Algeria : inflation in the transition

U s THE WORLD BANK Internal Discussion Paper EUROPE, MIDDLE EAST AND NORTH AFRICA REGION Report No. IDP-0084 Algeria: Inflation in the Transition by Ali Khadr and James Parks February 1991 Office of the Vice President Europe, Middle East and North Africa Region Discussion Papers are not formal publications of the World Bank. They present preliminary and unpolished results of country analysis or research that is circulated to encourage discussion and comment; citation and the use of such a paper should take account of its provisional character. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they represent. ALGERIA: INFLATION IN THL TRANSITION All Khadr and James Parks, EM2CO February 1991 The authors are Economists in the Europe, Middle East and North Africa Region (EMENA). This paper attempts to gauge the potential impact of full price liberalization on the general price level and future open inflation. First, the available evidence is evaluated to assess the degree of repressed inflation in Algeria and its potential impact on open inflation. The evaluation is based on a descriptive analysis of past macroeconomic policies, statistical comparison, and econometric analysis. Although there are clear indications that official prices are repressed in Algeria, the degree of repression and the impact of price liberalization on economic behavior are much less certain. Evaluation of the evidence suggests that, immediately lifting all remaining price controls would result in an immediate one-time jump in the general price level as measured by the official CPI. However, because the extent of parallel market activity is uncertain, the size of this increase and its impact on consumer behavior and future open inflation remain unclear. Given the uncertain impact of price liberalization, therefore, a flexible policy package is required for controlling inflation during the remainder of the transition. A tight medium-term monetary stance- -bolstered by fiscal restraint- -would be required if inflation accelerates and a more accommodating monetary stance preferred if precautionary measures taken to absorb liquidity and contain cost-push factors turn out to be deflationary. ALGERIA: INFLATION IN THE TRANSITION TABL! OF CWN=NS Page No I. BACKGROUND AND SUMMARY . . . .................... 1 II. OBJECTIVESOFTHISPAPER. . . . . . . . . . ............ 2 III. KEY RESULTS AND POLICY IMPLICATIONS . . . . . . . . . . . . . . . . 3 Key Results . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Policylaplications........................ 5 IV. PAST MACROECONOMIC POLICY AND REPRESSED INFLATION . . . . . . . . . 7 V. STATISTICAL INDICATIONS OF EXCESS LIQUIDITY . . . . . . . . . . . . 12 VI. EVIDENCE OF REPRESSED INFLATION FROM ECONOMETRIC ANALYSIS . . . . . 15 Determinants of Domestic Inflation . . . . . . . . . . . . . . . . 16 Price Inflation for Tradable Goods . . . . . . . . . . . . . 18 Price Inflation for Non-Tradable Goods . . . 18 Deriving and Estimating the Reduced Form Inflation Equation. 21 EstimatingMoneyDemand...................... 25 Testing Directly for Repressed Inflation . . . . . . . . . . . . . 31 Repressed Inflation Tests Based on Money Demand . . . . . . . 31 Repressed Inflation Tests Based on Private Savins . . . . . 34 BIBLIOGRAPHY . . . . . . . . . ... . . . . . . . . . 39 Tdo a m81111: *nepeaio eq- e I n ftl %poW asW thoe O& V i r au t n 6001 llgneo a oure e6 s "poessm The authors would like to thank Ajay Chhibber for his contribution and support, and Wiltse Bailey for his able research assistance. Helpful comments on an earlier version were provided by Fabrizio Coricelli, Bill Easterly, Laurie Effron, Wafik Grais, Nemat Shafik, and Steve Webb. ALGERIA: INFLATION IN THE TRANSITION I. BACKGROUND AND SMOIARY 1. Algeria has embarked on the transition from a centrally-planned to a market-oriented economy. While some price liberalization has occurred, reform efforts to date have focused mainly on the legal and institutional framework. One natural concern with designing and implem-inting a strategy for completing price liberalization is the potential impact on inflation. In the Algerian context, the concern with inflation is heightened by the possibility that past macroeconomic policies consistent with the build up of repressed inflation have resulted in excess liquidity in the economy. If there is excess liquidity, complete price liberalization could result in a substantial jump in the general price level which would risk setting off an inflationary spiral. 2. Although there are clear indications that official prices are repressed in Algeria, the degree of repression and the impact of price liberalization on economic behavior are much less certain. Evaluation of the evidence described below suggests that, immediately lifting all remaining price controls would result in an immediate one-time jump in the general price level as measured by the official CPI. However, because the extent of parallel market activity is uncertain, the size of this increase and its impact on consumer behavior and future open inflation remain unclear. If the parallel market is, in fact, small, then the inflationary consequences of price liberalization are potentially significant. In this case, precautionary measures to rapidly absorb -2- liquidity and contain cost-push factors would be required to limit the risk of an inflationary spiral. However, if the parallel market is large, then these measures run the risk of deflating the economy. 3. Given the uncertain impact of price liberalization, therefore, a flexible policy package is required for controlling inflation during the remainder of the transition. A tight medium-term monetary stance--bolstered by fiscal restraint- -would be required if inflation accelerates and a more accommodating monetary stance preferred if precautionary measures taken to absorb liquidity and contain cost-push factors turn out to be deflationary. II. OBJECTIVES OF THIS PAPER 4. This paper attempts to gauge the potential impact of full price liberalization on the general price level and future open inflation. First, the available evidence is evaluated to assess the degree of repressed inflation in Algeria and its potential impact on open inflation. The evaluation is based on a descriptive analysis of past macroeconomic policies, statistical comparison, and econometric analysis. While not fully conclusive, this assessment serves as a aseful basis for discussing different policy options for dealing with inflation in the remainder of the transition. 5. The remainder of the paper is divided into four sections. Section III presents the key results and their implications for policy. Section IV provides a descriptive analysis of past macroeconomic policies in Algeria and demonstrates how these policies were consistent with repressed .3- inflation. Section V evaluates the available statistics for evidence of a money overhang and compares liquidity measures in Algeria with those of neighboring countries and other developing countries. Section VI presents an econometric analysis of repressed inflation in Algeria, including short studies of irflation determination and money demand, as well as direct tests of repressed inflation. III. KT RESULTS AND POLICY IMPLICATIONS Ky Results 6. Analysis of the available evidence from the past two decades provides substantial indication of repressed inflation in Algeria with respect to official prices. For most of the period following independence in 1962, Algeria pursued macroeconomic policies, including monetization of public sector deficits and imposition of pervasive price controls, consistent with the buildup of excess liquidity, or equivalently, the repression of inflation. Currently, as a result of these policies, measures of liquidity relative to the level of official transactions do in fact indicate repressed inflation. 7. Results based on econometric analysis of official data provide additional support for the hypothesis of repressed inflation by demonstrating that: (i) standard cost-push and demand-pull factors are not reflected in the behavior of official consumer price index (CPI) inflation in Algeria, (ii) money demand cannot be captured by standard specifications and more complex models indicate a considerable disequilibrium in the money market, and (iii) direct - 4 . econometric tests for repressed inflation, while not unambiguous, do provide some indication that official prices are below market-clearing prices. 8. Estimates indicate that repressed inflation in Algeria in 1989-- from statistical comparison and econometric analysis--may have been as high as 60% to 80%. In other words, these estimates indicate that, had prices been fully decontrolled in 1989, the CPI could have experienced a one-time increase as large as 60% to 80%. Such a jump would have pushed official inflation for the year to some 70% to 90%, compared with the approximately 10% registered. 9. Of course, all of these statistical and econometric measures may be misleading because they are based on official data which do not fully take into account parallel market activity, which may be significant. To the extent that non-recorded parallel market activity exists, it can affect the true size of the money overhang and .he inflationary potential of price liberalization. First, if a large number of goods are being produced and imported through unofficial, unrecorded channels, then the quantity of goods in the economy is being underestimated and the overhang is lower than it appears from official statistics. In this case, freeing prices will cause a smaller jump in the CPI than predicted by official data. Second, if many "official" goods, i.e., those goods produced or imported through official channels, are being resold at higher prices on the parallel market, then a large amount of the repressed inflation has already been "freed*, or validated, on the parallel market. This second type of parallel market activity does not reduce the jump in the CPI following price liberalization as predicted by official data, but consumers- -having already adjusted to higher parallel market prices--will react less. -5- 10. While no comprehensive data exist on parallel market activity in Algeria, a preliminary study by the National Planning Council (CNP) suggests that it is fairly small relative to official GDP.1 Moreover, the study suggests that only a small fraction of officially repressed inflation has already been validated by resale of official goods on the parallel market. However, the CNP study is not based on a detailed survey of the parallel market and therefore cannot be considered definitive. The parallel market may in fact be large enough to account entirely for the money overhang indicated by official data. Polic Imolications 11. Clearly, rapid price liberalization could cause a substantial official price shock. The results presented here indicate that in 1989 a one-shot liberalization could have entailed a one-time jump in the official CPI as large as 60% to 80%. While not definitive, the CNP's study on parallel markets suggests that only about 4% of this jump has already been validated by parallel market prices. If the CNP study is correct, then the remaining increase would represent a substantial price shock. Of the potential measures to absorb liquidity, monetary reform would appear to be the most able to help reduce this initial shock. Other measures, such as sale of state assets and repayment of central bank debt by the Treasury would likely require a significant amount of time to take effect. Monetary reform could involve converting a portion of 1/ National Planning Council, "Documentation relative A l'dvaluation do I'dconowle inforelle et ses cons4quences sur les dqullbres macro- 6conomlques", CNP, 1990. -6* deposits into longer-term Treasury bonds. Explicitly backing these bonds by future petroleum tax revenues would enhance the reform's credibility.V 12. If a large amount of liquidity were not rapidly absorbed prior to completing price liberalization, and official prices did jump substantially, then measures to control subsequent cost-push factors would be necessary t% limit the risk of a subsequent, accelerating spiral of open infiation. Wages are probably the key cost-push factor over the near to medium term. While expanding the current system of consumer price subsidies could cushion the initial price shock somewhat, an unvalidated jump in the CPI on the order of 55% to 75% would put almost irresistible pressure on wages. To limit these increases, the government could lead the way by limiting civil service salary increases to the extent possible. Moreover, strong enforcement of recent legislation to liberalize the labor market would allow the some 23% of the labor force currently unemployed to act as a significant restraint on wages. 13. However, if the parallel market is indeed small, then the results presented here indicate that it is unlikely--no matter which measures are undertaken--that sufficient liquidity absorption and cost-push containment can occur to eliminate any subsequent inflationary pressure. In this case, consistently tight monetary policy over the medium term, i.e., a policy which does not fully accommodate wage and price increases, would be necessary. By generating fiscal surpluses and recycling them through the banking systga, tight fiscal policy would help the Central Bank limit money creation. Although it In 1989, Treasury petroleum tax revenues amounted to 12.5% of GDP. - 7 - could entail costs in terms of temporary output losses, a combination of tight monetary policy and fiscal restraint would be the only way to eventually tame inflationary pressures set off by price liberalization. On the other hand, if the true money overhang is small, then a more accommodating medium-term monetary policy may be required to offset the deflationary effects of any precautionary measures previously taken. 14. In fact, tighter monetary and fiscal policies have already eroded the money overhang. The ratio of M2 to GDP, which peaked at over 95% in 1988, dropped to about 85% in 1989. Preliminary data indicate that this trend continued in 1990 as additional liquidity was absorbed. However, annual inflation increased to 17% and is expected to exceed 20% in 1991. This acceleration, together with the recent agreement between the government and Algeria's main trade union--the Union G6n6rale des Travailleurs A1gdrlens--to increase the minimum wage by some 80%, indicates that even tighter macroeconomic policy will probably be needed to contain inflation in the future. IV. PAST MACROECONOMIC POLIC! AND REPRESSED INFLATION 15. Historically, Algeria's macroeconomic policy has been characterized by: (i) large, heavily monetized, public sector deficits, (ii) price controls and (iii) goods rationing. Centrally managed public enterprises (PEs), operating under pervasive price controls, have allocated resources inefficiently. Excessive real wages and overemployment have inflated PE operating expenditures, while a deliberate policy of capital intensive development has inflated investment requirements. In an atmosphere of lax financial discipline, PEs have * 8 - borrowed heavily from the banking system to cover the resulting operating losses and resorted to soft loans and capital transfers from the Treasury to finance investment. In parallel with PEa, the Treasury has been running its own deficits. A lack of transparency and efficiency have resulted in a shrinking tax base and over-dependence on petroleum tax revenues. A rapidly growing government labor force and generous wage increases have increased Treasury current expenditures, while the proliferation of large-scale public works projects has kept Treasury investment expenditures high. 16. The public sector deficits resulting from PE inefficiency and lax fiscal policy have resulted in large Treasury borrowing requirements and high monetary growth. Except for 1981, the Treasury's annual borrowing requirement exceeded 5% of GDP between 1975 and 1988. Between 1982 and 1988, it averaged more than 10% of GDP (see Graph #1). Because Algeria's financial system is relatively undeveloped, this borrowing requirement has been covered largely by advances to the Treasury from the monetary system. For example, the stock of monetary system claims on the government as a percentage of GDP doubled from 20% in 1980 to 40% in 1987 (see Graph #1). Monetary growth has been correspondingly high. Nominal broad money (M2) grew at an average annual rate of 23.6% between 1971 and 1980 and 14.2% between 1980 and 1989 (see Graph #2). Narrow money (Ml) growth paralleled that of broad money (see Graph #2). - 9 - Graph #1 Treasury Borrowing Requirement ana ctaisf of th Montary syetem ÞPrcent of G0D -10 8 70 71 7 73 74 7e 7 77 7e 70 8o 8q 83 048e6887 8 89 Yes>r -= Trwuy 9ar~Ig~qIramnt St~ M of netey ymom cir on rea. Graph #2 NominaLl Broad Money Growth ana cPii nfiat ion Prcent 140 71 72 73 74 75 7t 77 70 74 80 8i Gå 83 64 65 86 87 88 89 - Nom Now h fe4~ CP i l nf lati on HM m bI =,MNrrw G~e ~wr 10 - 17. Due to pervasive price controls, officially measured inflation has remained quite low, despite the high rate of money growth. As shown in Graph #2, inflation--as measured by the urban consumer price index--has for the most part remained in the 5-15% band over the past two decades. The historical extent of price controls is evident from a recent World Bank estimate that, as of 1986, 47% of goods in the Algerian consumption basket were subject to fixed prices and an additional 11% were subject to price homologation. Furthermore, the study found that, as of 1984, 72% of industrial output was subject to fixed, stabilized, or special prices wile an added 11% was subject to homologation. Although a new price law was passed in July 1989 creating a category of "declared" prices which are freely determined subject only to ex ante notification and ex post monitoring for abuse, the prices of some 53% (by weight) of the CPI basket remain subject to direct controls. 18. The combination of large, monetized public deficits and pervasive price controls has led to widespread rationing in the private sector and forced saving. Private investment in Algeria, including residential investment, has been hampered by a strict rationing of credit as well as capital goods, particularly imported capital goods, since the private sector has also been deprived of foreign exchange. At the same time, pervasive consumer price controls and goods rationing appear to be preventing consumer goods markets from clearing. The apparent money overhang implies that "forced saving" has been Kumar, A., "Measuring the Extent of Price Controls and Measuring the Share of Imports in Consumption", World Bank mimeo, 1989. In the Algerian context, homologation refers to government surveillance with potential for occasional intervention to control prices. - 11 - imposed on the private sector, i.e., at official prices people would prefer to consume more and save less than they are currently able to., 19. The resulting asset accumulation by the private sector has further increased demand pressure, leading to the emergence of parallel markets. The asset accumulation implied by the cumulative excess of private saving over investment has been substantial. By one World Bank estimate, the excess of private saving over private investment reached a high of almost 15% of GDP in 1988. Compounding demand pressures, this accumulation has largely taken the form of narrow money because: (i) other domestic financial assets are cumbersome to use and have poor yields, and (ii) the capital account in the balance of payments is strictly administered, which prevents the private sector from acquiring foreign assets through officially recognized channels.' The buildup of private sector demand pressure has resulted in significant parallel market activity. In more technical terms, if the expected relative official absorption price were equated with the marginal rate of substitution between present and future absorption, the implied rate of present absorption would be higher. For an insightful description of past macroeconomic policies in Algeria and the implications for repressed inflation, see: A. Gelb and P. Conway, "Algeria: Windfalls in a Socialist Economy" in A. Gelb and Associates, Oil Windfalls: Blessing or Curse?, World Bank, 1988. At the end of 1989, the ratio of parallel to official exchange rates (AD/US$) stood at some 460% and unofficial imports during the year amounted to some US$633 million, about 1.3% of official GDP. - 12 - V. STATISTICAL IDICATIONS OF ECESS LIOUIDITY 20. Official statistics reflect the buildup of liquidity to apparently excessive levels, indicating repressed inflationary pressure. The ratio of broad money (M2) to official GDP, which stood at 65% in 1978, reached a high of over 95% in 1988 before dropping to about 85% in 1989 (see Graph #3). In comparison, the 1988 ratios of M2 to GDP for neighboring Morocco and Tunisia were some 48% and 521, respectively; while the 1988 ratios for Turkey and Pakistan were some 28% and 42%, respectively. The contrast with respect to narrow money (Ml) is even more striking. The ratio of Ml to GDP in Algeria reached some 82% in 1988, before declining to 69% in 1989. Comparable ratios for Morocco and Tunisia in 1988 were about 34% and 28%, while those for Turkey and Pakistan were approximately 12% and 28%, respectively. LIquidity Ratios c in wercent: Parcent 200 100 717724757770700 100200 57 09 - Officlai Mi to GOP a-SOffictal Ma to GDP X-4OffIntel M2 to Private Conamuption * 13 - 21. Ratios of money to private consumption show a pattern similar to that of GDP-based liquidity measures. The official ratio of (2 to private consumption in Algeria was about 180% in 1988 before dropping to 158% in 1989. In 1988, comparable ratios were: 70% in Morocco, 81% in Tunisia, and 57% in Pakistan.U Such high liquidity ratios indicate a money "overhang" in Algeria, at least with respect to official prices. An increase in official prices of some 77% would be required, for example, to match the ratio of M2 to private consumption in Tunisia. 22. The increase in these money overhang indicators over the past decade suggests that the degree of repressed inflation is increasing. Interestingly enough, the ratio of M2 to private consumption indicates a steeper trend (see Graph #3) than the GDP-based measures. This may suggest that, if repressed inflation is indeed present and has increased over the 1980s, a significant proportion of the corresponding increase in private sector rationing which would account for it may be occurring in markets for consumption rather than investment goods. 23. However, to the extent that they do not fully incorporate parallel market activity, these official liquidity measures actually overestimate the true size of the money overhang. Incorporating parallel market activi reduces the overhang in two ways. First, a portion of the apparent "excess" liquidity finances transactions of unofficial goods produced and imported on the parallel V Ratios for Algeria are from official sources, those for Morocco, Tunisia, Turkey, and Pakistan are from International Financial Statistics, IMF, 1990. The private consumption figure for Turkey in 1988 is not available in IFS. - 14 - market. Adding these transactions to official volumes reduces the overhang. Second, some of the "excess" liquidity finances the resale of official goods at higher parallel market prices. Valuing such goods at the higher parallel market prices further reduces the overhang. 24. While no comprehensive estimate of parallel market activity in Algeria is currently available, preliminary information would indicate that it is not large enough to eliminate the entire money overhang as measured by official data. First, with respect to the construction sector, where parallel market activity is concentrated, the official GDP figures already incorporate parallel market output.V Second, a recent CNP study on parallel market activity indicates that in 1989: (i) nominal parallel market value-added in sectors other than construction equalled some 7% of official GDP, (ii) that nominal informal private consumption amounted to some 28% of official private consumption, and (iii) of this 28%, only some 4% was due to resale of official goods at higher prices on the parallel marketY LI The volume of informal output in the construction sector is computed using: (i) the volume of intermediate inputs into the construction sector which is assumed to be produced entirely in the official economy and, therefore, is measurable; and (ii) an established input/output coefficient for the sector. Subtracting the total production given by (i) and (ii) from official production gives an estimate of parallel market production, which together with (ii) gives a figure for parallel market value-added. V See: "Documentation relative A l'6valuation de 1'6conomle informelle et ses consdquences sur les dquilibres macro-dconomIques", CNP, 1990. The discrepancy between the value-added and consumption corrections is caused by the large number of unofficial, but legal, imports, referred tol as 01aportations sans palement". - 15 - 25. Incorporating the CNP study' s corrections into the liquidity measures for 1989 gives a ratio of M2 to GDP of 79%, a ratio of Ml to GDP of 64%, and a ratio of M2 to private consumption of 123%. All of these figures remain significantly above comparable ratios for Tunisia, Morocco, Turkey and Pakistan. According to the CNP study, if one assumes that unofficial goods are sold at market-clearing prices, an increase of some 71% in the official CPI would still be required to reach the ratio for Tunisia.1 VI. EVIDENCE OZ REPRESSED INFIATION FROM ECONOMETRIC ANALYSIS 26. Beyond descriptive evidence of past macroeconomic policies consistent with forced savings and statistical indications of excess liquidity, econometric analysis provides some additional evidence in support of the repressed inflation hypothesis. This evidence takes three forms. First, attempts to explain the behavior of official consumer price index (CPI) inflation in Algeria demonstrate that it is not a function of standard cost-push and demand-pull factors, thereby implying that these factors have been repressed. Second, money demand behavior is not well captured by conventional specifications which assume money market equilibrium, indicating a forced accumulation of monetary assets. Finally, while In interpreting an apparent money overhang in the light of parallel activity, it is important to distinguish between the resale of official goods at higher parallel market prices and transactions of unofficial goods. Consider a given value of parallel market activity and a given, officially measured, money overhang. If parallel market activity consists essentially of the resale of official goods, then freeing prices will result in a relatively large jump in the official CPI--with a relatively large portion of that jump having been already validated in the parallel market. However, if parallel market activity consists mainly of transactions of unofficial goods, then freeing prices will cause a smaller jump in the official CPI--the difference being that a smaller portion of that jump would have already been validated by the parallel market. - 16 - the results are not unambiguous, some direct econometric tests detect a statistically significant degree of repressed inflation.1V Determinants of Domestic Inflation 27. The purpose of this section is to dewonstrate that the effects of demand-pull and cost push factors are not well reflected in official CPI inflation. If inflation were not suppressed, one should be able to capture it as a function of standard demand-pull and cost-push factors. Our inability to do so indicates that these factors have been repressed in Algeria through a combination of price and quantity controls.1W 28. Common approaches to modelling consumer price inflation with a pass- through equation start by specifying CPI inflation as a function of price inflation for tradables and price inflation for non-tradables. For a small country, in the absence of price controls and non-tariff barriers, inflation in tradables is then assumed to be a function of FCU import price inflation and the ly Other econometric studies have indicated the existence of repressed inflation in Algeria. For example, Gelb and Conway have estimated that "without rationing the Algerian price level would have had to increase in 1974-78 by some 40% more than its historical trend to hold consumption at pre-1973 levels, even under the favorable assumption that desired real balances were unaffected by inflation*, (Gelb and Conway, 1988). Using tests based on switching regressions, Standaert (1989) also obtains results confirming the frequent occurrence of repressed inflation over the 1970s. Fimally, following the methodology of Feltenstein et al. (1987) and using data through 1986, Rocha (1988) finds a significant degree of repressed inflation in Algeria. 1V However, this result must be qualified by the fact that a lack of complete wage data prevented the effects of wages on CPI inflation from being explored. - 17 - exchange rate. Assuming cost-plus pricing for non-tradables, inflation in non- tradables is determined by a markup equation over costs that generally incorporates, inter alia, the unit costs of intermediate goods and labor.1V This methodology yields a reduced form pass-through equation for inflation that can be estimated directly. 29. In developing a specific inflation equation for Algeria, we followed the general approach of a similar study for Zimbabwe (Chhibber et al. [1989]), which specifies consumer price inflation as a weighted average of price inflation for tradable goods and price inflation for non-tradable goods. In line with the approach outlined above, price inflation for tradables is assumed to be determined exogenously by movements in import prices and the exchange rate. Price inflation for non-tradables is determined by a markup pricing rule that takes into account unit labor costs and the costs of imported intermediate goods.1 In addition, the markup itself is assumed to be a function of aggregate demand pressure. 30. Formally, consumer price inflation is given by: f - PVT + (1-P)IrN (5) 1V Common proxies for demand pressure are deviations of real income from trend and measures of the excess supply of real money balances. IV In Algeria, one would expect a third category of goods, namely commodities with controlled prices, to be extremely important in determining inflation. We have not exhaustively explored the determinants of changes in controlled prices in detail. However, as axplained below, tentative proxies for price controls performed poorly in the regression equation for inflation. * 18 - where w represents consumer price inflation,1V and the subscripts refer to the price indices for tradable and non-tradable consumer goods. Note that: O<p<l. Time subscripts (denoted by t) are omitted when there is no risk of confusion. 31. In what follows we give an account of price inflation in tradable and non-tradable goods in turn, then we turn to the derivation and estimation of the reduced form inflation equation. Price Inflation for Tradable Goods 32. We assume the following: X F - VF + 6o (6) where F is the inflation rate of the foreign currency price of imports, and 6o is the proportional rate of change of the official exchange rate. Equation (6) is the conventional specification of the rate of change of the prices of tradables. Price Inflation for Non-Tradable Goods 33. Price inflation for non-tradable goods is assumed to be given by the 1V This is computed as the proportional rate of change, or log difference, between period t and period t-l. - 19 - following markup equation: iN - a, me + a2 wc + a3 ED (7) where: mc and we denote the proportional rates of change of the unit costs of imported intermediate goods and wages, respectively. ED is a term (to be defined shortly) that capturea aggregate demand pressure. 34. We now look at how the three components of equation (7) are determined in turn. 35. The unit cost of imorted intermediate goods: This is assumed to change at the same rate as the price of tradables:-W mC - WF + 60 (8) 36. Unit labor costs: A natural measure of the change in unit labor costs is the difference between the change in the real wage and the change in the productivity of labor. For small variations of wages, prices, and productivity, this can be written as: we- - w - At (9) 1W That is, we do not distinguish between the prices of tradable consumer goods and the prices of imported intermediate goods at this stage. * 20 - where: W and AL represent the rate of growth of nominal wages and labor productivity, respectively. In words, equation (9) states that, if the real wage is growing at the same rate as productivity, the cost of a labor-hour in terms of goods is changing at the same rate as the yield of a labor-hour, so the real cost to employers per unit of labor is constant. 37. The difficulty, of course, arises in finding an adequate measure of the change in productivity. In the absence of data on productivity, one approach is to assume that it follows a time-trend.1t The simplest such case is to assume that productivity growth is a constant, 1: AL - 1 (10) Substituting (10) into (9) then yields our measure of the change in unit labor costs that goes into the inflation equation for non-tradables, given by (6). 38. Aregate demand Dressure: In countries where asset markets are not well-developed, it is reasonable to assume that any disequilibrium in the goods market spills over mainly into the money market. That is, any unfulfilled demand for goods takes the form of (involuntary) accumulation of money balances. Using this reasoning, we can write: ED - log(M/P) - log(fo) (11) IV Another approach that may be worth pursuing is to relate labor productivity to the productivity of capital as captured by une ICOR. - 21 - where M is the nominal stock of money and ad is the demand for real money balances. In words, the excess demand for goods is reflected in an excess supply of money. Money demand is specified in the usual way: log(ml) - ba + b, log y + b2 i + b3 We (12) where y is real income, i is the nominal interest rate, and we is expected inflation. For the moment, we assume that expected inflation is given simply as: me - N (13) Substituting (12) and (13) into (11) and rearranging the result yields: ED - log(/P,) - (1 - b3) x - bo + b, log y + be i (14) Derivin and Estimating the Reduced Form Inflation Eauation 39. Using equations (6) - (13) in (5), we obtain the following reduced- form pass-through equation for the rate of inflation: W W Ao* +A, WF + a o& + A3 + A4 log(Mt/Pt-) + As log y + As 1 (15) where the coefficients are given as follows: AO - - (1-p) (a,1 + abo) /D, 22 * A, - (p + As *A4, A, - aa(l-p)/D, As - a(l-p)/D, A, - - ab(-)D As - - asb,(l-)/D, and D - (1 + (1-p)[a, + a3(1 + b,)]). Given the form of the structural equations, D should be positive. If so, then a priori we would expect all the coefficients to be positive, except A5 and A, (and possibly the constant term). 40. Equation (15), then, describes inflation as linear combination of cost push and demand pull factors. The constant term captures, inter aIla, the effect of changes in productivity. This term states that, cateris paribus, increases in productivity growth will reduce the inflation rate. Terms two and three capture the effects of world inflation and devaluation, respectively. The fourth term captures the positive effect, ceterls paribus, of nominal wage growth on inflation. The last three terms introduce the effect of excess demand as a function of money market disequilibrium, with increases in money balances aggravating inflation, ceteris paribus; while increases in real output and inflation have the opposite effect. 41. Several different specifications based on equation (15) were tried, but all yielded very poor results. We began by regressing CPI inflation on US$ * 23 - import price inflation and official devaluation via-A-via the US$. The result is a very poor fit, given by:W a - 0.089 - 0.027 ap + 0.001 6o (16) (7.88) (-0.33) (0.01) (Sample 1971-1989; RO - 0.007; DW - 0.92) 42. In other words, the government, through the use of subsidies and price controls, appears to be successfully protecting consumer prices from any variations attributable to changes in the prices of imported goods. This result, combined with the fact that the estimated average economy-wide markup over costs has fluctuated considerably in the 1980s,1V indicates that consumers have been protected from world price shocks at the expense of PE losses. 43. To incorporate excess demand terms, we then added broad money deflated by the preceding year's CPI, the log of real GDP, and the six-month deposit rate, i, which follows the same path as most other rates in Algeria. As the results show, none of the coefficients turned out to be significant. Moreover, the exchange rate and interest rate terms have the wrong signs. w - 2.909 + 0.076 w, - 0.171 do + 0.258 log(N/P,) - 0.366 log y (17) (1.42) (0.89) (-1.18) (1.89) (-1.52) + 0.281 1 (0.21) (Sample 1971-1989; Ra - 0.405; DW - 1.87) In all the regressions that follow, t statistics are given in parenthesis. ill Over the 1980s, the estimated average markup fluctuated from a high of some 401 in 1984 to a low of about 21% in 1989. - 24 - 44. Finally, to investigate the possibility that observed CPI inflation in Algeria is a direct function of the system of price controls, we extended the inflation equation by including a term that captures the difference between inflation in the "miscellaneous" component of the CPI and overall CPI inflation (this difference is denoted DPI). The hypothesis is that: (i) the "miscellaneous" component of the CPI reprezents those goods whose prices have been least held down, and (ii) overall CPI inflation and DPI are negatively correlated as a general tightening of price controls slows inflation and increases DPI. Equation (18) demonstrates that the explanatory power of this price control proxy is not very good, although the sign is at least consistent with the hypothesis. The exact specification used was as follows: CPI inflation was regressed on LCU import price inflation, broad money deflated by the lagged CPI, the log of real GDP, the six-month deposit rate, and the proxy for price controls referred to above (DPI). w - 2.779 + 0.005 wF* + 0.234 log(M/P,.,) * 0.340 log y (18) (1.34) (0.06) (1.75) (-1.41) - 0.885 1 - 0.288 DPI (-0.73) (-1.42) (Sample 1971-1989; Ra - 0.387; DW - 1.72) 45. In the above estimations, only the money supply and GDP components of the excess demand term and the price control proxy give any hint of affecting CPI inflation. Unfortunately, lack of a complete data series has prevented wage inflation from being included in the regressions at this stage. Nonetheless, the impact of other cost-push factors is completely absent. Such a pattern of * 25 - results corroborates the hypothesis that, because of a combination of pervasive price controls and goods rationing, the official consumer price index is not accurately reflecting inflationary pressures in the Algerian economy, especially those resulting from external cost-push factors. Estimating Money Demand 46. The purpose of this section is to support the hypothesis of money market disequilibrium, i.e., the presence of a money overhang, by demonstrating that the evolution of officially measured real money balances in Algeria is not well captured by money demand specifications which assume money market equilibrium. To start, we estimated two specifications of a conventional static money demand equation. In the first, the logarithm of the stock of broad money deflated by the CPI was regressed against the six-month deposit rate, the log of real GDP, and CPI inflation as a proxy for expected inflation. The result is given by: log(M2/P) - -14.873 + 1.742 log y - 1.046 1 - 0.233 v (19) (-18.82) (24.82) (-0.59) (-0.47) (Sample 1971-1989; Ra - 0.984; DW - 1.34) A similar result was obtained for narrow money demand: log(Ml/P) - -14.351 + 1.699 log y - 3.691 1 - 0.197 w (20) (-16.69) (22.24) (-1.93) (-0.37) (Sample 1971-1989; Ra - 0.979; DW - 1.21) - 26 - 47. At first glance, these specifications appear to fit reasonably well, with high Rls and all the appropriate signs. However, the income elasticities are extremely high. Moreover, the Durbin-Watson statistics are low, indicating that equations (19*20) are mis-specified. These facts point to a spurious relationship bFetween real money demand and real income, with the high income elasticity of demand reflecting a forced accumulation of liquid assets, or equivalently, an excess of liquidity. 48. A re-estimation of money demand in differences clearly demonstrates that equations (19-20) are mis-specified: Alog(M2/P) - 0.051 + 0.856 Alog y - 3.491 Ai - 0.572 A* (21) (2.30) (3.09) (-1.68) (-1.39) (Sample 1972-1989; R' - 0.543; DW - 1.41) Alog(Ml/P) - 0.048 + 0.849 Alog y - 5.529 Ai - 0.574 As (22) (2.02) (2.82) (-2.44) (-1.28) (Sample 1972-1989; RI - 0.576; DW - 1.21) As equations (21-22) demonstrate, re-estimating money demand in differences results in a much lower income elasticity. This result confirms the hypothesis of a spuriously strong relationship in levels between real money balances and income. Consequently, it corroborates the finding of excess liquidity. * 27 - 49. A test for cointegration shows, however, that the simple difference equations (21-22) remain mis-specified. Two non-stationaryEw series are cointegrated if they tend to move together in time such that a particular linear combination of the series is stationary. In other words, cointegrated series can be thought of as having a long-run, stationary, steady-state relationship despite short-run divergences from this long-run relationship. RV This leads one to include an "error-correcting term" in the dynamic specification of the functional relationship between cointegrated variables. This term acts as a "feedback loop" which corrects for short-run divergences and tends the relationship back to its long-run "equilibrium". In this instance, the error correction term would be formally defined as the difference between the level of real money balances last period and the equilibrium level of money demand last period as predicted by income, interest rates, and inflation. 50. Based on the Dickey-Fuller test, both broad and narrow real money balances are cointegrated with real money demand as specified in equations (19-20). The first requirement for cointegration is that the real money supply be non-stationary. A Dickey-Fuller test indicates that, at the 5% level, we cannot reject the hypothesis that broad and narrow real money balances are non- stationary (see Table 1). Although equilibriu. real money demand is unmeasurable in this case, Table 1 also shows that real income, inflation and interest rates all appear to be non-stationary. Given non-stationarity, if real money supply 2 A series is non-stationary if it does not fluctuate around a constant mean value with a constant variance and autocovariance. IV Another example of cointegrated series might be consumption and permanent income. - 28 - and demand are cointegrated, then their long-run relationship should be captured by equations (19-20). Therefore, we can test for cointegration by testing the residuals (e2 and el) of equations (19-20) for stationarity. The results of these tests are also shown in Table 1. Clearly, both broad and narrow real money balances are cointegrated with real money demand as specified. Table #1 Dickey-Fuller Test ResultsW Variable # of Lagged DF-Statistic Durbin-Watson Dependent Variables Statistic log(N2/P) 1 1.54 1.86 log(Ml/P) 1 0.91 1.81 log y 1 3.37 1.32 1 1 1.69 1.12 Jr 1 -0.36 2.07 e2 1 -4.51 2.31 el 1 -4.79 2.32 Critical value of the D-Statistic at 5% level: -1.95 For a given variable X, the Dickey-Fuller test used here involves: (i) running the following regression: AX, - aX- + ]86A4 + Et (ii) forming the t-statistic, known in this context as the DF statistic, for a, and (iii) testing the null hypothesis that a is greater than or equal to 0, i.e., that the series is non-stationary, by comparing the DF statistic with its critical value. If the DF-Statistic exceeds the critical value, we fail to reject the null hypothesis. The order of i is set to ensure that the residual series is white noise. For more information concerning cointegration and the specification of Dickey- Fuller tests, see: A. Harvey, The Econometric Analysis of Time Series., 2nd Ed., MIT Press, 1990, Chapter 8; and R. MacDonald and C. Kearney, "On the Specification of Granger-Causality Tests Using the Cointegration Methodology", Economics Letters, No. 25, 1987. - 29 - 51. Consequently, we re-estimate equations (21-22), adding error- correction terms, denoted EC2 and EC1, respectively. For the purposes of estimation, EC2 and EC1 are given by the residuals from the corresponding static regressions above, i.e., equations (19-20), respectively. Note that equations (23-24) have also been corrected for first-order autocorrelation in the residuals by applying the Cochran-Orcutt procedure. The autocorrelation coefficient is denoted RHO and all variables are shown in transformed form, e.g., Ai* - ALt - RHO*Aij.j. Alog(M2/P)* - 0.037 + 0.622 Alog y' - 1.200 Ai - 0.760 Aw' + 0.895 EC2* (1.26) (2.29) (-0.79) (-3.34) (4.04) RHO - 0.621 (23) (3.27) (Sample 1972-1989; R2 - 0.761; DW - 2.31) Alog(Ml/P)* - 0.038 + 0.536 Alog y' - 3.535 Ai* - 0.769 As* + 0.830 EC1' (1.26) (1.90) (-2.31) (-3.27) (3.93) RHO - 0.619 (24) (3.25) (Sample 1972-1989; S2 - 0.809; DW - 2.16; ) 52. Relative to the conventional specification in levels, the above results give substantially lower income elasticities, supporting the hypothesis of a spuriously strong relationship between the level of real money balances and GDP. Compared to results from other countries, this last pair of income elasticities is more reasonable than the estimates from the conventional specifications. - 30 - 53. However, the good fit of equations (23-24) cannot be used as evidence of money market equilibrium in Algeria. The coefficients on the error- correcting terms EC2 and EC1 are positive, indicating instability. In other words, if real money balances exceeded long-run demand yesterday, then through EC2 and EC1, equations (23-24) state that they will likely exceed long-run demand by even more today, and so on. In this way, the error-correction terms in (23- 24) capture the money market disequilibrium- -while the terms for inflation, interest rates, and GDP, represent equilibrium money demand. Consequently, the disequilibrium effect is separated from the income effect, giving a more realistic income elasticity. 54. Because they can be interpreted as distinguishing between equilibrium and disequilibrium effects, equations (23-24) can be used to gauge the present degree of money overhang. From equations (23-24) one can estimate a path for equilibrium money demand by: (i) assuming that the money market was in equilibrium in 1971, the first data year; and (ii) thereafter setting supply equal to the demand given by the inflation, interest rate, and CDP terms (i.e., setting ECa and EC, to zero). These assumptions give an equilibrium ratio of N2 to GDP of about 53% for 1989, 32 percentage points lower than the official ratio of 85%. For Ml, they give some 40%, 29 percentage points higher than the official ratio of 69%. Therefore, this exercise implies that an increase in the official price level of somewhere between 60% (from (23)) and 73% (from (24)) is required, caterls parlbus, to restore current real balances to their underlying steady-state demand. . 31 - Testing Directly for Repressed Inflation 55. In this section we test directly for repressed inflation using two distinct tests. The first test is based on money demand and the second on private savings. The second test was previously applied to Algeria by Rocha (1988). The test results are somewhat ambiguous. While the first test does not indicate repressed inflation, the second test supports the repressed inflation hypothesis. The second test gives an estimate of repressed inflation for 1989 of some 66%. This result is consistent with the estimates obtained from the error-corrected money demand regressions above as well as the estimates of between 40% to 50% for the years 1983-86 obtained by Rocha (1988).W R12ressed Inflation Tests Based on Money Demand 56. The money demand-based test for repressed inflation leveloped by Feltenstein (1989) and subsequently applied to C6te d'Ivoire has two basic elements. First, a distinction is drawn between the observed index of consumer prices, P, and a "true", or notional, price index PT, which would clear the goods market in the absence of price controls. The wedge between the two indices is assumed to be proportional to some measure of the money overhang. Second, real demand for quasi-money, qi, depends on real income and the expected rate of inflation in P.. To complete the specification, we assume that the real stock of quasi-money adjusts immediately to (ex ante) real demand for quasi-money and at R. Rocha, 9g. Sit., p. 30. - 32 - model inflationary expectations as static. In equation form, these assumptions can be expressed as: log P - log P + a log (Mi/C) (26) qa - qmr ao + a, log y + a2 rW (27) IrT" I sy(28) where: XT - log PT -log PT., (29) qmr - log (QM/PT) - qad (30) 57. In this specification, y is "real" GDP measured in volume terms, Ml is the stock of money and QM is the nominal stock of quasi-money. Equation (26) relates the "true" price level to the observed price level. The wedge between the two is assumed to be proportional to the ratio of narrow money to nominal private consumption (C). This part of the argument thus hinges on a transactions velocity.W Equation (28) then gives the expectations mechanism for the evolution of the virtual price level. Finally, the "true" purchasing power of money is then used to deflate quasi-money (equation 30) to determine the demand I WU One would also expect prices on parallel markets to be an indication of the extent of aggregate demand pressure, and thus inflationary pressure. Conceptually, this spillover into parallel markets is captured by the black market premium. Along these lines, Feltenstein (1989) outlines a third test for repressed inflation based upon the foreign sector using the linkage between changes in foreign reserves and net domestic credit. This third test is not applied to Algeria in this paper. - 33 * for it and equilibrium in the market for quasi-money (the market for money holdings as a store of value). 58. In levels, the reduced form equation that emerges from this system is the following: log (QM/P) - o + a, log y + a r + a log (M1/C) + a4 log (MI/C).i (31) where: A m a (1+as) (32) A4 -a (33) 59. Estimates of a provide a direct test of repressed inflation. If a is significantly greater than zero, then PT will be higher than P, indicating a significant degree of repressed inflation. First, we ran equation (31) in levels. Next we re-ran it in differences including an error-correction term (ECQM). As before, the error-correction term is given by the residuals from the level regression. The results for the dynamic specification, corrected for - 34 * first-order autocorrelation, are as follows: Alog (QM/P)* - 0.086 + 1.512 Alog y - 0.114 A* + 0.213 Alog (Ml/C) (1.45) (1.32) (-0.13) (0.56) - 0.168 Alog (M1/C).,* + 0.494 ECQM` (-0.43) (2.72) RHO - 0.134 (34) (0.54) a - a3 + a4 - 0.045 (0.07) (Sample 1973-89; R' - 0.389; Durbin-Watson - 1.68) 60. Clearly, a is insignificant and, therefore, the test fails to indicate repressed inflation. The fit of this quasi-money demand specification incorporating a "true" price level is very poor. Moreover, from the sign of ECQM, we see immediately that the equation is unstable. This result indicates that the money overhang terms--as specified- -are not capturing any disequilibrium in the market for quasi-money.3W Repressed Inflation Tests Based on Private Savins 61. This second test for repressed inflation is based on the goods, rather than the money, market. In essence, this test substitutes a private savings function for the quasi-money demand equation above. Otherwise, the basic 3' Use of the ratio of M1 to GDP in equation (34) also failed to indicate a significant degree of repressed inflation. . 35 - idea is the same. An alternative price index P. is defined as a function of the official price index and some measure of the money overhang. PT represents a notional price at which consumers would be willing to consume the potentially restricted amount of goods available at official prices.W If PT is estimated to be significantly higher than the official price P, then repressed inflation will be confirmed. Applying this second test to China, Feltenstein (1987) confirms the presence repressed inflation in that country. 62. We start by estimating a permanent income specification of real private savings behavior, defining real private savings as nominal private savings deflated by the official CPI: log (S/P) - A + A, log I. + A log It (35) In this specifications, S and I represent nominal private savings and real disposable private income, respectively; while I, and I, are the permanent and temporary components, respectively, of real disposable private income. 63. For the purposes of estimation, we take S as the difference between gross national savings and central government savings. Unfortunately, a reliable time series for real private disposable income is not available for Algeria. Therefore, we use real GDP measured in volume terms as a proxy. Permanent private income was estimated from the fitted values of a regression of log (GDP) For further discussion of the analytics behind virtual prices, rationing, and the impact on savings behavior, see pp. 26-30 of Rocha (1988), 2R. =., as well as Feltenstein (1987), g. s=., pp. 1-20, * 36 - vs. time. Temporary private income was defined as the residuals from this regression. The result is as follows: log (S/P) - -10.308 + 1.246 log IP + 1.217 log It (36) (-5.91) (8.54) (2.23) (Sample 1971-89; RI - 0.830; Durbin-Watson - 2.49) 64. While equation (36) fits the data very well, the exceptionally high elasticities of permanent and temporary income corroborate previous indications of forced savings. Permanent income theories assert that (i) the elasticity of real private savings with respect to I, is close to zero, and (ii) the elasticity with respect to I,, while relatively high, is less than one. The very large elasticities estimated above clearly point to a spuriously high correlation in levels between income and savings. The signs of serial correlation in (36) corroborate this mis-specification hypothesis. Therefore, equation (36) provides additional evidence of forced savings and repressed inflation. 65. We now test directly for repressed inflation by substituting PT for P in the private savings equation. Using the previously defined relationship between PT and P (see equation 26), we get: * 37 . log (S/P) - AO + A, log I + A. log I, + A8 log (Mi/C) (37) where: A3 - a (38) Estimating this equation gives: log (S/P) - -2.998 + 0.599 log IP + 0.657 log I, + 1.286 log (Mi/C) (-0.81) (1.83) (1.18) (2.17) (39) (Sample 1971-89; Ra - 0.870; Durbin-Watson - 2.14) 66. The results of equation (39) indicate repressed inflation. The estimate of a is significantly greater than zero, indicating that PT is significantly higher than P, and corroborating the repressed inflation hypothesis. While still high, equation (39) gives much lower income elasticity estimates.UJ In addition, the Durbin-Watson statistic no longer shows any signs of serial correlation. These results confirm that the previous private savings equation (35), based on the official price level (P), was mis-specified. E The fact that real GDP is not a very precise proxy for disposable private income helps explain why equation (39) gives an insignificant estimate for the savings elasticity with respect to temporary income. - 38 - 67. Given an estimate for a of 1.286 from (39), the ratio of virtual to official prices from equation (26) works out to 166% in 1989, indicating a degree of repressed inflation presently on the order of 66%.W This estimate is based on a ratio of MI to C normalized to 1 in 1970, which is equivalent to assuming that the official price level was equal to the virtual price level in that year. * 39 - Ahmed, S. and A. Chhibber, "Non-Inflationary Growth with Debt Management: Indonesia's Achievements and Options", World Bank Mimeo, 1989. Bailey, W., "An Econometric Analysis of Inflation in Algeria: The Monetary Approach", Wor: d Bank Mimeo, August 1990. Chhibber, A., J. Cottani, R. Firuzabadi, and M. Walton, "Inflation, Price Controls, and Fiscal Adjustment in Zimbabwe", PPR Working Paper No. 192., World Bank, 1989. Chhibber, A. and Nemat Shafik, "Inflation, Exchange Unification and Fiscal Adjustment in Ghana", World Bank Mimeo, 1988. A. Gelb and P. Conway, "Algeria: Windfalls in a Socialist Economy", in A. Gelb and Associates, Oil Windfalls: Blessing or Curse?, World Bank, 1988. National Planning Council, "Documentation relative A I'6valuation de 1'dconoale informelle et ses cons6quences sur les 6quilbres macro-6conomLques*, CNP, 1990. Feltenstein, A., "Inflation, Price Controls, and Fiscal Adjustment: A Survey of the Literature and Analytical Framework", University of Kansas Mimeo, 1989. Feltenstein, A., D. Lebow, and S. van Wijnbergen, "Savings, Commodity Market Rationing and the Real Rate of Interest in China", World Bank Mimeo, 1987. Fornasari, F. and M. Leduc, "Algeria: Managing the Budget Constraint during the Transition Period", World Bank Mimeo, 1989. Harvey, A., The Econometric Analysis of Time Series, 2nd Ed., MIT Press, 1990. International Monetary Fund, "International Financial Statistics", IMF, August, 1990. - 40 * Khadr, A., J. Parks, and M. Walton, "Inflation in Malavi", World Bank Mimeo, 1989. Kumar, A., "Measuring the Extent of Price Controls and Measuring the Share of Imports in Consumption*, World Bank Mimeo, 1989. MacDonald, R. and C. Kearney, "On the Specification of Granger-Causality Tests Using the Cointegtation Methodology", Economics Letters, No. 25, North Holland, 1987. Rocha, R., "The Black Market Premium in Algeria", World Bank Mimeo, 1989. Rocha, R., "Goods Market Rationing and Savings in Algeria", World Bank Mimeo, 1988. Standaert, S., "Did Algeria Experience Repressed Inflation in the 1970s?", Journal of Econometric Modelling, January, 1989. NP 5.0 M:\JI\ALIUIN6.DOC