A Brief History of Inflation in Poland, 1989-94

Figure 1 illustrates the precipitous increase in inflation in Poland in 1989.¹ By October 1989, CPI inflation had reached 55 percent a month, and, in early 1990, Poland was on the verge of hyperinflation (for the year, inflation averaged 585.8 percent). The stabilization program brought inflation down, but was accompanied by an unexpectedly high drop in output (see Ebrill and others, 1994). That period also witnessed marked changes in the terms of trade and in real wages.

View Full Size

Poland: Wage and Price Inflation

Sources: Polish authorities; and IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

During these chaotic developments on the macroeconomic front, movements in the overall CPI masked considerable divergences in relative prices. Past repressed inflation in the service sector suddenly materialized through a marked increase in relative prices. In particular, the drift in the price of services persisted in 1992, even though overall price inflation was tamed (see Figure 2). By 1993, however, it seemed that the bulk of relative price adjustment among food, services, and manufactured goods had already taken place.

View Full Size

Poland: Components of Consumer Price Inflation

Sources: Polish authorities; and IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

Relative price adjustment was, however, not over. At a more disaggregated level, relative price changes remain substantial although less visible. For instance, between January 1992 and May 1995, the relative cost of water tripled. Rental prices have also largely exceeded average inflation. Substantial relative price adjustment can therefore take place not only in the atmosphere of hyperinflation and stabilization, but also during spells of moderate inflation.

The causality between these relative price changes and inflation remains open to question. On the one hand, at low levels of inflation, even small changes in relative prices may have a large proportionate effect on inflation. On the other hand, annual inflation in the range of, say, 20-40 percent can also create large relative price shifts if certain prices are sticky.

The remainder of this section fleshes out the argument linking relative prices to inflation. Following Ball and Mankiw (1995), it is shown that there is a relationship between inflation and the skewness of the distribution of sectoral price changes and that the latter helps explain inflation. This skewness of the distribution can be seen as a proxy for relative price changes, suggesting that they are linked to inflation. To the extent that there is a need for further relative price adjustment—as prices catch up to international levels or as utility prices are increased to recover costs—this suggests a source of future inflationary pressure in Poland.

Relative Price Adjustment, Asymmetric Price Adjustment, and Inflation in Poland

In recent years, economists have paid renewed attention to relative price adjustments. This interest has been fueled by consistent statistical evidence that these adjustments are asymmetric (prices are more flexible upward than downward) and that relative price adjustments are correlated with the level of inflation. Fischer (1981) stressed, for instance, that U.S. inflation is highly correlated with a higher variance of sectoral price inflation. Similarly, Gerhaeusser (1988) found instantaneous Granger-causality between inflation and relative price volatility in Germany.²

The empirical relationship between relative price changes and inflation has until recently been regarded as an analytical puzzle. Indeed, at first glance, there need be no relationship between relative price changes and the overall price level (for example, if money determines the overall level of prices). According to this view, price increases in certain sectors would be offset by price decreases in others, leaving the overall price level in the hands of monetary policy. Of course, if prices were sticky downward, relative price changes could occur only in an inflationary context. But downward price adjustment is difficult to justify on theoretical grounds (unless adjustment costs are, for some reason, asymmetric). A solution to this puzzle must therefore be based on different grounds.

Such a solution has recently been proposed by Ball and Mankiw (1995), who argue that price stickiness (both upward and downward) arising from “menu costs” can cause higher inflation if the distribution of relative price changes is skewed.³ Their argument runs as follows.

In the presence of menu costs, firms will change their prices only if the costs of inaction exceed the menu costs associated with changing prices. Accordingly, small shocks to the economy will have little effect on the price level, because individual firms will be reluctant to change prices. Large shocks will have a disproportionate effect on individual prices—but not necessarily on the aggregate price level: if the distribution of shocks is symmetric, price increases and decreases at the firm level cancel each other out. Thus, greater variance of shocks need not lead to overall price inflation.

Rather, Ball and Mankiw’s approach can justify a relationship between the skewness of price changes and overall inflation. In a skewed distribution, most outliers are on one side of the distribution. Suppose that the distribution of relative price shocks is skewed positive (that is, there is a small number of large positive price shocks, offset by a large number of small negative price shocks). In the presence of menu costs, the large number of firms with small negative price shocks leave their prices unchanged, while firms with large price shocks increase their prices, raising the average price level. Conversely, if the distribution of price shocks is skewed negative, the overall price level will fall.

Empirical analysis of price dispersion in Poland during 1990-94 yields the following results. First, the distribution of price changes is heavily skewed, and not only during the first months of the transition. On average, once every two months there is some change in price more than two standard deviations away from the mean price increase. This fat-tailed distribution of price changes primarily reflects sizable price increases, largely concentrated in particular sectors (for example, utilities). Thus, there is considerable evidence of relative price adjustment within overall price level changes.

Second, monthly inflation (ΔCPI) is highly correlated with the standard deviation (SD) and skewness (SK) of the distribution of sectoral price changes that month.⁴

R²=0.59 DW=1.67

Sample period: December 1989–June 1995

Number of observations: 66

This result suggests that relative price shocks are closely linked to aggregate inflation.

However, causation is uncertain. Just as menu costs can explain why relative price increases can cause inflation, they can also explain why trend inflation owing to other causes can lead, in any given period, to a positively skewed distribution of price changes (Ball and Mankiw, 1994). First, in the presence of menu costs and trend inflation, those sectors facing relative price declines can simply avoid paying the menu cost, keep nominal prices unchanged, and allow overall inflation to generate the desired erosion in relative prices. In contrast, those sectors needing to establish relative price increases must increase prices over and above the inflation rate. With no change in certain sectors and increases in others, the result is a positively skewed distribution of price changes. Second, menu costs may differ across sectors. For example, assume that average long-term inflation does not differ between sectors. Sectors with small menu costs will change prices continuously, but those with large menu costs will change prices infrequently—but by large amounts when they do. The result is apparent skewness in the distribution of price changes, without implication for the long-term inflation rate.

The question is whether skewness causes inflation, or whether inflation causes skewness. In the first interpretation, skewness and relative price movements are part of the inflation process; in the second, they are simply an epiphenomenon. If the relative price explanation is correct, one should observe in the same sectors skewed price increases and a lasting improvement in their relative prices. If the second explanation is correct, the skewed price increases should be essentially random, without any lasting effect. To distinguish between these explanations, the change in a sector’s relative price (RCPI_i) between December 1989 and June 1995 was regressed against the number of times that sector registered a skewed price increase (NSK_i):

This cross-section regression indicates a strong correlation between a sector’s ability to generate a sustained increase in relative prices and the number of times its price changes are outliers. This is consistent with the claim that skewness captures lasting relative price changes and that these relative price changes were a key cause of Polish inflation.

The Dynamics of Wages, Prices, and the Exchange Rate

To gauge the extent of wage and price indexation, we estimated a standard markup model of the Polish economy. Given that the estimation period is restricted to March 1991-May 1995,⁵ simple dynamic specifications were favored. This exercise sheds some light on the nature of the interplay among prices, wages, and the exchange rate, as well as of the lags involved.

Structure of the Model

For the producer price index (PPI), the following markup pricing is assumed:

$\begin{array}{cc}PPI=\left(1+\mu \right){\left[\frac{P_{IC}IC+WL}{Y}\right]}^{\alpha }\mathrm{.}& \left(3\right)\end{array}$

Thus, producer prices are a markup on unit variable costs, defined as the sum of wage costs (WL) and the costs of intermediate inputs (P_ICIC), divided by output (Y).

Output affects prices through at least two distinct channels. The first is through the effect on marginal cost. Unless the production function has constant returns to scale, average and marginal productivity differ. For example, with increasing returns to scale and holding factor prices constant, marginal cost will decline as output increases. The second is through the effect on the markup. Some authors have argued that, in industrial economies, price competition fades somewhat during growth spurts. Therefore, the markup μ of prices on costs is a function of capacity utilization, simply proxied here by changes in output.

The impact of the exchange rate (exr) on producer prices is also twofold. First, it directly affects the cost of imported intermediate inputs. Second, foreign competition may also influence the markup of domestic firms: when the currency appreciates, margins shrink in those sectors most exposed to foreign competition. Overall, a reduced-form equivalent of equation (3) can be expressed as

$\begin{array}{cc}log\left(PPI\right)=a_1\log \left(W\right)+a_2\log \left(exr\right)+a_3\log \left(Y\right)+a_4\mathrm{.}& \left(4\right)\end{array}$

The CPI equation is largely an accounting identity. Most models simply assume that the CPI is a weighted average of producer price indices and import prices. This CPI equation also incorporates wages. This could be justified on the following grounds. First, consumer goods are often more labor intensive than typical industrial goods. Hence, simply using an aggregate PPI would underestimate the influence of wages. Second, wages also represent an important variable cost for retailers. Thus, for consumer prices, the following equation is assumed:

$\begin{array}{cc}\log \left(CPI\right)=b_1\log \left(PPI\right)+b_2\log \left(exr\right)+b_3\log \left(W\right)+b_4\mathrm{.}& \left(5\right)\end{array}$

For the labor market, a standard relationship between unemployment and real wages is assumed:

$\begin{array}{cc}\log \left(Wn\right)=\log \left(CPI\right)-c_1\log \left(UR\right)+c_2,& \left(6\right)\end{array}$

where UR represents the unemployment rate, and wages are considered net of tax.⁶

To capture the dynamics associated with high-frequency data, these equations were all estimated in an error-correction form. This involves estimating, first, a long-term (cointegrating) relationship between the various variables and, second, a short-term dynamic equation that determines the speed of adjustment toward the long-term target. Econometric results are reported in Table 1.⁷

^{Table 1}.

Poland: Wage and Price Equations

Note’. All variables in logarithms. Δ denotes first difference. Error-correction terms ecmw_-1, ecmp_-1, and ecmc_-1; are, respectively, the lagged residuals from the long-run wage, producer price, and consumer price equations of the previous table. Series names: Iwnr = real wage (deflated by consumer price index); lur = unemployment rate; Ippi = producer price index; lip = industrial production; lcu = unit labor costs; lexr= exchange rate; Icpi = consumer price index; and Iwg = gross wages.

^{Table 1}.

Poland: Wage and Price Equations

Iwnr	=	$\underset{\left(26.3\right)}{1.00}-\underset{\left(5.6\right)}{0.076}\mathit{lur}+\mathit{ecmw}$
R2	=	0.83 DW = 0.44
Ippi	=	$\underset{\left(5.4\right)}{-1.37}+\underset{\left(11.0\right)}{0.81\mathit{lip}}+\underset{\left(9.7\right)}{0.54\mathit{lcu}}+\underset{\left(3.3\right)}{0.24\mathit{lexr}}+\mathit{ecmp}$
R2	=	0.99 DW = 0.87
Icpi	=	$\underset{\left(12.2\right)}{-0.74}+\underset{\left(11.3\right)}{0.43\mathit{lwg}}+\underset{\left(9.8\right)}{0.56\mathit{lppi}}+\underset{\left(1.9\right)}{0.07\mathit{lexr}}+\mathit{ecmc}$
R2	=	0.99 DW = 0.96
Δlwn	=	$\underset{\left(14.5\right)}{-0.14}+\underset{\left(2.5\right)}{0.42}\mathit{\Delta lcpi}-\underset{\left(2.3\right)}{0.21}{\mathit{ecmw}}_{-1}$
R2	=	0.95 DW = 2.02
Δlppi	=	$\underset{\left(2.4\right)}{0.01}+\underset{\left(2.3\right)}{0.06}\mathit{\Delta lcu}-\underset{\left(1.3\right)}{0.06}\mathit{\Delta lexr}-\underset{\left(2.7\right)}{0.13}{\mathit{exmp}}_{-1}$
R2	=	0.50 DW = 2.03
Δlcpi	=	$\underset{\left(0.42\right)}{0.002}+\underset{\left(4.2\right)}{0.20}\mathit{\Delta lwg}-\underset{\left(3.7\right)}{0.58}\mathit{\Delta lppi}-\underset{\left(2.6\right)}{0.30}{\mathit{ecmc}}_{-1}$
R2	=	0.71 DW = 1.35

Note’. All variables in logarithms. Δ denotes first difference. Error-correction terms ecmw_-1, ecmp_-1, and ecmc_-1; are, respectively, the lagged residuals from the long-run wage, producer price, and consumer price equations of the previous table. Series names: Iwnr = real wage (deflated by consumer price index); lur = unemployment rate; Ippi = producer price index; lip = industrial production; lcu = unit labor costs; lexr= exchange rate; Icpi = consumer price index; and Iwg = gross wages.

View Table

^{Table 1}.

Poland: Wage and Price Equations

Iwnr	=	$\underset{\left(26.3\right)}{1.00}-\underset{\left(5.6\right)}{0.076}\mathit{lur}+\mathit{ecmw}$
R2	=	0.83 DW = 0.44
Ippi	=	$\underset{\left(5.4\right)}{-1.37}+\underset{\left(11.0\right)}{0.81\mathit{lip}}+\underset{\left(9.7\right)}{0.54\mathit{lcu}}+\underset{\left(3.3\right)}{0.24\mathit{lexr}}+\mathit{ecmp}$
R2	=	0.99 DW = 0.87
Icpi	=	$\underset{\left(12.2\right)}{-0.74}+\underset{\left(11.3\right)}{0.43\mathit{lwg}}+\underset{\left(9.8\right)}{0.56\mathit{lppi}}+\underset{\left(1.9\right)}{0.07\mathit{lexr}}+\mathit{ecmc}$
R2	=	0.99 DW = 0.96
Δlwn	=	$\underset{\left(14.5\right)}{-0.14}+\underset{\left(2.5\right)}{0.42}\mathit{\Delta lcpi}-\underset{\left(2.3\right)}{0.21}{\mathit{ecmw}}_{-1}$
R2	=	0.95 DW = 2.02
Δlppi	=	$\underset{\left(2.4\right)}{0.01}+\underset{\left(2.3\right)}{0.06}\mathit{\Delta lcu}-\underset{\left(1.3\right)}{0.06}\mathit{\Delta lexr}-\underset{\left(2.7\right)}{0.13}{\mathit{exmp}}_{-1}$
R2	=	0.50 DW = 2.03
Δlcpi	=	$\underset{\left(0.42\right)}{0.002}+\underset{\left(4.2\right)}{0.20}\mathit{\Delta lwg}-\underset{\left(3.7\right)}{0.58}\mathit{\Delta lppi}-\underset{\left(2.6\right)}{0.30}{\mathit{ecmc}}_{-1}$
R2	=	0.71 DW = 1.35

Note’. All variables in logarithms. Δ denotes first difference. Error-correction terms ecmw_-1, ecmp_-1, and ecmc_-1; are, respectively, the lagged residuals from the long-run wage, producer price, and consumer price equations of the previous table. Series names: Iwnr = real wage (deflated by consumer price index); lur = unemployment rate; Ippi = producer price index; lip = industrial production; lcu = unit labor costs; lexr= exchange rate; Icpi = consumer price index; and Iwg = gross wages.

A Highly Indexed Economy

Poland is a well-indexed economy and therefore prone to relatively high inflation. Such findings are not unexpected after a hyperinflation spell (see Commander and Coricelli, 1991). However, certain aspects of the econometric results deserve specific comment.

First, wages are much more indexed than prices. After a consumer price shock, the mean lag of wage adjustment amounts to about two months. Hence, an inflationary shock to the CPI or wages is likely to reduce enterprises’ markups rather than real wages. This result may seem counterfactual given that, at the beginning of the transition, average real wages declined strongly. However, there are various ways to reconcile these two facts. For one, Polish households may have learned from the hyperinflation period how to protect themselves against price changes. Also, in addition to the precipitous decline in productivity, the initial real wage decline was facilitated by centralized negotiations, quite independently of any inflationary surprise.⁸

Second, the producers’ markup reacts rapidly to changes in output and is negatively affected by sudden changes in costs. The first feature may partly explain the growth of the PPI in early 1995 given the brisk pace of output growth. The magnitude of the effect of output on prices is, however, surprising. It may have resulted from a shortage of capital in industry during the recovery of output since 1991 and, subsequently, higher margins accompanying the boost in demand. As expected, exchange rate appreciation compresses producers’ margins. However, such an impact on inflation materializes only with a mean lag of seven to eight months. The pass-through from unit labor cost is somewhat more rapid.

Third, CPI inflation is driven mainly by wages and the domestic PPI. Possible explanations for the strong direct impact of wages on the CPI have already been mentioned. The main surprise is the relatively weak role of the exchange rate: the direct impact appears relatively small, and the mean lag is three to four months. A more disaggregated analysis shows the exchange rate’s impact is statistically significant only for manufactured consumption goods, but not for foodstuffs or services.

A casual inspection of the structure of Polish imports provides an explanation for such findings (Table 2). Before the transition, imports of consumer goods were constrained to fairly low levels in Poland, as in other Eastern European countries. Since then, a strong appetite for Western products combined with a more liberal trade policy has boosted imports of consumer goods. Their share in total imports has more than doubled since 1988 to reach approximately 20 percent.⁹ As such, since imports and consumption represent, respectively, 25 percent and 70 percent of GDP in 1994, an appreciation of the currency by 10 percent mechanically reduces the CPI only by 0.7 percent.¹⁰ This result is in line with our econometric estimates. Overall, the stronger impact of the exchange rate on the PPI than on the CPI mainly reflects the large share of intermediate inputs in total imports (Table 2).

^{Table 2}.

Structure of Polish Imports, 1994

Source: Polish Statistical Office.

^{Table 2}.

Structure of Polish Imports, 1994

Goods	In Percent of Total Imports
Investment goods	13.6
Raw materials	66.3
Consumer goods	19.3

Source: Polish Statistical Office.

View Table

^{Table 2}.

Structure of Polish Imports, 1994

Goods	In Percent of Total Imports
Investment goods	13.6
Raw materials	66.3
Consumer goods	19.3

Source: Polish Statistical Office.

It would be utterly misleading, however, to infer from this that the exchange rate is useless in controlling inflation. Indeed, the weak initial impact from the exchange rate on both CPI and PPI is magnified by the high indexation in the economy. To illustrate this point, this model was simulated in two different ways. The first simulation simply consists of a 10 percent devaluation. Consumer prices increase by 2.5 percent after one year and by about 6 percent after three years. In the second simulation, it is assumed that, after the initial shock, the exchange rate is depreciated in line with inflation to offset the loss of competitiveness. This would amount to adding to the previous three equations the following one:

which represents an accommodating policy reaction function. With this addition to the model and, by setting d₁ equal to 0.5, assuming only a partly accommodating exchange rate policy, the inflationary effect of the devaluation is severely compounded and exceeds 10 percent. Thus, despite its slow impact, the exchange rate is key as an anchor.

Finally, the wage-price model was used to gauge the inflationary impact of energy price adjustments, to illustrate the interaction among relative price changes, indexation, and inflation. Consider the effect of a 60 percent increase in utility prices—at the lower end of the range that Freund and Wallich (1995) consider necessary for Poland. At first glance, the immediate impact on the CPI would be approximately 6 percent, given the share of energy in the consumption basket. This effect would be stronger if one were to take into consideration the effects of utility price increases on the prices of other public services (for example, transportation) and on household rents.

In the well-indexed Polish economy captured in our wage-price model, any such energy price shock would also induce a considerable multiplier effect. With fixed exchange rates, such a shock would increase the overall level of prices by approximately 20 percent after a four-year period. Additional exchange rate depreciation would compound the inflationary impact. With a partly accommodating exchange rate rule (d₁= 0.5), consumer prices would increase by 40 percent after 4 years and by 70 percent after 10 years.¹¹

This exercise in itself need not imply that inflation should be high if energy prices are liberalized, or if there are other market relative price increases. Indeed, one could argue that a tighter monetary policy would keep overall inflation subdued. What is important, however, is that such an outcome would hinge on a severe price adjustment in other sectors. In the face of nominal price rigidity, changes in relative prices can be achieved only through higher inflation.

Inflation and Monetary Policy in a Restructuring Economy

So far, three potential sources of entrenched inflation in Poland have been isolated: a drift of certain relative prices, de facto indexation of wages, and some evidence of an accommodating monetary policy. In this section, it is argued that the nature of Polish firms’ restructuring process may be an ultimate underlying cause. Indeed, certain firms have restructured largely through attrition, accompanied by gradual output contraction and accumulated losses. It is argued that, because of this slow restructuring, firms are likely to increase their relative prices or to run deficits. As a result, they are less responsive to traditional macropolicy instruments. Therefore, to the extent that such firms represent a sizable proportion of output, we show that a “hard-currency” strategy to reduce inflation may be unexpectedly costly.¹⁴

The move to a market economy was accompanied by a relative price shock of unprecedented magnitude. This shock created a fault line between two groups of firms: those that benefited from it and those for which it proved detrimental.¹⁵ In the face of this shock, restructuring in the sectors adversely affected can be expected to follow one of three paths: (1) cutting wages to restore profitability; (2) boosting productivity by shedding labor or, in certain (rare) instances, by investing in new equipment; or (3) a “wait-and-see” approach, where firms merely shrink through attrition. In this situation, they run losses and possibly increase their prices. In a sense, (3) is a slower version of (2), leading to a depletion of the firms’ assets through accumulated losses. The wait-and-see approach seems to capture certain traits of the restructuring process in some sectors of the Polish economy. Indeed, Pujol (1996) showed that, during the early transition years, Polish firms tended to adjust employment levels rather than change relative wages. Employment has fallen most in the loss-making sectors, but did not lead to profitability. He argues that, during this period, the hard budget constraint was not always sufficiently binding to force firms into needed adjustments.

However, it is important to bear in mind two things. First, this characterization only applies to sectors in need of restructuring and that have delayed it. Privatized or newly created firms have been less prone to procrastination and, in certain cases, have made giant strides toward Western productivity standards. In essence, the transition created a dual economy.¹⁶ Second, a pure wait-and-see approach is unlikely given that no firm is totally immune from restructuring pressure. As we discuss below, firms are better sheltered if they can secure additional financing (for example, from rolling over loans, tax arrears, or inter enterprise credits). What the above description emphasizes, however, is that the process is relatively slow because managers try to shelter workers from painful wage adjustment or labor shedding. A similar point is made by Berg and Blanchard (1994).

The coal mining sector provides an interesting illustration of this behavior. First, the coal industry has restructured, as evidenced by sizable productivity gains (Table 5). These gains have certainly exceeded what a pure wait-and-see approach would have delivered and have contributed to the improvement in the sector’s profitability. However, despite substantial losses, coal miners have improved their relative wages, which still stand strikingly above the average for industry as a whole. In other words, a large fraction of productivity gains have been erased by additional wage costs. These developments have three main implications. First, the price of Polish coal has increased despite a 15 percent decline in international prices between 1991 and 1994. Second, the sector’s assets have been severely eroded, threatening its long-term viability. Third, the completion of restructuring has been delayed. A recent report issued by the Polish Ministry of Mining now considers that restructuring may take another five years and involve the loss of another 90,000 jobs.¹⁷

^{Table 5}.

Restructuring in the Coal Industry

Source: Polish Statistical Office, Biuletyn Statystyczny, various issues. Note: “Employment” is defined as the change in average employment over the year. The coverage of the series changes in 1992. “Relative wage” is the ratio of gross wages in the coal sector to average gross wages in industry. “Relative price index” represents the change in the relative price of sold production. “Profitability” is the ratio of net losses to total income. “Assets-liabilities” is the ratio of short-term assets to short-term liabilities.

^{Table 5}.

Restructuring in the Coal Industry

	1991	1992	1993	1994
Change in employment (in percent)	9.5	-4.9	-14.3	-7.6
Output index	108.5	100.0	97.6	100.4
Relative wage index	164.5	158.3	170.1	90.6
Change in relative price index (in percent)	40.9	2.2	16.9	14.5
Profitability (in percent)	-4.9	-23.0	-10.4	-0.8
Assets-liabilities	1.4	—	0.6	0.75

Source: Polish Statistical Office, Biuletyn Statystyczny, various issues. Note: “Employment” is defined as the change in average employment over the year. The coverage of the series changes in 1992. “Relative wage” is the ratio of gross wages in the coal sector to average gross wages in industry. “Relative price index” represents the change in the relative price of sold production. “Profitability” is the ratio of net losses to total income. “Assets-liabilities” is the ratio of short-term assets to short-term liabilities.

View Table

^{Table 5}.

Restructuring in the Coal Industry

	1991	1992	1993	1994
Change in employment (in percent)	9.5	-4.9	-14.3	-7.6
Output index	108.5	100.0	97.6	100.4
Relative wage index	164.5	158.3	170.1	90.6
Change in relative price index (in percent)	40.9	2.2	16.9	14.5
Profitability (in percent)	-4.9	-23.0	-10.4	-0.8
Assets-liabilities	1.4	—	0.6	0.75

Source: Polish Statistical Office, Biuletyn Statystyczny, various issues. Note: “Employment” is defined as the change in average employment over the year. The coverage of the series changes in 1992. “Relative wage” is the ratio of gross wages in the coal sector to average gross wages in industry. “Relative price index” represents the change in the relative price of sold production. “Profitability” is the ratio of net losses to total income. “Assets-liabilities” is the ratio of short-term assets to short-term liabilities.

How relevant is this example? One may argue that the mining industry is an extreme case, reflecting largely the historical and political clout of miners in Polish society. In a sense, this is precisely the lesson drawn from this example: for whatever reasons, certain sectors have successfully circumvented the hard budget constraint. And, because these sectors represented a substantial fraction of industrial employment before the transition (approximately 12 percent for the coal mining sector), they are more than a statistical curiosity. Indeed, various studies have documented insiders’ resistance to painful restructuring. Aghion, Blanchard, and Burgess (1994) have modeled how certain groups can impose the status quo to avoid mass layoffs. Pujol (1996) also documents various rigidities in the labor market and the lack of reaction of relative wages to the financial situation of firms.

At a macroeconomic level, the slow adjustment to the new post-transition conditions is also well illustrated by the financial results of Polish enterprises. Polish firms restructured between 1992 and 1994, as evidenced by the recovery in profitability (Table 6). However, during this period, the proportion of loss-making firms, as well as their losses as a percentage of total income, remained sizable, consistent with the view that some firms adjusted by slowly reducing activity rather than by raising profitability.¹⁸ For the economy as a whole, losses have dwindled partly because some firms are profitable again, but also because loss-making firms have shrunk.

^{Table 6}.

Financial Results of Polish Enterprises

(in percent)

Source: Polish Statistical Office, Biuietyn Statystyczny, various issues.

¹In percent of total income.

^{Table 6}.

Financial Results of Polish Enterprises

(in percent)

		1989	1990	1991	1992	1993	1994
Gross profits1		34.6	23.1	4.6	2.1	2.8	4.5
Net profits1		22.3	10.6	-1.3	-1.5	-0.5	1.7
Loss-making firms
	Percentage of firms	n.a.	n.a.	46.3	49.2	45.3	39.6
	Share of income	n.a.	n.a.	28.6	29.5	29.4	19.6

Source: Polish Statistical Office, Biuietyn Statystyczny, various issues.

¹In percent of total income.

View Table

^{Table 6}.

Financial Results of Polish Enterprises

(in percent)

		1989	1990	1991	1992	1993	1994
Gross profits1		34.6	23.1	4.6	2.1	2.8	4.5
Net profits1		22.3	10.6	-1.3	-1.5	-0.5	1.7
Loss-making firms
	Percentage of firms	n.a.	n.a.	46.3	49.2	45.3	39.6
	Share of income	n.a.	n.a.	28.6	29.5	29.4	19.6

Source: Polish Statistical Office, Biuietyn Statystyczny, various issues.

¹In percent of total income.

What are the macroeconomic implications of sizable loss-making sectors undergoing only (relatively) slow transformation? To answer that fundamental question, we present a model in which relative price adjustment plays a key role. Such a model is somewhat unrealistic, but nonetheless useful: it provides a way to synthesize certain stylized facts concerning the transition.¹⁹ Its main conclusion is that the sector undergoing restructuring will gradually increase its relative prices as it reduces output and labor hoarding, adding to medium-term inflationary pressures. True, if the slowly restructuring sector faces a high price elasticity of demand, achieving such a relative price increase is unlikely, and the inflationary pressure is correspondingly lower. However, in these circumstances, the slowly restructuring sector will incur higher losses for a longer time. For given money supply growth, the credits that the slowly restructuring sector will require to stay afloat will unduly crowd out lending to the expanding sector, prolonging the restructuring process. If these credits remain in place, the burden of tighter monetary policy to reduce inflation will then fall disproportionately on the expanding sector.

Starting with the model, we assume that the economy comprises two distinct groups of firms, say, distressed (sector 1) and profitable (sector 2).²⁰ In sector 1, activity dwindles at a pace determined by the rate of attrition of the labor force. This rate of attrition includes an exogenous factor (say, retirement) and an endogenous one, linked to employment growth in sector 2. The intuition between such a linkage is as follows. Growing employment in the emerging sector is the main source of job creation. Therefore, separations from distressed firms are facilitated by an important pool of job creation. Conversely, quits tend to occur during a period of intense job destruction. As a result, during the transition in Poland, job-to-job movements have become increasingly important as a way to avoid unemployment spells (see Coricelli, Hagemejer, and Rybinski, 1995). Also, for certain occupational groups, intense job creation has led to shortages and, in turn, to higher wages to lure workers away from sector 1. This equation may therefore capture this wage competition between the two sectors.

Consider supply. In sector 1, at any instant, employment is assumed to determine output:

This implies that output in sector 1 is exogenous and does not respond to wage or price changes. However, employment declines by proportion Δ through natural attrition and by proportion β as labor demand increases in sector 2. For simplicity, it is assumed that this latter effect is minor (β ≈ 0). As employment in sector 1 declines, so does output.

In the profitable second sector, output is given by equating the real wage, w/P₂, to the marginal product of labor:

It is further assumed that nominal wages are given, so that increased prices of goods in sector 2 lower the real product wage and increase employment and output in sector 2.²¹ In contrast, employment in sector 1 is assumed to be excessive, so that the real wage always exceeds the marginal product of labor. In addition, it is assumed that labor shedding has only a limited impact on output. One measure of progress toward restructuring is the extent to which the marginal product of labor and the real wage in sector 1 can be brought closer together.²² With fixed nominal wages, progress can be achieved only by reducing employment or by increasing output prices.

The demand side of the model is standard and simply assumes a certain substitution of domestic demand for goods in the two sectors depending upon their prices (see Blanchard and Kiyotaki, 1987). It is assumed that the output of sector 1 is consumed entirely at home, whereas the output of sector 2 is both exported and consumed domestically. The share of domestic consumption of goods 1 and 2 as a proportion of domestic income depends on the relative price of the two goods:

where q = P₂ / P₁ Income = ΣP_iY_i,

and where g(.) and h(.) are, respectively, increasing and decreasing functions of q: if the relative price of good 2 increases, consumption of good 1 increases and consumption of good 2 declines. The sum of g(.) and h(.) is constant and is less than unity. Equivalently, households save a constant share of their nominal income.²³

It is assumed that the foreign demand for exports of the profitable firms of sector 2 declines as the ratio of the domestic producer price to the exchange rate increases:²⁴

To determine equilibrium, we equate supply with demand in the two goods markets:²⁵

and

The underlying mechanisms for the impact of prices on output can be summarized briefly. Demand for good 1 increases with a higher relative price of good 2, through a substitution effect; it is reduced by higher prices in sector 2 relative to the exchange rate because exports, and hence household income, diminish (income effect). Conversely, in sector 2, higher prices of good 2 relative to good 1 reduce domestic consumption, and higher prices relative to foreign prices reduce export demand; in addition, the marginal product of labor conditions determines total supply. Figure 4 depicts these two equilibrium loci.

View Full Size

Poland: Equilibrium in a Two-Sector Model

• Download Figure
• Download figure as PowerPoint slide

As time goes by, exogenous labor shedding lowers the supply of good 1. Barring any change in households’ demand, producers in this sector will gradually increase their relative price, restoring profitability but at the expense of a lower domestic market share. Conversely, sector 2 is faced with growing demand. Its firms react by increasing their prices slowly, leading to an increase in supply as well as a reduction in exports.²⁶

Overall, this setup provides a potential explanation for the gradual increase in the relative price of sector 1 during the transition. As sector 1 contracts, supply falls and the relative price of good 1 increases. As sector 2 grows, income increases, raising the demand for and relative price of good 1. This relative price shift is facilitated if the elasticity of substitution between goods 1 and 2 is low.

One may draw the following parallel. Some authors have argued that inflation is required in transition economies to eliminate the initial monetary overhang. In our setup, labor hoarding in sector 1 plays a similar role. As labor is dishoarded, output in sector 1 declines, and its relative price increases. However, the size of the relative price increase will depend on the size of the price elasticity of demand. Whether or not it leads to inflation will depend on the extent of monetary accommodation.

In this setting, a hard-currency option will deliver lower domestic prices only at a substantial real cost. Real appreciation combined with imperfectly indexed wages entails lower output growth in sector 2 and, subsequently, in the economy as a whole. Moreover, owing to subdued demand, firms in sector 1 have to postpone their relative price increase, and their profitability deteriorates. As a corollary, the end of the restructuring—when the marginal product and real wage equate in sector 1—is postponed. The greater the spillover from growth in sector 2 to declining employment in sector 1, the stronger this conclusion is. Note, however, that if sector 2 Is already at capacity, enjoying the benefits of an undervalued exchange rate and buoyant foreign demand, more pressure may be needed to promote genuine restructuring. In these circumstances, real appreciation is still appropriate.

Serious caveats apply. As presented, the model is well equipped only to describe the impact of slow restructuring when competition, possibly through foreign imports, is weak. This could well be the case in the energy sector, in nontradables, or where protection is strong. But there are other possibilities. First, in sectors where the price elasticity of demand is high or in sectors exposed to foreign competition, a marked relative price increase is not a viable option. Various segments of Polish industry seem to be in this alternative situation (for example, shipyards and tractors). Second, certain sectors may be producing goods that no one wants. In such a case, there will be no trend increase in relative prices, and there will be no demand spillover from growth in sector 2. Ultimately, there is no alternative to closure. In both of these two possibilities, firms must seek outside financing to cover their losses. In the next section, it is argued that financing losses in these sectors may render the conduct of monetary policy more difficult, hence indirectly hampering the fight against inflation.

Loss-making firms have mainly relied on three different sources of external financing: short-term loans, the accumulation of tax arrears, and interenterprise credits. A World Bank survey (Belka and others, 1995) indicated that, in state-owned firms, the payment of taxes and social security obligations had a much higher priority than obligations to banks or other firms. However, these authors also found that severely distressed firms tended to share the burden more equally between creditors; ultimately, they ended up accumulating substantial debt against the government. As a result, in addition to additional stress on the fiscal front, the Polish banking industry acquired many problem loans, with the potential for crowding out “good” borrowers.

Poorly performing loans have made up a substantial fraction of Polish banks’ assets. Moreover, these loans were highly concentrated in approximately 10 percent of Polish firms (Gomulka, 1994). Baer and Gray (1995) offer an interesting account of how Polish banks handled these nonperforming borrowers. Overall, banks seem to have had only limited leverage to push these firms into restructuring. For instance, these authors mention that only 50 percent of nonperforming borrowers in December 1991 had signed conciliation agreements with banks by April 1994 to avoid lengthy legal procedures. An even smaller fraction, 19 percent, had resumed debt service. As a result, banks are less likely to force companies into bankruptcy, because they come far down in priority as creditors (after employees and liabilities related to social security). Therefore, to keep themselves and the distressed sector afloat, banks will tend to roll over past loans.²⁷

Therefore, an immediate, undesirable outcome of slow restructuring is that it increases the exposure of the banking system to nonperforming loans.²⁸ Against this background, how might banks rebalance their portfolios or alter their lending rates? They could charge higher premiums to distressed firms. Such a move, however, will be largely cosmetic if the book value of these loans is already low. Instead, banks may pursue a quite different strategy to reduce the risk of their portfolio while increasing their margins. It involves increasing their holdings of government paper to reduce their overall credit-risk exposure,²⁹ and reducing loans or increasing the premiums charged to sector 2 firms.

An additional perverse feature of this behavior is readily apparent. Good enterprises, in particular startup firms, may be those most penalized. This crowding-out effect, and the ensuing higher interest rates, hampers capital accumulation in the emerging sector.

Other factors may well undermine the effectiveness of monetary policy. One may think that this undue premium cannot be imposed indefinitely upon good borrowers. Indeed, given that the premium charged to these firms is justified not by the borrowers’ credit risk, but instead by the situation of lenders, other sources of financing may become available. The crowding out could be circumvented, in particular, through the formation of new financial intermediaries, through borrowing abroad, or through long-term capital inflows. If crowding out is circumvented, the problem in the banking sector will rematerialize as a capital-inflow problem. Of course, this could be offset by a steep appreciation of the currency to make zloty-denominated investments riskier, but such a move would again exacerbate the pressure imposed on restructuring firms.

Against this background, a tight monetary policy or a hard-currency option may prove costlier than in more standard conditions. The underlying cause is that they are both rather indiscriminate instruments. As a result, whatever the transmission channel (competitiveness effect or financial crowding out), most of the burden of appreciation is shouldered by the “wrong” sector. Indeed, instead of fostering a more intense restructuring in sector 1 along with a desirable reallocation of resources into sector 2, it simply slows down the emergence of the latter.³⁰

How can these costs of tighter monetary policy be avoided? The message is that tighter monetary policy, be it in the form of exchange rate appreciation or reduced money supply growth, cannot be a substitute for restructuring. Indeed, if it unduly forces the burden of adjustment onto the emerging private sector, tighter monetary policy, unaccompanied by structural reform, might actually delay the restructuring on which future growth will depend. The first-best solution is to extend the discipline of hard budget constraints to all sectors of the economy, to promote restructuring, and to ensure that the burdens of tighter monetary policy are borne proportionately.