than in the free market is expected to exceed the appreciation of the dollar in terms of the zloty. In this case, a higher variance of inflation is associated with a larger demand for zloty money. The implication of the variance of consumption is ambiguous and depends on the third derivative of the utility function, u‴.6 The expected utility excess return on foreign currency is positively related to the covariance between consumption and the return differential; the intuition is that if, for example, appreciations of the dollar tended to be associated with states of the world in which consumption is low (or, conversely, depreciations are associated with high consumption), the attractiveness of dollars as a store of value would be enhanced. The attractiveness of foreign money is also negatively associated with the covariance between inflation and consumption, since if, for instance, higher inflation were more likely to occur in states with lower consumption, the erosion of the return differential due to inflation would matter more. Finally, the attractiveness of dollars is positively related to the covariance between consumption and the return differential: a high return on dollars, if associated with periods of low consumption, would increase the attractiveness of holding dollars.

Next, consider the case in which dollars are expected to be a more attractive temporary abode of purchasing power—that is, where Eat > Eθt+1. In this case, under what circumstances will any zlotys be held? In equation (16), if Eat > Eθt+1, the first three terms are positive; it is thus the three covariance terms that can make EUER < 0.7 Thus, for any zlotys to be held, the covariance of consumption and the expected return differential on dollars must be sufficiently positive, the covariance of consumption and inflation must be sufficiently negative, or the covariance of inflation and the return differential must be sufficiently positive to make the right-hand side of equation (16) negative. For instance, in an economy in which supply shocks are prevalent and where adverse supply shocks result in low consumption, high inflation, and rapid depreciation of the zloty, the second and third conditions would be met. If demand shocks result in an association between high consumption, high inflation, and depreciation of the zloty, this could satisfy the first and third conditions. Thus, some combination of supply and demand shocks might lead utility-maximizing households to hold some zlotys, even if their expected return—allowing for the fact that zlotys can be exchanged for goods at controlled prices in the official stores—is lower than that on dollars.

After some manipulation of equation (16), using equation (3), the demand for domestic money can be derived and expressed as follows:8

This demand for money, derived from a stylized optimizing framework, combines the considerations underlying the transactions and precautionary demand for money—the fact that domestic currency is needed for purchases in the official shops, but that the quantity of goods available for purchase on any given day is uncertain—with the portfolio approach, as reflected in the role of the variances and covariances of relative returns on the two monies and on goods.

In the remainder of the paper, the demand for money will be analyzed empirically. In order to lay the groundwork for this analysis, it is also necessary to discuss another issue: whether and how the money market clears. The traditional view, discussed in the introduction, implies that, although desired money balances may be determined as in this model, actual money balances may differ from desired levels, because there are shortages, and not enough variables can adjust to equate actual and desired balances. The present model, to the contrary, suggests several variables that can adjust to clear the money market. First, individual households can adjust their money balances by varying their expenditures on goods in the official shops, to the extent that such goods are available—that is, if constraint (6) is not binding; such expenditures would affect the total stock of money, in nominal terms, held by the household sector. The authorities may, to some extent, adjust the official prices of goods, but these adjustments may be expected to take place gradually.

Individual households can also vary their expenditures on goods in the black market; however, in this model, black market prices adjust to clear that market, so that, in the aggregate, adjustment takes place primarily through prices rather than gradually through quantities. Households may also attempt to adjust between holdings of domestic currency and foreign exchange; although in the short run the unofficially held stock of foreign exchange may be relatively inflexible from the standpoint of the economy as a whole, the black market exchange rate, et, adjusts to clear the market, thus affecting the demand for money via the intertemporal budget constraint (3). The framework also implies some persistence in desired real balances, since past money balances also enter the intertemporal budget constraint; however, this will not delay the adjustment of aggregate money balances to the extent that other variables can adjust relatively quickly (Lane (1990)).

In summary, the framework presented here suggests that equilibrium in household money holdings may be achieved despite shortages: in the short run, through adjustment of household expenditures as well as black market prices and exchange rates; and in the long run, through adjustment of official prices and of the quantities of goods in the black market and the stock of unofficially held foreign exchange. These considerations suggest that, if the present model is relevant, one would not necessarily expect to see particularly slow adjustment of money balances in a socialist economy.

II. The Data: Poland in the 1980s

In the remainder of the paper, the model described in the previous section will be explored empirically, with data for the household sector in Poland over the period December 1979 to October 1988. This was a turbulent period in the Polish economy, beginning with a phase of political, economic, and social unrest associated with the Solidarity movement. These developments were short-circuited by the imposition of martial law on December 13, 1981, Both the Solidarity and the martial law years were accompanied by economic reforms, which not only attempted to bring about needed price adjustments to rectify macro- and microeconomic imbalances, but also to modify the system of central planning with a view to ending the economic stagnation (Balcerowicz (1989)). By and large, these reforms were unsuccessful, and a “second stage of reform” was launched with a referendum, held in November 1987, to approve the government’s program. Although the government lost the referendum, it nonetheless introduced substantial price adjustments in the so-called price-wage maneuver of February 1988—a price increase accompanied by a smaller increase in wages. The failure of this maneuver inaugurated the deterioration of economic conditions that led up to the economic crisis of 1989, with the slide into hyperinflation and the accompanying political revolution.

The data set that will be used for this period was assembled at the National Bank of Poland, and has already been used for some empirical analysis of the demand for money (Olés and others (1987)).9 Household sector figures were derived from those for the nonsocialized sector by adjusting for the money balances of nonsocialized enterprises.

The empirical work will concentrate on narrow zloty money, whose behavior over the sample period is depicted in Figure 1. Why narrow? Because a substantial proportion of zloty time and savings deposits held by households consisted of specially earmarked deposits that needed to be set aside for specific large purchases, such as automobiles; such holdings of money are not amenable to explanation in terms of the usual analysis of the demand for a medium of exchange. Why zloty? Because foreign currency money includes not only officially registered foreign currency deposits, but also mattress money, and deposits made informally (and illegally) in foreign banks; no reliable data exist on the latter quantities. Moreover, the regulations for foreign currency deposits were liberalized, drawing money out of the mattresses into the banks. In particular, foreign currency deposits by Polish residents originally required a declaration of the source of the funds (for example, royalties, copyright, gifts from nonresidents); then, beginning in April 1985, a new form of account was introduced, which could be freely credited with convertible currency without a declaration of source, but funds in these accounts could not be transferred or used to finance travel abroad. In July 1988 the two types of accounts were consolidated, and the restrictions on the use of funds abolished. Meanwhile, the state savings bank issued deposit certificates in U.S. dollars, which, beginning in November 1987, it undertook to convert into zlotys at close to the parallel exchange rate; and effective June 16, 1988, it agreed to buy or sell these certificates in exchange for zlotys. These changes in regulations—which may well have affected the relative attractiveness of different ways (legal or illegal) of holding foreign currency, without much affecting the total amount of foreign currency held—would make it unlikely that registered foreign currency deposits were a stable proportion of total foreign currency holdings of Polish households.

Widespread unofficial holdings of foreign currency (especially dollars) by the household sector were associated with a very active parallel or black market for foreign exchange. Although this market was not strictly legal until March 1989, black market exchange rates were widely known and quoted (even by government officials), and there are good data on these rates. Exchange rates are shown in Figure 2; black market rates are of the greatest relevance to the decisions of households, but the official rate is also shown as a reference point. The black market rate shows a rapid depreciation of the zloty in the second half of 1981, associated with the political crisis; the rate dropped again in connection with martial law (as well as the devaluation of the official zloty). From 1982 to 1987 there was a gradual official depreciation of the zloty in terms of the dollar, with small adjustments on a weekly basis to maintain the exchange rate in terms of a basket, interspersed with larger adjustments whose target was to guarantee the continuing profitability of 75 percent to 85 percent of convertible currency exports; beginning in April 1987, a new system was established, with smaller and more frequent adjustments. Throughout this period, the black market rate was at a substantial premium (generally 100 percent to 300 percent) vis-à-vis the official rate; this premium shot up dramatically in late 1987 and continued to rise rapidly, reaching over 500 percent in late 1988.

Figure 1.Household Narrow Money

(Billions of zlotys)

Figure 2.Exchange Rates

(Zlotys per dollar)

These exchange rate developments were taking place against the backdrop of variations in inflation. The retail price index, shown in Figures 3 and 4, is a notoriously incomplete measure of inflation because of the exclusion of goods purchased in the informal market, as well as, in some cases, the deliberate concealment of price changes behind changes in product classification—which together are known as “hidden inflation” (Nuti (1986) and Kalicki in Cassel and others (1989, pp. 73–91)).

Figure 3.Consumer Price Index

(1980 = 100)

Figure 4.Consumer Price Inflation

(Percent a month)

The retail price index does, however, show some general trends on a monthly basis. Official prices were held relatively steady until the martial law period, which was followed by rapid increases, peaking at a monthly rate of 21.8 percent—equivalent to an annual rate of 963 percent—in March 1982. Measured inflation followed a more regular pattern during 1983 to 1986, and then took off sharply with the reforms of late 1987 and early 1988, with monthly inflation reaching a temporary peak of over 15 percent—an annual rate of nearly 500 percent—in February 1988.

Net household income also showed substantial variation over the sample period, as shown in Figure 5.10 Measured incomes rose as a result of large wage settlements during the period 1979 to 1981; the imposition of martial law was followed by a drop in household incomes of about 35 percent from peak to trough. Incomes fluctuated, but on average edged up slightly during the rest of the 1980s, although never regaining their 1981 peak.

Figure 5.Net Household Income

(Billions of 1980 zlotys)

There is an important omission from the data set: data are not generally available on the premium of black market commodity prices above official prices, θt, a premium that affects the relative attractiveness of zloty as opposed to foreign money, as shown in the model in Section I. To the extent that this premium was relatively stable and uncorrected with other variables in the model, the omission of this variable would not be a serious problem. However, the discussion in this section suggests that there were potentially (at least) two regime changes over the sample period: one associated with the imposition of martial law in December 1981, and the other with the inception of the second stage of reform in 1987. It will be important to consider whether these events were associated with structural breaks in the household money demand relationship that might arise from a change in the relationship between commodity prices on official and black markets.

The next section describes the method used to estimate the money demand relationship and reports the results.

III. Empirical Analysis

The household demand for money relationship given in equation (17) was estimated using Polish data from the period December 1979 to October 1988, as described in the previous section.

An important feature of the analysis is that, in the model in Section I the expected values, variances, and covariances of the inflation rate and the rate of appreciation of the black market exchange rate are important. A common approach to incorporating expectations in an empirical model is based on rational expectations: the realized values of the variables whose expectations appear in the model are taken as unbiased estimates of their expected values. This introduces an errors-in-variables problem, which is addressed using an instrumental variables procedure; variables whose realization was known when expectations were being formed are used as instruments. The realized values of the inflation rate, πt, and the appreciation of the black market exchange rate, at, are thus included as explanatory variables. In order to allow for the possibility that the variances and covariance of πt, and at varied over the sample period and played a role in household demand for money as the model in Section I suggests, the realizations of their squares and cross products are also included.

In estimating demand for money, dynamic structure is also particularly important, because it may shed light on the mechanism by which the money market is brought into equilibrium. This issue can be addressed via the two-stage error-correction model, as suggested by Hendry (1986), Granger (1986), and Engle and Granger (1987). The first step in this approach is to test the order of integration of the variables in the model. Dickey-Fuller tests were conducted, indicating that the log of household holdings of zloty currency and sight deposits, In mt, the log of the retail price index, In pt, the log of the black market exchange rate, In et, and the log of real net household income, In yt, were all I(1). Given these results, an error-correction approach can be sought: first a “static equation,” a relationship among the levels of the variables, is estimated;11 for this approach to be valid, nonstationarity of the residual must be rejected, which means, by definition, that the nonstationary variables included in the equation are cointegrated. In the second stage, a “dynamic equation” for the first difference of the dependent variable is estimated; all of the explanatory variables in this equation must be stationary. The dynamic equation includes the lagged disturbance term from the first-stage static equation as an error-correction term.

A static equation was first estimated. A cointegrating relationship was found between In mt, In pt, and In yt, provided that dummies were included to divide the three regimes discussed in Section II. Dummies were included for the period of unrest prior to the imposition of martial law on December 13,1981 (UNREST), and for the reform program that began after the November 1987 referendum (REFORM); in this context, the dummies indicate permanent shifts in money demand associated with a regime change. Initial estimates showed a coefficient on the log of the price level, In pt, that was very close to unity, as implied by economic theory; accordingly, this condition, implying a demand for real money balances, was imposed.12 The resulting static equation in the levels of the variables, estimated over the period December 1979 to October 1988, is (figures in parentheses are t-statistics)13

  • Sum of squared residuals = 0.2715

  • Standard deviation of dependent variable = 0.1129

  • Standard error of regression = 0.051

  • R2 = 0.7989

  • Adj. R2 = 0.7930

  • F(3, 103) = 136.40

  • Durbin-Watson = 1.22

  • Number of observations = 103.

This equation implies a long-run income elasticity of money demand of about 0.6, which is toward the low end of the range of estimates typically found for developed countries; this is consistent with the theoretical model presented in Section I, as well as with simpler models of the transactions demand for money. One interpretation of the positive coefficient on the dummy, UNREST, is that during the period leading up to martial law, the public’s demand for zloty money may have been augmented by fear that an impending regime change would include measures, such as a clampdown on the black markets, that would make dollar holdings less desirable. The negative coefficient on the dummy variable, REFORM, may indicate that the increases in administered prices introduced in connection with the reform program may have been expected to reduce the premium of black market prices over official prices (the unobserved variable, θt), thus reducing the incentive to hold zloty money as reflected in equation (16).

The static equation’s predictions of demand for money, in relation to the actual outturns, are shown in Figure 6. Deviations between actual and predicted values appear to be serially correlated, and this is confirmed by the Durbin-Watson statistic; this finding does not affect the consistency of a cointegrating equation, however. The residuals from this static equation (19) were examined for stationarity, and a Dickey-Fuller test rejected the hypothesis of a unit root (with a test statistic of 0 – 7.34). This indicates (as shown by Engle and Granger (1987)) that the dynamic behavior of the money stock can be represented in terms of an error-correction model.14

Figure 6.Real Household Narrow Money

(In logs)

A dynamic equation for Δln mt was accordingly estimated, including the lagged residual from the static equation, ECt-, as one of the explanatory variables. The log first differences of real income, Δln yt, the retail price index, Δln pt, and the black market exchange rate, Δln et, were also included. However, the discussion at the end of Section I suggests that Δln et is a variable that adjusts to clear the money market in the short run, and thus should be treated as endogenous. Moreover, as discussed in Section II, Δln pt omits prices in the unofficial sector, and therefore its measurements of the change in the relevant price level contain considerable error. In addition, first differences of the expected change in the black market exchange rate, ΔEat, where at = (In et+1 - In et), and of expected inflation, ΔEπt, were included, as implied by the theoretical model; these were represented, through the assumption of rational expectations, by their respective realized values. Furthermore, as suggested by equation (16), the conditional variances of inflation and the black market exchange rate, and their conditional covariance, were also included as explanatory variables; using the same rational expectations assumption, they were represented by their realized values, Δ(πt2),Δ(at2), and Δ(atπt). Lagged values of all of the variables, including 12 monthly dummies, were used as instruments.15

Initial explorations confirmed the appropriateness of the error-correction model, since the error-correction term turned out to be statistically significant and negative. Moreover, coefficients on the first differences of prices, income, and the black market exchange rate were all significant and of the expected sign. No statistically significant effect of changes in expected inflation, ΔEπt, and expected exchange rate appreciation, ΔEat, could be found, however. This is perhaps not surprising, given the difficulty of forecasting prices. The variance of inflation also turned out to have no statistically significant effect—also perhaps unsurprising, given the theoretical ambiguity of the sign of this variable’s influence.

Monthly dummies were initially used in the equation. It was discovered that the only important seasonal effects were a December minus January effect and a March effect; one could not reject the hypothesis that the other monthly dummies and constant term could be excluded from the equation (F(10,84) = 0.49).

Accordingly, the following, more parsimonious specification of the equation was estimated for the March 1980 to September 1988 period (figures in parentheses are t-statistics):16

  • Sum of squared residuals = 0.04569

  • Standard deviation of dependent variable = 0.03500

  • Standard error of regression = 0.02205

  • R2 = 0.6414

  • Adj. R2 = 0.6109

  • F(9,99) = 20.06

  • E’PZ*E = 0.611109E-02.

Several features of this estimated equation can be noted. The equation seems to track changes in money balances fairly well, as suggested by Figure 7. The error-correction term shows up strongly, suggesting that household money balances do in fact adjust toward their long-run real level as represented in equation (18); this corroborates the reintegration results from the static model, and contradicts the notion—associated with the liquidity overhang story discussed in the introduction—that households’ money holdings are largely passive. The error-correction coefficient implies a mean lag of about one month in response to a change in the price level.17 This rapid adjustment is striking, and is broadly consistent with results obtained for Polish data by Hartwig (1987) and Olés and others (1987). However, the lag in adjustment is shorter than that obtained in many money demand studies for other economies apparently displaying much less structural rigidity.18 This suggests that Polish households were able to adjust their holdings of zlotys fairly rapidly in response to changes in the determinants of their desired holdings, or that black market exchange rates adjusted quickly to clear the money market.19

Figure 7.Changes in Nominal Money

(In logs)

Also of particular importance is the conditional variance of exchange rate movements and their conditional covariance with inflation, represented by Δ(atπt) and Δ(at2). The variance of exchange rate movements had a significantly positive effect, suggesting that (for a given domestic price level) a greater variance of the black market exchange rate makes dollars a less attractive temporary store of value for domestic residents.20 The covariance term, whose sign is theoretically ambiguous, turned out to have a significant negative effect on the change in money holdings. The intuition behind this result is that a higher covariance of the exchange rate and inflation makes dollars a better inflation hedge, and thus reduces the incentive to hold zlotys. These results corroborate the dynamic portfolio approach developed in Section I. They also stand in contrast to previous empirical work on market economies, which has not generally accorded an important role to such variance and covariance effects, even though such effects are an implication of the portfolio approach that was long part of the standard equipment of monetary theory (Laidler (1976)).

Why, then, do such effects show up in Poland in the 1980s? Two reasons come to mind. One is that, in Poland there are few alternatives to money with similar risk characteristics, so risk considerations may have a more noticeable effect on the demand for money than in developed market economies. Second, during the sample period there were substantial changes in the Polish economy, which quite conceivably could have brought about substantial enough changes in the variances and covariances of prices and exchange rates to have an influence on the behavior of individual households.

The next step is to test the stability of the estimated equation over the sample period. Two possible structural breaks associated with the possible regime changes will be considered: one is associated with the imposition of martial law in December 1982, and the other with the inception of the second stage of reform, following the referendum of November 1987. Structural breaks at either or both of these points would be unsurprising in terms of the theoretical model presented in Section I, since the model implies that changes in θt, the spread between prices in the black market and the official stores, would affect the demand for zloty money, given the observable explanatory variables. If a regime change alters the process generating prices and exchange rates, it would also be expected to alter the way in which expectations are formed.

The stability of the dynamic equation is first tested over the periods before and after the imposition of martial law—that is, between the March 1980–December 1981 and the January 1982–September 1988 subperiods. Here, the test statistic is 2.59, distributed as F(11,83); and the critical value is 1.91, so stability is rejected. The next step is to test whether the inception of the second stage reforms led to a structural break; that is, was the estimated relationship stable over the March 1980–October 1987 and November 1987–September 1988 subperiods? In this case, the test statistic is 2.00, so stability is again (although even more narrowly) rejected. Thus, significant, although not drastic, changes in structure appear to have been associated with both of these regime changes. This result is unsurprising in light of the extreme nature of the changes involved, as described in Section II. The imposition of martial law was accompanied by social disruption, massive price adjustments, and a sharp drop in household incomes, and the second stage was associated with, among other things, substantial price increases and liberalization of the conditions under which foreign currency deposits could be held. The result is also unsurprising in terms of the theory developed in Section I—which itself predicts that regime changes and the attendant changes in market conditions would likely be associated with structural changes in the demand for money.

This, of course, is not the end of the story. In late 1989 and through 1990, further and much more vigorous reforms were carried out (Lane (1992)). These reforms have entailed the freeing of many prices, beginning with most food prices in August 1989 and continuing through large administered price increases in October 1989 and the “big bang” freeing and/or adjustment of remaining prices in January 1990. These changes appear to have drastically reduced the gap between official and black market prices, so that by early 1990 food and other commodities were being sold in the burgeoning street market at prices that anecdotal evidence suggests were below official prices in some cases. Because the program of reform and stabilization was intended to alter the processes generating inflation and exchange rate movements in Poland and the way in which expectations are formed, it is not clear what the overall effect on demand for zloty money will be, but a further structural shift in money demand seems likely.

IV. Conclusion

This paper has developed a model of household demand for money in a reforming socialist economy such as Poland in the 1980s, in which the prevalence of black markets in commodities and foreign exchange has widened households’ range of choice. It suggests that, even though the Polish economy has been characterized by shortages in the official markets, household money demand within that economic environment can nonetheless be analyzed fruitfully in the framework of rational intertemporal choice, taking due account of the alternatives available to consumers.

The paper estimated the resulting money demand equation, using the error-correction framework to capture the dynamics. This empirical analysis yielded several important results. First, a cointegrating relationship was found between households’ real money balances and real incomes, allowing for intercept shifts associated with the imposition of martial law and with the inception of the second stage of reform. This result contradicts the traditional view that, in a socialist economy, money is largely passive. Since “real” here implies deflation by the official retail price index, this result also suggests that official prices adjust, in some degree, toward their equilibrium levels.

Another important aspect of the results pertains to the dynamic equation. This equation identifies an error-correction mechanism underlying the long-run relationship between money, prices, and output. The estimated error-correct ion coefficients suggest a rapid adjustment of household money balances toward their equilibrium levels. The dynamic equation also reveals a significant role for some other variables. The black market exchange rate is important, as suggested by the theoretical model and consistent with the view that the black market exchange rate adjusts to equilibrate the money market, given the sluggishness of official prices. Moreover, the variance of exchange rate changes and their covariance with inflation were found to have a significant influence on demand for money over this period. These factors may have been important because of the lack of alternatives to money as a store of value in Poland at this time and also because of the turbulence of the Polish economy, which may well have led to significant alterations in the variance-covariance structure of returns on domestic and foreign money in relation to goods.

The stability of the equation is narrowly rejected over both of the regime changes examined. This result, while disappointing, is unsurprising given the drastic nature of these regime changes; it is also consistent with the theoretical model, which suggests a rationale in terms of some omitted variables—notably, the ratio of black market and official prices of goods—which were clearly altered with these regime changes, and would be expected to have an important influence on demand for money.

Reforms in Poland, which have freed most prices, have greatly changed the transactions environment, and quite probably led to a further shift in demand for money. The framework presented in this paper may therefore now be more readily applicable to economies at an earlier stage in the reform process—and notably to the emerging states of the former Soviet Union.

This paper, in both its theoretical and empirical parts, contradicts the view—associated with the notion of a liquidity overhang—that household money holdings in a socialist economy can be regarded as adapting passively to government decisions on wages and on the availability of consumer goods. This conclusion suggests that the monetary economics of socialist economies is a subject that has been unfairly neglected.


    AbelIstván and István P.Székely“Money and Causality in CPEs (The Case of Hungary),”Department of Economic Planning and Modelling Working Paper 88/3 (Budapest: Karl Marx University of EconomicsAugust1988).

    BabaYoshihishaDavid F.Hendry and Ross M.Starr“U.S. Money Demand 1960–1984” (unpublished; Oxford: Nuffield College1987).

    BalcerowiczLeszek“Polish Economic Reform, 1981–88: An Overview,”in Economic Reforms in the European Centrally Planned EconomiesUnited Nations Economic Commission for Europe Economic StudiesNo. 1 (1989) pp. 11722.

    BirmanIgor“The Financial Crisis in the USSR,”Soviet StudiesVol. 32 (January1980) pp. 84105.

    BirmanIgor and Roger A.Clarke“Inflation and the Money Supply in the Soviet Economy,”Soviet StudiesVol. 37 (October1985) pp. 494504.

    BoyckoMaxim“The Velocity of Money in a Shortage Economy” (unpublished; Moscow: Institute of the World Economy and International RelationsSeptember1991).

    BurtonDavid and JimingHa“Economic Reform and the Demand for Money in China,”IMF Working Paper 90/42 (Washington: International Monetary FundApril1990).

    CasseiDieterWladislawJaworskiDietmarKathTadeuszKierczynskiKarolLutkowski and Hans-JoachimPaffenholzeds.Inflation und Schatten-wirtschaft in Sozialismus (Hamburg: S&W Steuer-und Wirtschaftsverlag1989).

    CharemzaWojciech W. and SubrataGhatak“Demand for Money in a Dual-Currency, Quantity-Constrained Economy: Hungary and Poland, 1956–1985,”Economic JournalVol. 100 (December1990) pp. 115972.

    ClowerRobert W.“A Reconsideration of the Microfoundations of Monetary Theory,”Western Economic JournalVol. 6 (1987) pp. 19.

    CottarelliCarlo and MarioBlejer“Forced Savings and Repressed Inflation in the Soviet Union; Some Empirical Results,”Staff PapersInternational Monetary FundVol. 39 (June1992) pp. 25686.

    EngleRobert F. and Clive W. J.Granger“Co-integration and Error Correction: Representation, Estimation, and Testing,”EconometricaVol. 55 (March1987) pp. 25176.

    FeltensteinAndrewDavidLebow and Swedervan Wijnbergen“Savings, Commodity Market Rationing, and the Real Rate of Interest in China,”Journal of Money Credit and BankingVol. 22 (May1990) pp. 23452.

    GrangerClive Y. J.“Developments in the Study of Cointegrated Economic Variables,”Oxford Bulletin of Economics and StatisticsVol. 48 (August1986) pp. 21327.

    HartwigKarl-Hans“Theoretische und empirische Ansätze zur Bestimmung der Geldnachfrage in sozialistischen Planwirtschaften” in Geldtheorie: Entwichlung Stand und systemvergleichende Anwendunged. by H.Jörg Thieme and others (Baden-Baden: Nomos Verlagsgesellschaft1985).

    HartwigKarl-Hans“Die Geldnachfrage in sozialistischen Planwirtschaften: Polen 1954–1984,”Kredit und KapitalNo. 3 (1987) pp. 378401.

    HendryDavid F.“Monetary Economic Myth and Econometric Reality,”Oxford Review of Economic PolicyVol. 1 (1985) pp. 7284.

    HendryDavid F.“Econometric Modelling with Cointegrated Variables: An Overview,”Oxford Bulletin of Economics and StatisticsVol. 48 (August1986) pp. 20112.

    KornaiJanosEconomics of Shortage,Vol. B (Amsterdam; New York: North-Holland1980).

    KremersJeroen J. M. and Timothy D.Lane“Economic and Monetary Integration and the Aggregate Demand for Money in the EMS,”Staff PapersInternational Monetary FundVol. 37 (December1990) pp. 777805.

    LahiriAshok“Money and Inflation in Yugoslavia,”Staff PapersInternational Monetary FundVol. 38 (December1991) pp. 75188.

    LaidlerDavid E. W.The Demand for Money: Theories and Evidence,2nd ed. (New York: Harper & Row1976).

    LaidlerDavid E. W.“On the Demand for Money and the Real Balance Effect,”in Monetarist Perspectives (London: Philip Allan1982) pp. 3766.

    LaneTimothy D.“Costly Portfolio Adjustment and the Short-Run Demand for Money,”Economic InquiryVol. 28 (July1990) pp. 46687.

    LaneTimothy D.“Inflation Stabilization and Economic Transformation in Poland: The First Year,”Carnegie-Rochester Conference Series on Public PolicyVol. 36 (July1992) pp. 10556.

    LucasRobert E. Jr.“Equilibrium in a Pure Currency Economy,”Economic InquiryVol. 18 (April1980) pp. 20320.

    LucasRobert E. Jr.“Money in a Theory of Finance,”Essays on Macroeconomic Implications of Financial and Labor Markets and Political Processesed. by KarlBrunner and Allan H.MeltzerCarnegie-Rochester Conference Series on Public PolicyVol. 21 (Amsterdam: North-Holland1984) pp. 945.

    MihaljekDubravko“Wealth Effects of Inflation and Devaluations in Yugoslavia: Who Gains and Who Loses?”Economic Analysis and Workers’ ManagementVol. 23No 3 (1989) pp. 20752.

    NutiMario“Hidden and Repressed Inflation in Soviet-Type Economies: Definitions, Measurements and Stabilization,”Contributions to Political EconomyVol. 5 (March1986) pp. 3782.

    NutiMario“Internal and External Aspects of Monetary Disequilibrium in Poland,”in Economic Transformation in Hungary and Poland European EconomyNo. 43 (March1990) pp. 18196.

    OlésMarekJerzyPawilno-Pacewicz and MariuszSumlinski“Popyt Gospodarstw Domowych Na Pieniadz,”FinanseNo. 10 (1987) pp. 2133;English version entitled “Households Money Demand Function for Poland, 1960–1986” (unpublished; Warsaw: National Bank of Poland. 1987).

    PayneJames E.“A Model of Money Demand in Yugoslavia: 1952–1985,”Economic Analysis and Workers’ ManagementVol. 24No. 1 (1990) pp. 8394.

    PindákFrantisek“Inflation Under Central Planning,”Jahrbuch der Wirt-schaft Osteuropas/Yearbook of East-European EconomiesVol. 10 (1983) pp. 93131.

    PodkaminerLeon“Disequilibrium on Poland’S Consumer Markets: Further Evidence on Intermarket Spillovers,”Journal of Comparative EconomicsVol. 12 (March1988) pp. 4360.

    PodkaminerLeon“Macroeconomic Disequilibria in Centrally Planned Economies: Iden-tifiability of Econometric Models Based on the Theory of Household Behavior under Quantity Constraints,”Journal of Comparative EconomicsVol. 13 (March1989) pp. 4760.

    PortesRichard“On the Theory and Measurement of Macroeconomic Disequilibrium and Shortage in Centrally Planned Economies,”in Models of Disequilibrium and Shortage in Centrally Planned Economiesed. by ChristopherDavis and WojciechCharemza (London; New York: Chapman and Hall1989).

    PortesRichardRichard E.QuandtDavidWinter and StephenYeo“Macroeconomic Planning and Disequilibrium: Estimates for Poland, 1955–1980,”EconometricaVol. 55 (January1987) pp. 1941.

    SzékelyIstván“The Role of Money and Monetary Policy in a Modified CPE (The Case of Hungary)” (unpublished; Cambridge: Cambridge UniversityJanuary1990).

    TysonLaura D’Andrea“Enterprise Demand for Money and Illiquidity Crises in Yugoslavia,”European Economic ReviewVol. 12 (February1979) pp. 5371.

    WeitzmanMartin L.“Price Distortion and Shortage Deformation, or, What Happened to the Soap?”American Economic ReviewVol. 81 (June1991) pp. 40114.

Timothy D. Lane, an Economist in the Research Department, received his doctorate from the University of Western Ontario.

The author wishes to thank Pierre-Richard Agénor, James Boughton, Fabrizio Coricelli, Marek Olés, Rolando Ossowski, Jacek Rostowski, Mariusz Sumlinski, and participants in seminars at the University of Florida and the World Bank for helpful comments.

The cash-in-advance approach has been used in many studies, few of which, however, pertain to socialist economies. Some exceptions are Mihaljek (1989), who used cash-in-advance as a framework for money demand in Yugoslavia, and Boycko (1991), who used a similar approach to formalize the notion of a liquidity overhang.

Another motive for holding money that, for simplicity, will not be addressed here is target savings. In centrally planned economies, there is little consumer credit, so households have to accumulate the purchase price of consumer durables before purchase. For durable goods that are in short supply, such as automobiles, households are often also required to deposit funds m special accounts in order to be given a place on the waiting list. See, for example, Kalicki in Cassei and others (1989, pp. 195–212).

The error-correction approach has been applied to other socialist economies: in particular, to Yugoslavia by Lahiri (1991), and to China by Burton and Ha (1990).

Similarly, Tyson (1979) used the percentage growth of base money as an indicator of credit conditions in examining enterprise demand for money.

This ignores hoarding of commodity inventories, which are typically quite important in planned economies.

For example, in the case of an isoelastic utility function, u = cγ/γ, where 0 ≤ γ ≤ 1,u‴ = (γ - 2)(γ - 1)cγ-3 > 0, so the demand for zloty money depends positively on the variance of consumption.

This assumes that u‴ > 0, as in the isoelastic case.

The details of this derivation are presented in an earlier version of this paper, which is available from the author on request. A similar derivation is presented in Lane (1990).

Special thanks are due to Marek Olés for sharing these data.

Figures are deflated by the retail price index.

Ordinary-least-squares estimates of this equation are consistent even if the disturbance term is not well behaved—for example, even if it is serially correlated or is not orthogonal to the regressors.

The estimated coefficient on In pt is 0.9692, and the reported standard error is 0.0198, which implies that one cannot reject the null hypothesis of a unitary coefficient at a significance level of 0.05. Strictly speaking, though, hypothesis testing using this estimated standard error is not valid, due to the nonstationarity of the variables.

These/-statistics are reported purely as a matter of convention; it is recognized that the associated standard errors are invalidated by the nonstationarity of the data.

Estimation of a dynamic equation raises the possibility of autoregressive conditional heteroscedasticity (ARCH) in the exchange rate and inflation rate, so it is important to test that this does not also affect the residuals of the static equation. No evidence was found of ARCH(l) in the residuals of equation (18); the test statistic, distributed as F(1, 100), was 2.32.

In an earlier version of the paper, time-series predictions were used to represent Eπt and Eat. The conditional variances and covariances were proxied by the predictions of ARCH models using the squares and product of residuals from these time-series representations. The results were similar to those reported here.

Instrumental variables include Δln pt−1, Δln Pt−2, Δln et−1, Δln et−2, Δln yt, Δln yt−1, Δln yt−2, Δln mt−1, Δln mt−2, Δt−2, Δt−3, at−2, at−3, Δt−2at−2, Δt−3at−3, ECt−1, and 12 monthly dummies.

In the presence of a lagged dependent variable as well as an error-correction term, the mean lag is (1 - β0 - β2)/β1, where β0 is the coefficient on the price level in the dynamic equation (given a unit coefficient in the static equation), β1 is the error-correction coefficient, and β2 is the coefficient on the lagged dependent variable.

Kremers and Lane (1990) found an error correction of -0.95 from quarterly data, implying a mean lag of about one month, for an aggregate of European Monetary System countries. Most country studies, however, report much smaller error-correction coefficients; for example, Hendry (1985) and Baba and others (1987) found error-correction coefficients, of 0.1 and 0.14, respectively, for quarterly data, implying much longer mean lags.

Price flexibility would lead to rapid adjustment of actual to desired money balances in the economy as a whole, even if adjustment at the level of the individual household were costly (Laidler (1982) and Lane (1990)).

Overheard in a Warsaw restaurant in November 1989: “No, we couldn’t possibly accept dollars; the dollar has been going up and down like a yo-yo!”

Other Resources Citing This Publication