Money, barter, and inflation in Russia

Journal of Comparative Economics 32 (2004) 297–314 www.elsevier.com/locate/jce

Money, barter, and inflation in Russia Byung-Yeon Kim a,∗ and Jukka Pirttilä b a Sogang University, CPO Box 1142, Seoul 100-611, South Korea b Bank of Finland Institute for Economies in Transition, PO Box 160, FIN-00101 Helsinki, Finland

Received 13 March 2003; revised 27 February 2004 Available online 21 April 2004

Kim, Byung-Yeon, and Pirttilä, Jukka—Money, barter, and inflation in Russia This paper analyzes relationships among money, barter, and inflation in Russia during the transition period. Following the development of a theoretical framework that introduces barter into a standard macroeconomic model for a small, open economy, we estimate the model using structural cointegration and vector error-correction methods. Our findings suggest that Russian barter resulted partly from output losses, and, to a lesser extent, from the reduction in real money balances. We also find that increases in barter are affected by inefficiencies in the banking sector. Our results imply that increased barter contributed to generating persistent inflation and output decline in Russia. Journal of Comparative Economics 32 (2) (2004) 297–314. Sogang University, CPO Box 1142, Seoul 100611, South Korea; Bank of Finland Institute for Economies in Transition, PO Box 160, FIN-00101 Helsinki, Finland.  2004 Association for Comparative Economic Studies. Published by Elsevier Inc. All rights reserved. JEL classification: C32; E31; E51; P24

1. Introduction The transformation of the former Soviet republics and countries in Eastern Europe can be viewed as a transition to a market economy in which money is the universal medium of exchange. Money plays a passive role in centrally planned economies because the allocation of resources for production is driven by planning decisions. Thus, the transition process replaces a plan-based coordination mechanism with a market-based one that should be based on money and price signals. Contrary to this expectation, the Russian economy experienced severe demonetization during 1992 and 1998. Price liberalization in 1992 * Corresponding author.

E-mail address: [email protected] (B.-Y. Kim). 0147-5967/$ – see front matter  2004 Association for Comparative Economic Studies. Published by Elsevier Inc. All rights reserved. doi:10.1016/j.jce.2004.02.007

298

B.-Y. Kim, J. Pirttilä / Journal of Comparative Economics 32 (2004) 297–314

appears to have had a positive impact on monetization but reliance on barter increased rapidly in 1994. According to the Russian Economic Barometer (2001), over 50% of industrial transactions measured by volume of total enterprise transactions did not involve money in 1998. Rather, enterprises used monetary surrogates such as barter and interenterprise arrears.1 Interestingly, the demonetization trend reversed suddenly after the financial crisis in the autumn of 1998. By the end of 2000, only about 20% of enterprise transactions were still conducted with monetary surrogates. Demonetization in Russia differs from classical demonetization occurring during periods of high inflation or hyper-inflation. Demonetization intensified after 1994 as inflationary pressure was subsiding. Following an extremely unstable period at the outset of market reforms, Russia implemented a successful monetary stabilization program in 1995. Nevertheless, as annual inflation fell from 840% in 1993 to 11% in 1997, the share of barter transactions soared from 9% to 42% (Goskomstat, 2001; Russian Economic Barometer, 2001). These observations raise several provocative issues relevant not only to transition economies, but also to market economies.2 The macroeconomic causes of the emergence and growth of barter transactions may be related to tightened monetary policy.3 In a slightly different context, Ramey (1992) postulates that a real shock may account for increases in the demand for non-monetary forms of exchange because barter could be used by enterprises having difficulties with production or sales. The Russian experience provides a unique opportunity to test empirically for the causes of barter and the consequences of barter on inflation and output. In this paper, we analyze money, barter, and inflation in an integrated structural system. Section 2 reviews the related literature on money, barter, and inflation in Russia and offers a brief survey of the key events in the Russian economy during the 1990s. Section 3 develops a model incorporating barter in a standard macroeconomic model of a small, open economy. Section 4 provides the results of empirical analysis based on a long-run cointegration and a short-run error-correction model. Section 5 presents extensions by considering a long-run model that includes a bank-lending channel and by explaining the reversal of the demonetization process after the 1998 financial crisis. Section 6 concludes with some policy implications.

2. Inflation, money, and barter in Russia The literature contains several explanations of the factors driving Russians to barter. Karpov (1997) and Gaddy and Ickes (1998) emphasize that firms may hide revenues or 1 Russian barter instruments include the exchange of goods, i.e, pure barter, offsets whereby firms write off mutual debt, and veksels, which are bills of exchange issued by corporations, banks, or local and regional governments. Nevertheless, pure barter was the dominant form of non-monetary transactions, as Aukutsionek (1998) reports. 2 Prendergast and Stole (1996) and Marvasti and Smyth (1999) provide evidence that barter transactions have increased rapidly as well in advanced market economies such as the United States. 3 Meltzer (1960), Brechling and Lipsey (1963), Prendergast and Stole (1996) and Petersen and Rajan (1997) construct theories based on trade credit to support this possibility.

B.-Y. Kim, J. Pirttilä / Journal of Comparative Economics 32 (2004) 297–314

299

exaggerate costs using non-transparent barter transactions to avoid taxation. Hendley et al. (1997) suggest firms may resort to barter transactions due to the poor payments system provided by Russian banks. Pinto et al. (2000) stress that implicit subsidies from the government to the private sector through tax arrears, tax offsets, and energy supplied below market prices were key reasons for the rise of barter and inter-enterprise arrears. Finally, Guriev and Kvassov (2004) argue that firms may use barter for price discrimination; their model predicts a positive relationship between barter and the concentration of market power, for which they find empirical evidence. Although, many of these factors are likely to be operating, the main debate involves two competing schools of thought. Gaddy and Ickes (1998) and Guriev and Ickes (2000) argue that barter is a result of lagged restructuring. Alternatively, Marin et al. (2000), Commander and Mumssen (2000), and Marin (2002) assert that a lack of liquidity, caused either by insufficient money supply or the fragility of the banking system, was the prime reason for reliance on barter. Guriev et al. (2002) contends that the two hypotheses are interconnected because the lack of liquidity results in insufficient credit, which is linked to a lack of investor control over managers. Much of the empirical evidence is based on cross-section or panel microeconometric analyses of firm-level data and appears to support the liquidity hypothesis. Carlin et al. (2000) and Commander et al. (2002) find that firms having limited access to financial resources and suffering from liquidity problems are more likely to rely on barter. While other reasons, such as a failure of the banking system, may explain the liquidity squeeze, macroeconomic policies, in particular, tight monetary policy may be responsible for the lack of liquidity in the economy. Because variables using firm-level data are based on reports of managers who may wish to hide the true information on the conditions of their firms, an analysis of barter transactions in Russia based on aggregate data may give more reliable results. Brana and Maurel (1999) investigate the determinants of barter transactions from 1992 to 1998. Based on a cointegration relationship among barter, interest rates, and the financial situation of firms, these authors find that barter is positively correlated with the interest rate and negatively correlated with the index of financial position of firms. The implication is that barter is caused by financial distress. They also use a micro data set to show that barter provides a way for indebted firms to avoid costly restructuring. The literature investigating the causes of inflation in Russia does not include factors other than money. Buch (1998) considers a single equation error-correction model for Russian money demand based on the quantity theory of money. Korhonen (1998) applies cointegration analysis and vector-error correction models (VECM) to estimate the money demand relationship.4 Our paper differs from this literature in several important ways. First, we use an explicit macroeconomic IS-LM framework that leads us to consider a larger set of variables in our structural cointegration analysis. Second, we test the two contesting hypotheses regarding the causes of barter in a single macroeconomic framework to investigate whether the financial situation through the real money supply and interest rates or real-economic difficulties captured indirectly by output drive barter transactions. In con4 Studies examining other causes of inflation in a vector autoregression (VAR) framework are undertaken for certain transition countries. These include Brada and Kutan (2002) and Dibooglu and Kutan (2001) for Hungary and Poland, Blangiewicz and Charemza (1999) and Kim (2001) for Poland, Kalra (1999) for Albania. Cottarelli and Doyle (1999) provide a survey on cross-country and panel data analysis of inflation in transition economies.

300

B.-Y. Kim, J. Pirttilä / Journal of Comparative Economics 32 (2004) 297–314

trast to Brana and Maurel (1999), we assess the relative importance of these two competing explanations of barter. Third, our structural system enables us to investigate whether barter affects the macroeconomic variables, in particular, inflation and the money supply. At the outset of Russia’s transition in 1992, price liberalization and the monetary overhang inherited from the Soviet era combined to generate high inflation, as Fig. 1 shows. However, the inflation rate did not exceed the 50% per month threshold suggested by Cagan (1956) to define hyperinflation. One reason for the persistence of high inflation from 1992 to 1994 was the existence of a ruble zone, allowing several former Soviet republics to issue rubles independently until the latter part of 1993. None of the several stabilization programs during this time were successful. In the summer of 1995, the Russian government adopted an exchange rate-based stabilization program in which the nominal value of the ruble was anchored by a crawling peg. Annual inflation fell from 129% in 1995 to 11% in 1997; however, fiscal consolidation failed and government debt grew rapidly. When the Asian crisis hit the global economy in late 1997, Russia’s situation deteriorated rapidly. In mid-August 1998, Russia was forced to devalue the ruble and the government stopped servicing its debt. These events led to a full-scale financial crisis and paralyzed Russia’s already poorly functioning banking sector. Following the crisis, authorities managed to establish control over inflation quickly. The economy began to recover in 1999 due to the devaluation of the ruble and the high export prices of oil. Annual inflation, which exceeded 80% in 1998, fell to 37% in 1999. Barter was relatively rare at the start of the transition period, as Fig. 2 indicates. However, beginning in mid-1993, barter transactions as a percentage of total industrial transactions began a steady increase. The biggest increase corresponds to a period when inflation was declining from 1995 to 1997. The 1998 devaluation led to improvements in competitiveness and the liquidity position for Russian firms with a concurrent reduction

Fig. 1. Inflation dynamics in Russia 1992–2000, monthly inflation. Source: The Central Bank of Russia (various editions).

B.-Y. Kim, J. Pirttilä / Journal of Comparative Economics 32 (2004) 297–314

301

Fig. 2. Barter as a share of total industrial transactions (%). Source: Russian Economic Barometer (various editions).

in the use of barter. As inflation fell again in 1999, barter transactions also declined. Hence, no clear relationship between changes in inflation and barter transactions in Russia is discernible from these data.

3. A theoretical framework To provide the theoretical formulation for our empirical work, we present a simple model of money demand, barter, and the price level. The starting point is an IS–LM model of a small, open economy consisting of five variables, namely, the price level, broad money, the exchange rate, the interest rate, and output. To simplify this framework before introducing barter, we do not model output, the exchange rate, and the interest rate explicitly. For output, we assume that the transitional recession and structural change are more important determinants than short-run demand management policies (Gaddy and Ickes, 1998). To capture interest rate or exchange rate movements with an interest rate parity relationship is difficult (Korhonen, 1998). Hence, we begin with a model consisting of a money demand equation and a price equation based on Purchasing Power Parity (PPP).5 We introduce barter by adding an equation so that we have the following three-equation system: mt − pt = α + β(yt − pt ) + γ (bt − pt ) + δrt + λet ,

(1)

5 Due to the paucity of data, we consider only the overall price level rather than introducing separate prices for tradable and non-tradable sectors.

302

B.-Y. Kim, J. Pirttilä / Journal of Comparative Economics 32 (2004) 297–314

bt − pt = ρ + ζ (yt − pt ) + µrt + ψ(mt − pt ), pt = π

+ σ et + θpt∗

+ η(bt − pt ).

and

(2) (3)

In these equations, pt denotes the price level, mt is the nominal demand for broad money, e is the exchange rate defined as the ratio of the ruble to the US dollar, pt∗ is the foreign price index, yt is nominal output, rt is the nominal interest rate, and bt is the nominal value of barter. All variables are specified in natural logs. The real money demand equation (1) modifies the standard specification in an open economy by including barter.6 Commander et al. (2002) assert that barter may be used as a substitute for money in transactions. Hence, an increase in barter payments reduces the transaction demand for real money balances. In addition, barter functions as trade credit; as Meltzer (1960) and Brechling and Lipsey (1963) argue, the impacts of trade credit on monetary transmission mechanisms and inflation should be taken into account. If monetary policy is tightened, firms may substitute trade credit for bank lending. Therefore, the use of trade credit may offset some of the impacts of traditional monetary transmission through the bank-lending channel. Petersen and Rajan (1997) and Kohler et al. (2000) find empirical support for this argument by showing that firms facing liquidity constraints are more prone to rely on trade credit, especially during periods of tight monetary policy. Hence, we include barter in the real demand for money equation and expect γ < 0. The expected signs for the other parameters in Eq. (1) are based on the standard macroeconomic theories so that we expect β > 0, δ < 0 and λ < 0. The real value of barter equation reflects potential substitutability between money and barter; real money is included in this equation to allow for money and barter being mutual substitutes. However, money and barter may be imperfect substitutes because barter transactions involve an implicit interest rate.7 If the interest rates on bank loans increase, an arbitrage relation between different forms of assets implies that the interest rate on trade credit and barter will also increase. If the demand for credit decreases as interest rates increase, barter will be a decreasing function of interest rates. In addition, Gaddy and Ickes (1998) and Guriev and Ickes (2000) argue that barter may result from lagged restructuring of firms in conditions of declining demand and a reduction in output. If firms use barter as a survival strategy, a decline in output would increase the use of barter. Thus, the expected signs of the parameters in Eq. (2) are ζ < 0, ψ < 0, and µ < 0. The price equation is derived from the PPP relation with barter added. Hence, the price level is an increasing function of both the exchange rate due to depreciation and the foreign price level. Because of the relatively minor impact of foreign price inflation compared with domestic inflation in Russia, we exclude the foreign price in our empirical exercise to preserve degrees of freedom. Barter may affect prices in several ways. The increased liquidity resulting from a rise in barter credit might lead to a higher price level. However, increased barter also indicates demonetization of the economy. Since monetization usually 6 As Walsh (1998) suggests, the standard LM or money demand relationship can be derived from microeconomic foundations to justify the transaction-demand for money. 7 In trade credit transaction, the customer typically gets a discount if he pays by cash immediately but he must pay the full amount if he uses the extended maturity. In Russia, cash prices appear to be lower than barter prices, which confirms that barter transactions incur interest (Commander and Mumssen, 2000).

B.-Y. Kim, J. Pirttilä / Journal of Comparative Economics 32 (2004) 297–314

303

leads to inflation (Li, 1997), demonetization may decrease the price level. Hence, the expected signs of the coefficients in Eq. (3) are σ > 0, θ = 0, and η is ambiguous. Based on the discussion in the previous section, barter may help firms to survive in the short run so that barter would have a positive impact on production. However, barter trade involves significant transaction costs and it may also sustain unproductive structures and inhibit restructuring. These features lead to barter having a negative impact on output in the long run. Guriev et al. (2002) suggest that the existence of barter chains among some firms have negative external effects on profitable firms. Carlin et al. (2000) find empirical evidence that barter has had a clearly negative impact on productivity in Russia. Hence, although we do not model directly a long-run equation for output, we hypothesize that barter boosts output in the short run but that this impact eventually reverses itself.

4. The empirical analysis Based on our theoretical framework, we consider six variables in logs, namely, prices (p), money (m − p), interest rates (r), exchange rate (e), output (y − p), and barter (b − p) in our empirical work.8 Table 1 provides definitions of the variables and the statistical sources of the data. The data are monthly and cover the period from January 1994 to December 2000. For a long-term analysis, the time span is unfortunately short. Although this is common for any time-series analysis of transition countries, we must keep the limitations from the short time span in mind when assessing our results. For the output data, we use industrial output instead of GDP because official monthly GDP data are unavailable. We conduct both augmented Dickey–Fuller (ADF) and Phillips–Perron tests for unit roots for the sample from 1994 to 2000. We consider specifications with a constant and a trend and also only a constant. The tests confirm that all the variables are considered to be integrated of order one, i.e., I (1).9 The results from the cointegration analysis with unrestricted constant term and no trend are provided in Table 2.10 Both the trace and the max test statistics indicate three cointegration vectors (CV).11 Following normalization and 8 The extent of reliance on barter in Russia may be measured in several ways. The simplest approach is to use the index based on the enterprise survey conducted by the Russian Economic Barometer. This variable (sb) refers to the log of the share of barter in total enterprise transactions. To pursue the notion that barter may be a potential source of credit, we use the log of the volume index of barter, (b − p), obtained by multiplying the index of industrial output by the share of barter trade. We use the share of barter to check for robustness in the next section. 9 A table of the test results is available from the authors upon request. 10 Estimations were conducted using PcGive and Microfit statistical software. We use an F -test to identify the correct lag length. We include the dummies for October 1994, June 1995, and September 1998. The first and last dummies are intended to capture the ruble crisis periods. The dummy for June 1995 attempts to capture political instability due to the war in Chechnya and related events in the Budyonnovsky region in addition to preparation for a macroeconomic regime change when Russia adopted an IMF stabilization program with a fixed exchange rate system in July. 11 Based on the small sample-corrected critical values, the number of cointegration vectors could be two or three at 95% critical values or three at 90% critical values. We consider three cointegration vectors, partly because this is consistent with our theoretical framework and partly because it can be supported by the results of cointegration tests.

304

B.-Y. Kim, J. Pirttilä / Journal of Comparative Economics 32 (2004) 297–314

Table 1 List of variables Variables

Description

Sources

p m−p e

Log of consumer price index Log of real domestic M2 Log of Ruble/USD exchange rate, period avg.

r

Log of commercial banks’ lending rate

y−p sb b−p

Log of real industrial production Share of barter in sales of industry Log of the real volume of barter (b = sb ∗ y)

Central Bank of Russia (various editions) Russian Economic Trends (various editions) Central Bank of Russia (various editions), Russian Economic Trends (various editions) 1992–1994: World Bank (various editions), 1995–2000: IMF (various editions) Goskomstat (various editions) Russian Economic Barometer (various editions) Own calculations

Table 2 Cointegration tests H0 : rank = p

max test

using T -nm

95%

p=0 p
1 p
2 p
3 p
4 p
5

80.04** 54.04** 41.53** 14.5 7.877 0.158

57.17** 38.6* 29.66* 10.36 5.627 0.1129

39.4 33.5 27.1 21.0 14.1 3.8

trace test 198.1** 118.1** 64.06** 22.53 8.035 0.158

using T -nm

95%

141.5** 84.36** 45.76 16.1 5.739 0.1129

94.2 68.5 47.2 29.7 15.4 3.8

Notes. The max test is the maximal eigenvalue test for rank and trace is the trace test for rank. T -nm denotes the small sample-adjusted critical values for rank. * Significant at the 5% level. ** Idem., 1%.

the exact identification suggested by our theoretical framework, we produce the estimation results presented in Table 3.12 Overall, the estimation results contain reasonable parameter signs and estimates. Table 4 presents the results from likelihood ratio-based tests for linear restrictions. The restriction test results indicate that money demand was unaffected by barter because (b − p) = 0 in the first CV. Moreover, the hypothesis that barter is not caused by real money is rejected because (m − p) = 0 in the second CV. Finally, the joint restriction that money and barter are independent of each other, i.e., (b − p) = 0 in the first CV and (m − p) = 0 in the second CV, is also rejected. This evidence indicates that money and barter are substitutes, which supports our view of barter as a form of trade credit. However, the link is one-directional in that barter does not affect real money in the long run, which may reflect the efficiency gain from using money rather than barter transactions. Perhaps our interesting finding is that liquidity problems, namely the reduction in available real money balances, are a cause of the emergence and growth of barter. An alternative hypothesis is that barter serves as a survival strategy for firms facing 12 We impose a zero restriction on the coefficient of the interest rate in the first cointegration vector and retain

the exchange rate because it provides a proxy for expected inflation. This specification seems to work better in the Russian context than does money demand with an interest rate.

B.-Y. Kim, J. Pirttilä / Journal of Comparative Economics 32 (2004) 297–314

305

Table 3 Estimation results from restricted cointegration analysis Parameters of cointegration vectors (CV) CV1 m − p = −0.1199(b − p) + 2.1205(y − p) − 0.4396e (0.9837) (0.4148) (0.0329) CV2 b − p = −1.1618(m − p) − 1.9719(y − p) − 0.3410r (0.1433) (0.4389) (0.0316) CV3 p = −0.3523(m − p) − 0.6179(b − p) + 0.6615e (0.3030) (0.1712) (0.0691) Diagnostic test results Alternative

Test

Value (probability)

Serial correlation Normality Heteroscedasticity

F (180, 132) χ 2 (12) χ 2 (1008)

1.2261 (0.1075) 17.219 (0.1415) 988.16 (0.6663)

Note. In the upper part of the table, absolute standard errors are in parentheses. Table 4 Hypothesis test results based on LR tests (rank = 3) Hypothesis (b − p) = 0 in CV1 (m − p) = 0 in CV2 (b − p) = 0 in CV1 and (m − p) = 0 in CV2 (y − p) = 0 in CV2 (b − p) = 0 in CV1 and (y − p) = 0 in CV2 r = 0 in CV2 (y − p) = 0 in CV1 (y − p) = 0 and (b − p) = 0 in CV1 e = 0 in CV1 (m − p) = 0 and CV3 (b − p) = 0 and CV3 (b − p) = 0 and (m − p) = 0 in CV3 e = 0 in CV3 (b − p) = 0 in CV1 and (m − p) = 0 in CV3

Test statistic χ 2 (1) = 0.4499 [0.5024] χ 2 (1) = 26.469 [0.0000]** χ 2 (2) = 27.008 [0.0000]** χ 2 (1) = 9.0596 [0.0026]** χ 2 (2) = 9.0608 [0.0108]* χ 2 (1) = 25.099 [0.0000]** χ 2 (1) = 13.539 [0.0011]** χ 2 (2) = 14.081 [0.0028]** χ 2 (1) = 21.716 [0.0000]** χ 2 (1) = 0.7070 [0.4004] χ 2 (1) = 6.4622 [0.0110]* χ 2 (2) = 8.0424 [0.0179]* χ 2 (1) = 6.1158 [0.0134]* χ 2 (2) = 1.0552 [0.5900]

Notes. CV1, CV2, and CV3 refer to the first, second, and third cointegration vector appearing in Table 3; p-values in brackets. * Significant at the 5% level. ** Idem., 1%.

declining production. The restriction that output does not affect barter, i.e., (y − p) = 0 in the second CV is rejected, implying that barter is a result of the production decline. Therefore, our analysis supports indirectly the view that barter is used to avoid restructuring, even though we have no direct measure of restructuring in our analysis. As in Table 3, the long-run relationship, i.e., (b − p) = −1.162(m − p) − 1.972(y − p) − 0.341r, implies that the impact of output on barter is about twice that of money on barter. The rejection of the zero restriction on interest rates in the second cointegration vector, i.e., r = 0 in CV2, implies that demand for barter as trade credit is a decreasing function of

306

B.-Y. Kim, J. Pirttilä / Journal of Comparative Economics 32 (2004) 297–314

its interest rate. This interest rate is positively related to the interest rate on bank lending due to the arbitrage conditions of lenders and borrowers. The PPP hypothesis is supported by the fact that the restriction on e = 0 in the third CV is rejected, which implies that the exchange rate is a key determinant of Russian inflation. The elimination of barter in the third CV is rejected, which indicates that the price level is negatively associated with real barter and establishes a negative effect of demonetization on prices. There are two channels through which barter might affect prices. First, barter may affect cash prices of goods directly due to enterprises’ dumping goods in cash as Oppenheimer and Granville (2001) argue. Second, barter may affect prices indirectly and negatively due to increased output, which would not be forthcoming without using barter, and the subsequent decline in the cash prices of these goods because of increased supply. Note that barter prices would not be lower than cash prices; substantial evidence indicates that the opposite is true in Russia (Commander and Mumssen, 2000). Our results do not refer to barter prices themselves, but rather to the impact of barter on cash prices, which are the basis of price statistics. The final estimates of the model, following the restriction tests as above, are presented in Table 5. From CV3 in Table 5, a one percentage-increase in real barter leads to a 0.45% decrease in prices. To examine further the effects of barter on the other variables, we use Generalized Impulse Response (GIR) functions. These describe the time profile of the effect of a unit shock to a particular equation on the relevant endogenous variables, taking into account the contemporaneous interactions of all the endogenous variables of the system. Unlike other impulse responses, the GIR is invariant to the ordering of the variables in the VAR (Pesaran and Shin, 1998). We assume that the shock is small enough not to change the parameters of the underlying VAR model. Figure 3 shows impulse responses of prices, money, and output to a one standard error unexpected increase in barter. Prices increase for a short period following the shock, but prices decrease in response to a rise in barter transactions over the long run. The effect of a shock to barter on money turns out to be always negative. The short run impact of barter on real output is positive, which confirms our hypothesis that firms engage in barter transactions to escape decreases in economic activity. However, the long-run impact of barter on output is negative, which

Fig. 3. Generalized impulse responses of money, prices and output to a one standard error increases in real barter.

B.-Y. Kim, J. Pirttilä / Journal of Comparative Economics 32 (2004) 297–314

307

Table 5 Final estimates of long-run model Parameters of cointegration vectors CV1 m − p = 2.4595(y − p) − 0.4856e (0.4372) (0.0349) CV2 b − p = −1.1259(m − p) − 1.8622(y − p) − 0.3188r (0.1326) (0.4215) (0.0247) CV3 p = −0.4511(b − p) + 0.73825e (0.1452) (0.0517) Speed of adjustment m−p

b−p

p

y−p

e

r

−0.1008 −0.2140 0.0034

−0.4998 −0.3707 0.1868

0.0148 0.0524 −0.0391

−0.0810 −0.2082 0.0138

0.0498 0.0808 −0.0506

−0.0816 0.0217 0.0477

Note. In the upper part of the table, absolute standard errors are in parentheses.

Fig. 4. Generalized impulse responses of barter, prices and output to a one standard error increases in real money.

suggests that inefficiencies arising from the use of barter and delays in restructuring dominate any positive impact over time. In other words, enterprises may avoid output losses or bankruptcy by relying on barter in the short run but the cost of barter transactions is high enough in the long run to cause output loss, possibly because of delays in enterprise restructuring. In Fig. 3, the short-run positive impact on output lasts for only about one year. Figure 4 displays impulse responses of prices, barter and output to a one standard error unexpected increase in money. An increase in the money supply is associated with decreases in barter transactions in the short and long runs. However, the impact of money on both prices and real output is almost the opposite of this effect; an unexpected increase in the money supply increases prices and output after only a brief adjustment period. This

308

B.-Y. Kim, J. Pirttilä / Journal of Comparative Economics 32 (2004) 297–314

result suggests that a looser monetary policy in Russia would have eventually contributed to long-run real economic growth mainly by reducing barter transactions. The above analysis suggests that barter contributed to persistent inflation and real output loss in Russia. A significant decrease in annual inflation from 220% in 1994 to 11% in 1997 can be explained partly by both the effect of a reduction in liquidity on prices and the impact of an increase in barter transactions on prices. However, inflation remained persistent from 1999 to 2000 in spite of favorable economic conditions after the financial crisis in autumn of 1998; annual inflation in 1999 and 2000 was 37 and 20%, respectively (Goskomstat, 2001). Decreases in barter transactions since autumn of 1998 can account for this persistent inflation from 1999 to 2000. When the economy was in recession, enterprises used barter to boost real output. However, the positive impact of barter on output was short-lived, only about one year, and the subsequent economic recovery was hampered by the substantial accumulation of barter transactions. Our empirical work is consistent with the hypothesis in Guriev et al. (2002) that barter transactions undermine efficiency by postponing enterprise restructuring and maintaining soft budget constraints. Hence, barter does not have a neutral impact on the economy in the long run. To consider short-run effects in more detail, we use a vector error-correction model (VECM). Table 6 presents the estimates of the parsimonious VECM.13 We do not model interest rates because the variable is exogenous to other variables.14 The system-based test results, which are in the last three rows of Table 6, indicate that the model does not suffer from diagnostic problems, although the single equation tests suggest a slight problem with the price equation.15 Chow tests do not detect any structural breaks in the system; however, small breaks are found in the price and real money equations during the latter half of 1998 that may result from the turbulence of the financial crises. All error-correction terms have the correct signs in their own equations, implying that our estimation results are fairly robust. Barter and money are complements in the short run, although they become substitutes in the long run. The significant error-correction term in the equation for change in output indicates that, when barter is above its long-run level, output falls. Thus, barter may lead to efficiency losses in the long run, even though an increase in barter raises real output in the short run.

5. Some extensions The money supply, measured by M2, may be a poor indicator of the financial conditions of enterprises. Marin et al. (2000), Schoors (2001) and Gara (2001) suggest that the reluctance of financial institutions to lend to enterprises, rather than an insufficient money supply, is a direct cause of barter. We test this hypothesis and check the robustness of our 13 We include the same dummies, i.e., October 1994, June 1995 and September 1998, as well as two additional dummies, November 1994 and August 1998. The new additional dummies are intended to capture periods of ruble crisis. 14 Weak exogeneity tests confirm this point so that interest rates do not adjust to deviations from the long-run equilibrium. The test results are available from the authors. 15 Stability analysis is not reported for brevity; the results are available from the authors upon request.

Table 6 VECM estimation results and diagnostics t-value

D(b − p)

t-value

Dp

t-value

D(y − p)

t-value

De

t-value

−0.134 0.020 0.069 0.098 0.147 −0.575 −1.021 −0.261 0.083 0.403 −0.007 −0.069 −0.152 −0.018 −0.035 −0.122 −0.078 0.090 −0.104 −0.280

−1.387 0.220 1.357 2.198 3.170 −2.193 −4.438 −3.318 0.806 4.188 −0.144 −2.757 −5.550 −0.987 −1.435 −4.029 −2.444 2.853 −3.587 −9.901

−0.568 1.395 −0.592 −0.530 −0.170 −0.707 −0.958 −0.217 0.362 0.423 0.684 −0.403 −0.203 0.165 −0.044 −0.078 −0.206 0.113 0.025 −0.055

−2.202 5.880 −4.346 −4.451 −1.374 −1.010 −1.561 −1.033 1.319 1.646 5.139 −6.027 −2.771 3.312 −0.674 −0.970 −2.428 1.341 0.324 −0.732

0.073 −0.098 −0.008 −0.016 −0.041 0.649 0.093 0.112 −0.193 −0.001 0.065 0.002 0.022 −0.023 0.025 0.084 0.058 −0.025 0.043 0.323

2.171 −3.155 −0.437 −1.043 −2.540 7.089 1.156 4.081 −5.368 −0.020 3.760 0.191 2.308 −3.492 2.884 7.972 5.240 −2.263 4.259 32.769

−0.738 0.743 0.037 −0.097 0.172 −0.565 −0.379 −0.253 0.091 0.161 0.391 −0.065 −0.150 0.028 −0.061 0.057 −0.054 0.054 −0.030 −0.074

−5.364 5.868 0.503 −1.535 2.596 −1.513 −1.155 −2.256 0.624 1.173 5.512 −1.829 −3.827 1.070 −1.757 1.328 −1.201 1.192 −0.714 −1.851

−0.204 −0.075 0.004 −0.034 −0.043 −0.408 0.477 0.147 0.243 −0.212 0.222 0.022 0.055 −0.050 0.031 0.182 −0.059 −0.114 0.079 0.742

−2.959 −1.183 0.122 −1.065 −1.313 −2.181 2.909 2.623 3.311 −3.087 6.253 1.254 2.812 −3.792 1.756 8.427 −2.613 −5.055 3.788 36.827

Constant Serial correlation, F (5, 58) Normality, χ 2 (2) Functional form, F (5, 53) Heterosce, F (35, 27) Vector autocorrelation, F (125, 172) Vector normality, χ 2 (10) Vector heterosce, 2 F (525, 242)

3.339 6.261 1.040 (0.403) 1.513 (0.469) 0.600 (0.700) 0.342 (0.998) 1.089 (0.302) 10.274 (0.417) 0.560 (1.000)

2.986 2.097 0.646 (0.666) 3.941 (0.139) 0.336 (0.889) 0.438 (0.989)

−0.369 −1.982 1.242 (0.301) 1.916 (0.384) 0.385 (0.857) 2.254 (0.016)

3.036 3.996 2.634 (0.033) 4.549 (0.103) 1.091 (0.376) 0.512 (0969)

−0.921 −2.420 1.441 (0.223) 1.904 (0.386) 0.545 (0.741) 1.179 (0.333)

309

Notes. 1. P -values for the diagnostic tests are provided in parentheses. 2. Five dummies are used in these estimations. Dummies for Oct. and Nov. 1994 and for Aug. and Sept. 1998 are intended to capture periods of ruble crisis. The dummy of June 1995 is intended to capture political instability from the war in Chechnya and related events in the Budyonnovsky district in addition to the preparation for a macroeconomic regime change. 3. On the basis of long-run cointegration results, the error correction terms are calculated as: ecm1 = (m − p) − 2.4595(y − p) + 0.4856e, ecm2 = (b − p) + 1.1259(m − p) + 1.8622(y − p) + 0.3188r, and ecm3 = p + 0.4511(b − p) − 0.7382e.

B.-Y. Kim, J. Pirttilä / Journal of Comparative Economics 32 (2004) 297–314

D(m − p) D(m − p)−1 D(m − p)−3 D(b − p)−1 D(b − p)−2 D(b − p)−3 Dp−1 Dp−2 D(y − p)−3 De−1 De−2 De−3 ecm1 ecm2 ecm3 Dr oct94 nov94 june95 aug98 sep98

310

B.-Y. Kim, J. Pirttilä / Journal of Comparative Economics 32 (2004) 297–314

baseline cointegration results using data on bank claims on the private sector.16 Lending by financial institutions to the private sector can be affected not only by the efficiency of the banking sector but also by monetary policy. To find an appropriate proxy for the efficiency of the banking sector, we regress bank claims on the private sector on the money supply and interest rates. The residuals of this auxiliary regression, denoted bsd, which are no longer affected by monetary policy or interest rates, indicate fragility of the banking system. These residuals are considered to be a proxy for the extent of development of the banking sector.17 In addition, this procedure ensures the orthogonality of the variable to other variables, e.g., the money supply. Table 7 presents the cointegration results.18 No significant changes from our previous results are found, even though interest rates are dropped from the system.19 As in the second CV of our baseline results, the share of barter is negatively correlated with real output and real money; however, the extent of development of the banking sector also affects barter.20 The homogeneity of the coefficient on bsd in the second CV of Table 7 indicates that a 1% increase in bank lending to the private sector that is not associated with monetary or interest rate policies causes a 1% decrease in the share of barter. In addition, as the second CV of Table 5 shows, the sum of the coefficients on the variables related to a financial shock on enterprises, namely an insufficient money supply and bank failures, is about the same as the size of the coefficient on a real output shock. The empirical work reported in Tables 5 and 7 can shed light on the reversal of the demonetization process after autumn 1998. Barter transactions are driven by an output decline, liquidity problems, and bank failures. The 1998 financial crisis affected all three Table 7 Parameters of cointegration vectors using a proxy for the development of the banking sector CV1 m − p = 0.0449sb + 2.4500(y − p) − 0.4701e CV2 sb = −1.0929bsd − 1.0223(m − p) − 2.2145(y − p) CV3 p = −0.0410sb + 0.4061(m − p) + 0.7032e Note. The data are from October 1995 to December 2000. 16 Using these data require us to consider a shorter sample period because they are available from June 1995

only. We reduce the number of variables to six by dropping interest rates from our estimation equation, assuming that the interest rate affects the economy mainly through bank lending to enterprises. Moreover, Domac and Elbirt (1998) and Kim (2001) argue that the effect of interest rates may not be significant in transition economies. 17 Similar residual approaches are used by Hall (1997), Schwert (1989), Harrison (2001), and Roche (2001) in different contexts. 18 We use the share of the volume of barter transactions in total enterprise transactions (sb) to check the reliability of our previous estimation that contained the real volume of barter (b − p). 19 There is a change in the sign of the coefficient on barter. However, the zero restriction on the coefficient is accepted. The statistic of the LR test, χ 2 (1), is 0.0531 (p-value: 0.8177). In the third cointegration vector, the real money supply now increases prices, although the share of barter is associated negatively with prices. Zero restrictions on the coefficients on both variables are accepted, suggesting that prices are affected mainly by the exchange rate and not the other variables. The statistic of the LR test, χ 2 (2), is 2.6726 (p-value: 0.2628). Separate tests on money and barter also suggest that a zero restriction is accepted. 20 The zero restriction on the coefficient is rejected. The statistic of the LR test, χ 2 (1), is 13.699 (p-value: 0.0033).

B.-Y. Kim, J. Pirttilä / Journal of Comparative Economics 32 (2004) 297–314

311

of these either directly or indirectly. First, the ruble’s sharp devaluation led to decreases in the real money supply. Given the estimated negative correlation between money and barter, the decrease in money should induce a rise in barter transactions. Second, the money supply increased due both to increases in output, possibly from import substitution, and to the positive effects of devaluation on exports. The large coefficient on real output in the money equation, which is four times as high as that on exchange rates, implies that the output effect dominates the exchange rate effect. In fact, the exchange rate increased by 2.6% per month from September 1998 to December 2000, while industrial output rose by 1.3% per month during the same period (Goskomstat, 2001). Hence, the negative effect of the exchange rates on the money supply, and consequently on barter, was dominated by the positive output effect. Third, increases in real output decreased barter directly; at the same time, increases in the money supply contributed to a reduction in barter transactions. Therefore, we conclude that improvements in the real sector after the financial crisis are a key element of the re-monetization process after autumn 1998.21 Our findings in Table 7 suggest that strengthening the banking sector after the crisis reinforced the re-monetization process. Before the crisis, Russian banks’ lending to the public sector from purchasing government bonds was high. Buch (1998), Gara (2001) and Schoors (2001) argue that lending to the private sector was crowded out by public borrowing. With the weakening of the government bond market during the financial crisis, lending from financial institutions to the private sector increased due to the focus on developing profitable commercial credit portfolios (OECD, 2002).

6. Conclusion Using data on Russia from 1994 to 2000, we analyze relationships among barter, money, and prices focusing on the causes and effects of barter relative to the other variables. Based on long-run cointegration results, we find that the emergence and growth of barter in the Russian economy are associated with both the motivation of firms to survive in the presence of sales or production difficulties and the lack of liquidity experienced by firms. Hence, the two competing explanations for barter, namely restructuring and liquidity, are supported by our empirical analysis of aggregate data. Nevertheless, the effect of the former is twice as large as that of the latter. Therefore, the dominant motive for barter trade among firms may be the avoidance of bankruptcy in the face of low production and perhaps low profitability. In addition, we introduce a proxy for bank lending and find that both of the financial variables, i.e., the money supply and bank lending, are significant. In fact, the sum of these two effects is approximately equal to the production effect on barter. Our empirical findings explain the re-monetization process after the 1998 financial crisis in Russia. The key factor leading to the reduction in barter was the improvement in the real sector, which was triggered by the substantial devaluation of the ruble. Our results suggest that barter contributes to persistent inflation. We find evidence that the effect of barter on prices is lower than that of money, which suggests a negative 21 Changes in government policy, including the refusal of receiving payment in kind, also contributed to a reduction in barter transactions.

312

B.-Y. Kim, J. Pirttilä / Journal of Comparative Economics 32 (2004) 297–314

relationship between barter transactions and prices. Decreases in barter transactions from 1999 to 2000 may be one reason for Russia’s persistent inflation ranging between 20 and 37% per annum in the same period. Our results also shed light on the persistence of real output decline in Russia. We find that firms experiencing difficulty in selling for money use barter transactions, which results in a larger total supply of goods in the market and consequently a reduction in cash prices. Following a positive impact for about one year, the effect of barter on real output becomes negative thereafter. Hence, Russia’s prolonged recession from 1992 to 1998 can be explained partly by the substantial barter transactions occurring in earlier periods. Our empirical work leads us to draw several policy implications. A liquidity crisis caused by insufficient money supply or poor functioning of financial intermediaries may affect the supply side of the economy through a persistent, negative effect of barter on economic growth. A delay in enterprise restructuring may also provide an environment conducive to barter transactions. Hence, our results provide empirical confirmation of the potential benefits associated with restructuring, reform in the banking sector, and sound monetary policy that supplies the appropriate amount of money without endangering the stability of the economy. To make the transition to a well-functioning market economy, these policies should be implemented together with the creation of appropriate institutions to prohibit the use of barter. From a theoretical perspective, our findings suggest that a macroeconomic empirical exercise that excludes barter as a regressor may fail to capture the relevant information for making inferences about money and inflation in an economy in which barter transactions are present. Macroeconomic analysis has the obvious limitation that firm behavior cannot be analyzed directly. Nevertheless, our results are consistent with earlier microanalysis and they shed new light on the effects of barter transactions on macroeconomic variables. However, we have not considered the issue of whether barter is a long-run phenomenon in Russia. Our results apply to a specific time period; hence, it would be worthwhile to investigate whether barter is sustainable in the long run. Such an empirical exercise requires a longer time series.

Acknowledgments We thank Mark De Broeck, Pertti Haaparanta, Mark Harrison, Iikka Korhonen, Antti Ripatti, Jouko Vilmunen, Matti Virén and two anonymous referees for their valuable comments and suggestions. The paper is based on an earlier working paper written with Jouko Rautava. We are very grateful to him for allowing us to build on that work. We also thank participants in the 2001 workshop on transition economics at the Bank of Finland’s Institute for Economies in Transition (BOFIT) and participants in the seminar at Warwick University for their useful comments. Part of this research was undertaken while ByungYeon Kim was a visiting researcher at BOFIT in 2001; he acknowledges gratefully the excellent research environment.

B.-Y. Kim, J. Pirttilä / Journal of Comparative Economics 32 (2004) 297–314

313

References Aukutsionek, Sergei, 1998. Industrial barter in Russia. Communist Economies and Economic Transformation 10 (2), 179–188. Blangiewicz, Maria, Charemza, Wojciech W., 1999. East European economic reform: Some simulations on a structural vector autoregressive model. Journal of Policy Modeling 21 (5), 535–557. Brada, Joseph, Kutan, Ali M., 2002. The end of moderate inflation in three transition economies? William Davidson Institute, University of Michigan. Working paper No. 433. Brana, Sophie, Maurel, Mathilde, 1999. Barter in Russia: liquidity shortage versus lack of restructuring. William Davidson Institute, University of Michigan. Working paper No. 271. June. Brechling, Frank P.R., Lipsey, Richard G., 1963. Trade credit and monetary policy. Economic Journal 73, 618– 641. Buch, Claudia M., 1998. Russian monetary policy—assessing the track record. Economic Systems 22 (2), 105– 146. Cagan, Philip, 1956. The monetary dynamics of hyperinflation. In: Friedman, Milton (Ed.), Studies in the Quantity Theory of Money. Univ. of Chicago Press, Chicago. Carlin, Wendy, Fries, Steven, Schaffer, Mark E., Seabright, P., 2000. Barter and non-monetary transactions in transition economies: evidence from a cross-country survey. CERT discussion paper No. 00/04. Heriott–Watt University. March. Central Bank of Russia, various editions. Main macroeconomic indicators. CBR, Moscow. Commander, Simon, Mumssen, Christian, 2000. Understanding barter in Russia. In: Seabright, Paul (Ed.), The Vanishing Rouble. University Press, Cambridge. Commander, Simon, Dolinskaya, Irina, Mumssen, Christian, 2002. Determinants of barter in Russia: an empirical analysis. Journal of Development Economics 67, 275–307. Cottarelli, Carlo, Doyle, Peter, 1999. Disinflation in transition. IMF occasional paper 179. Washington, DC. Dibooglu, Selahattin, Kutan, Ali M., 2001. Sources of inflation in Poland and Hungary: Implications for full membership in the European Union. Working paper, 01. Southern Illinois University at Edwardsville. July. Domac, Ilker, Elbirt, Carlos, 1930, 1998. The main determinants of inflation in Albania. World Bank working papers. Gaddy, Clifford, Ickes, Barry W., 1998. To restructure or not to restructure: Informal activities and enterprise behavior in transition. Mimeo. Brookings Institution. Gara, Mario, 2001. The emergence of non-monetary means of payment in the Russian economy. Post-Communist Economies 13 (1), 5–39. Goskomstat, various editions. Main economic and social indicators. Goskomstat, Moskow. Goskomstat, 2001. Main economic and social indicators. Goskomstat, Moscow. Guriev, Sergei, Ickes, Barry W., 2000. Barter in Russia. In: Seabright, Paul (Ed.), The Vanishing Rouble. University Press, Cambridge. Guriev, Sergei, Makarov, Igor, Maurel, Mathilde, 2002. Debt overhang and barter in Russia. Journal of Comparative Economics 30 (4), 635–656. Guriev, Sergei, Kvassov, Dmitri, 2004. Barter for price discrimination? A theory and evidence from Russia, Submitted for publication, in International Journal of Industrial Organization. Hall, Robert E., 1997. Macroeconomic fluctuations and the allocation of time. Journal of Labor Economics 15 (1), S223–S250. Harrison, Sharon, 2001. Do sunspots reflect consumer confidence? An empirical investigation. Mimeo. Columbia University. Submitted for publication, in: Eastern Economic Journal. Hendley, Kathryn, Ickes, Barry W., Ryterman, Randi, 1997. Remonetizing the Russian economy. In: Broadman, Harry G. (Ed.), Russian Enterprise Reform: Policies to Further Transition. World Bank, Washington, DC. International Monetary Fund (IFM), various editions. International Financial Statistics. IMF, Washington. Kalra, Sanjay, 1999. Inflation and money demand in Albania. Russian and East European Finance and Trade 35, 82–105. Karpov, Peter, 1997. O prichinah nizkoi sobiraemosti nalogov (neplatezhei fiskalnoi sistemi), obshih prichinah ‘krizisa platezhei’ i vozmozhnostyah vosstanovleniya platezhosposobnosti rossiiskih predpriatii’ Otchet Mezvedomstvennoi Balansovoi Komissii (On the causes of tax arrears, general causes of the payment crisis and

314

B.-Y. Kim, J. Pirttilä / Journal of Comparative Economics 32 (2004) 297–314

plausibility of financial recovery of Russian firms). Report of the Inter-Agency Balance Sheet Commission, Moscow. December. Kim, Byung-Yeon, 2001. Determinants of Polish inflation: a structural cointegration approach. Discussion paper No. 16. Institute for Economies in Transition, Bank of Finland. Kohler, Marion, Britton, Erik, Yates, Tony, 2000. Trade Credit and the Monetary Transmission Mechanism. Bank of England. Korhonen, Iikka, 1998. An error-correction model for prices, money, output and interest rates in Russia. Economics of Transition 5, 33–44. Li, Kui-Wai, 1997. Money and monetization in China’s economic reform. Applied Economics 29 (9), 1139–1145. Marin, Dalia, 2002. Trust versus illusion: What is driving demonetisation in Russia? Economics of Transition 10 (1), 173–200. Marin, Dalia, Kaufmann, Daniel, Gorochowskij, Bogdan, 2000. Barter in transition economies: competing explanations confront Ukrainian data. In: Seabright, Paul (Ed.), The Vanishing Rouble. University Press, Cambridge. Marvasti, A., Smyth, David J., 1999. The effect of barter on the demand for money: an empirical analysis. Economic Letters 64 (1), 73–80. Meltzer, Allan H., 1960. Mercantile credit, monetary policy, and the size of firms. Review of Economics and Statistics 42, 429–437. Organization for Economic Co-operation and Development, 2002. In: OECD Economic Surveys: Russian Federations 2001–2002, vol. 2002/5. Paris. Oppenheimer, Peter, Granville, Brigitte, 2001. Russia’s post-communist economy. World Economy 2 (1), 149– 168. Petersen, Mitchell A., Rajan, Raghuram G., 1997. Trade credit: theories and evidence. Review of Financial Studies 10 (3), 661–691. Pesaran, H. Hashem, Shin, Yongcheol, 1998. Generalized impulse response analysis in linear multivariate models. Economic Letters 58 (1), 17–29. Pinto, Brian, Drebentsov, Vladimir, Morozov, Alexander, 2000. Give macroeconomic stability and growth in Russia a chance—Harden budget constraints by eliminating non-payments. Economics of Transition 8 (2), 297–324. Prendergast, Canice, Stole, Lars A., 1996. Non-monetary exchanges within firms and industry. NBER working paper series No. 5765. September. Ramey, Valerie A., 1992. The source of fluctuations in money: evidence from trade credit. Journal of Monetary Economics 30 (2), 171–193. Roche, Maurice, 2001. The rise in Dublin city house prices: bubble, fad or just fundamentals. Economic Modelling 18 (2), 281–295. Russian Economic Barometer, various editions. Monthly bulletin. Institute of the World Economy and International Relations, Moscow. Russian Economic Barometer, 2001. Monthly bulletin. Institute of the World Economy and International Relations, Moscow, 10, 4. Autumn. Russian Economic Trends, various editions. The Russian European Centre for Economic Policy, Blackwell Publishing, Oxford. Schoors, Koen, 2001. The credit squeeze during Russia’s early transition. Economics of Transition 9 (1), 205– 228. Schwert, G. William, 1989. Why does stock market volatility change over time? Journal of Finance 44 (5), 1115– 1153. Walsh, Carl E., 1998. Monetary theory and policy. MIT Press, Cambridge, MA. World Bank, various editions. World Development Indicators. World Bank, Washington.