- ;~~~'s \'#l - V 'l; ' Policy Research ( WORKING PAPERS Macroeconomic Adjustmuetni and Growth Country Economics Department The World Bank May 1992 WPS 909 Parallel Markets, the Foreign Exchange Auction, and Exchange Rate Unification in Zambia Janine Aron and Ibrahim A. Elbadawi Zambia's failure with macroeconomic reform - including exchange-rate reform - is the result of macroeconomic (espe- cially fiscal) laxity. And the exchange-rate premium is likely to rise as terms of trade worsen, foreign aid declines, and expec- tations of devaluations rise. Policy Research Working lapers disscminate the findings of work in progress and encourageLhe exchange of idear amnngf3ank sLaff and all others interestec in devclopment issues hbese papers, disnbuted by the Research Advisor, Staff carry thc names of theauthors,refIeci only theurviews, and should be used and ciied accordingl) The findings, inierpretations, andconclusions arc theauthors'own Theyshould not be autnbuteid to the World 13;nk, its lioard of Dirctiors, ILK management, or an) of u.s membcr counLnes Policy Research Macfoeconromi Adjusiment and Growth WPS 909 This paper -a product of the Macroeconomic Adjustment and Growth Division, Country Economics Department- is part of a larger effort in Ihe department to study the macroeconomic implications of multiple exchange markets in developing countries. Copics of the paper arc available frec from the World Bank, 181 8 H Strect NW, Washington DC 20433. Please contact Victoria Barthelimes, room NI 1-025, extension 39175 (May 1992, 115 pages). Since Zambia's independence in 1964, a large, thriving increasing foreign aid Icads to a decline in the premium. parallel market for foreign exchange has coexisted wilh This effect dominates the indirect effects of real wealth a rich menu of official exchange rate policies aimed at and real appreciation, which work to increase the achieving a more flexible, exchange rate and price premium. Expansive fiscp! and monetary policy cause system as well as financial and trade liberalization. the premium to rise. Despite aggressive policies in these areas, particularly Of all the factors that influence the premium and for the exchange rate, Lhe black rnark,t premium have caused exchange rate unification to fail, terms-of- (defined as the ratio of the black market rate to the trade shocks dominate. But Lhc driving force behind official rate) remains high -averaging 10() percent for persistence of the premium was outright laxity in fiscal 1970-88 and more than 400 percent in recent years. and monetary policy - especially ia 1985 and the Aron and Elbadawi examine the origins of Lhe following two years of the exchange rite auctions, and parallel market, the statistical properties of the parallel during the collapse of economic reform in 1987. premium, and the shocks and macroeconomic policy Aron and Elbadawi conclude that exchange rate changes that influence its evolution. Using annual data, reform without fiscal reform may be futile. Fiscal they specify and estimate an eciectic error-correction retrenchment for the first two years of the crawling peg model for the premium. was iifluenccd by political considerations. Zambia is They find that the la?"c parallel market might havc one of the most urbanized countries in Africa: about caused problems in macroeconomic management and half the people live in urban areas, and the urban middle economic reform. A large exchange rate premium as an class wields considerable influence. This explains Lhe indicaton of foreign exchange shortages will have pervasive ensemble of price control and subsidy direct deleterious impact on copper production and schemes that have survived reform attempts. But the export. It can also indirectly hurt copper exports ratio of government revenue to G DP has never been through its negative effect on domnesic incentives for below 20 percent (one of the highest rates in Africa) the officially sanctioned copper economy. A high despite sharp declines in terms of trade. More emphasis premium was also found to encourage overimporting should be given to the political economy and distribu- (and probably overinvoicing) of officially traded tional consequences (especially the rural-urban nexus) imports. in the carly stages of economic reform. Aron and Elbadawi find that foreign inflation and Zambia's economy, heavily dependent on copper depreciation of the black market rate (in a cost-push exports, is particularly susceptibic to external shocks. It manner) directly increase domestic inflation. Deprecia- is important to liberalize major trade and financial lion of the black market rate also signals indirectly Lhat markets in such a way as to compress the parallel economic reform lacks credibility and that macroeco- market and prevent the premium from serving as a nomic policy is unsustainable. Short-term changes in major signal to Lhc economy. the premium reflect expected changes in policy and Increased foreign aid could help mitigate the politics. destabilizing effcts of terms-of-trade shocks. This is The major factor behind the failure of unification likcly to be most helpful in the carly stages of refoim, and economic reform is the fundamental endogeniety of when it can foster credibility and stabilize the free ratc the parallel premium in macroeconomic and tradc - before the forcign sector can begin to rcspond to real policy - as well as in exogenous terms-of-trade and depreciation. Then there will be less need for aid flows foreign aid shocks. Improving the terms of trade or to conLinue at their initial levels. The Policy Research Working llaper Seriesdisserninates the findings of w\ork under \k ay in ilie Bank. Anobjectiveofthe series is to geL these findings out quickly, even if presentations are less than fully polished. Thc findings. inlerprelalions, and Xconclusions in thcsc papers do nct necessarily represent official Bank policy. Produced by the Policy Research D)issernination Center CONTENTS I. INTRODUCTION . . . . . ... . . . . 1 II. THE DEVELOPMENT OF MULTIPLE MARKETS FOR FOREIGN EXCHANGE IN 2AMIA . . . . . . . o--oo. . *t. *. . . . . . ....... 6 1. A brief survey of official currency arrangements since 1964 . . 6 2. The emergence of an unofficial market in foreign exchange. . . . 9 3. An evaluation of exchange rate policy in Zambia. . . . . . . . 10 III. THE PARALLEL PREMIUM AND THE MACRO-ECONOMY IN ZAMBIA: 1965-1989 . . 16 1. Macroeconomic linkages by exchange rate episode . . . . . . . . 18 2. A Model of Expected Devaluation. ..... .... . . . 28 IV. A MODEL OF THE PARALLEL PREMIUM ..*...... .......... 32 1. The Basic Model . . . . . . . . . . . . . . . . . . * . . . . 33 2. Endogenizing the RER . . . . . . . . . . . . . .... . . . . 38 V. EMPIRICAL DETERMINATION OF THE PREMIUM . . . . . . . . . . . . . . 43 1. The Annual Model . . . . . . . . . . . . . . . . . . . . . . 43 VI. THE EFFECTS OF THE PREMIUM ON TRADE . . . . . . . . . . . . . . . . 47 1. Reported Trade Flows . . . . . . . . . . . . 47 2. The misinvoicing of foreign trade. . . .. . 55 2.1 Statistical methods for the detection of unrecorded trade. . 58 2.2 An approximate quantification of misinvo$icing in Zambia. . . 59 VII. THE EFFECTS OF THE PREMIUM ON DOMESTIC INFLATION AND FISCAL BALANCE 65 1. Domestic inflation. . . . . . . . . . . . . . . . . . . . . . . 65 2. The fiscal balance. . . . . . . . . . . . . . . .. . . 70 VIII. THE PROBLEMS OF UNIFICATION: SOME CONCLUDING REMARKS ON THE LESSONS FROM ZRENIA . . . . . . ....... ..574 REFERENCES .. .. .. .. .. .. .. .. .. .. .. .. .. ..... .................................... 105 LIST OF TABLES Table 1 (a) Exchange Rate Policy Episodes in Zambia 1964-90 . . . . . . . 79 Table 1 (b) Basic Statistics for the Black Market Premium (monthly data) . . . . . . . . . . . . . . . . . . . . 80 Table 2 (a) Major Parity Changes Since 1964 . . . . . . . . . . . . . . . 81 Table 3 (a) The Parallel Premium and the Macro-economy . . . . . . . . . . 82 Table 3 (b) Zambian Copper Dependence, 1964-89 . . . . . . . . . . . . . . 83 Table 3 (c) Short-run Influences on the Black Market Premium During the Auction Period . ................. . 84 Table 6 (a) Expected Disparities for Import Data Comparisons in the Presence of Trade Policy and a Black Market Premium . . . . . 85 Table6 (b) Partner-country-data Comparisons: Percentage Misinvoicing . . 86 Table 6(c) Quantifying Faked Invoicing Using Import Unit Value Comparisons (1975=100) ................... . 87 Table 6 (d) Statistics for the smuggling incentive and Trade Ratios . . . 88 Table 6 (e) Regressions of the Trade Ratios on the Smuggling Incentives . 88 Table 7 (a) An Anatomy of Possible r 'cal Implications of Exchange-rate Devaluation-driven Unification . . . 89 Table 7 (b) Central Bank Losses Due to Foreign Exchange Transactions . . 90 Table 7(c) An Empirical Assessment of the Location of Unification Episodes on the Seigniorage Laffer Curve . . . . . . . . . . . 91 Appendix Tables Table A.2 (a) A Summary of the Exchange Rate Components of I.M.F. Programmes with Zambia from 1976 . ...... . 108 Table A.3 (a) A Summary of the Weekly Auction Results . . . . . . . . 109 Table A.5 (a) The Parallel Premium and its determinantst Tests for Normality (Annual Data) . . . ... .. . 110 Table A.5 (b) Dickey Fuller and Augmented DLikey Fuller Tests for Unit Roots . . . . . . . . . a . . . . . . . . . . .. 110 Table A.5 (c) The Black Market Premium and Related Macroeconomic Variables in Zambia ... . . . . . . . . . . . . . . 111 LIST OF FIGURES Figure 1 The Quarterly Black Market Pzemium for Zambia showing exchange rate episodes . . . . . . . . . . . . . . . 92 Figure 2 (a) Nominal and Real Effective Exchange rates by Episode (1980-100 for the official rate) . . . . . . . . . . 93 Figure 3 (a) The Monthly Black Exchange Rate Premium: 1970-90 . . . . . . 94 Figure 3 (b) World Copper Prices: LME Prices in U.S. c/lb (Mt Wdeflated, 1963 terms) . . . . . . . . . . . . . . . . . 95 Figure 3 (c) Expectations of a maxi-devaluation: (monthly black market premium Jan. 1982-Jul..1983) . . . . . . . . . . . . . 96 Figure 3 (d) Zambia: nominal official and parallel market exchange rate depreciation: 1984-89 . . . . . . . . . . . . 97 Figure 3 (e) Expected devaluation estimated using monthly data . . . . . 98 Figure 6 (a) A proxy for trade policy in Zambia: ratio of imported to retail manufacturing price indices (1966-100) . . . . . . 99 Figure 6 (b) Smuggling incentives and trade data ratios . . . . . . . . 100 Figure 7 (a) Inflation (CPI, WP) and parallel rate depreciation . . . . 101 Figure 7 (b) Inflation (disaggregated index) and parallel rate depreciation . . . . . . . . . . . . . . . . . . . . . . . 102 Figure 7 (c) The Seigniorage Laffer Curve and the premium/inflation trade-off . . . . . . . . . . . . . . . . . . . . . . . . 103 Figure 7 (d) Inflation and exchange rate depreciation during the auction ..... . . . . ..... . . . . . . . . . 104 Appendix Graphs Figure A-1 Black Market Premiu ..112 Figure A-2 Real Exchange Rate ..113 Figure A-3 Real Money Supply . . . . . ..114 Figure A-4 Interest Rate Parity Differential . . . . . . . . . . . . 115 This paper is pai. of the CECMG project 'Macroeconomic Implications of Multiple Exchange Markets in Developing Countries". The paper has benefitted from discussions with and comments from Stephen O'Connell, Paul Collier, and participants at a CECMG workshop on the project. Also country desk provided helpful comments that have been received and taken into consideration. Any views expressed in this paper are not necessarily those of the World Bank or affiliated organization. The authors would like to acknowledge able research assistance from Nita Ghei. x *iNRODUCTION The economy of Zambia provides an example of a thriving and large black X market for foreign exchange. This black market ha. co-existed with a rich menu of official policies aimed at achieving more flexible exchange rats and price system as well as financial and trade liberalization. Despite aggressive policies in these areas especially with regard to the exchange rate', the black market premium (defined as the ratio of the black rate to the official rate minus one) remains high. For the 1970-88 period, the premium averaged 100 percent and exceeded 400 per cent in recent years. The presence of sizable parallel markets with substantial premiums, such as the one in Zambia, has Important macroeconomic implications. The theoretical models of dual exchange rates, e.g. Lisondo (1987 a,b) and Dornbusch (1986), and their extensiona and elaboration to the case of black markets, e.g. Dornbusch et al (1983) and Pinto (1987, 88a, 88b), derive a steady state solution for the premium that depends on the fiscal deficit ratio and the parameters of trade and exchange rate policy. Furthermore, key macroeconomic indicators such as inflation, foreign trade flows, and the real exchange rate can also be shown to be influenced by (or jointly determined with) the premium, e.g. Kamin and Elbadawi (1990), Elbadawi (1989). The empirical tradition on this literature however, is still in its infancy. One of the main objectives of this paper is to develop empirical models linking the premium to the above macroeconomic indicators and to use Zambian data to assess the extent and dynamics of these linkages. 'Since the country's ndependence in 1964 to present, six distinct episodes of exchange rate regimes - including an auction - can be identified (see Table 1 (a)). -2- Evidence from the above models can shed light on the recent experience in Zambia of aborted reforms, macroeconomic instability and recurrent policy reversals.2 The failure of the Zambian economic reform of the last decade, especially those aspects of the reform related to the foreign exchange auction experiment and the attempts at exchange rate unification, has been suggested to be the direct outcome of macroeconomic laxity, particularly with regard to fiscal policy, (Pinto, 1987). The evidence of our paper upholds this view. In addition, exogenous factors that operate through the trade accounts such as the terms of trade and foreign aid, and expectations reflecting asset market considerations, are important as potential determinants of the premium in Zambia. The empirical results of this paper predict a rising premium as a result of worsening terms of trade, declining foreign aid, or higher expectations of future devaluation. Official dual markets have twice been cr-ated in Zambia: a two tier auction operated for two months in 1987, and in February, 1990, a dual exchange rate system was established. These episodes are either too short-lived or recently instituted to merit analysis. Thus, the structure of multiple markets for foreign exchanle considered in this study will consist of an official market with an officially managed or auction-determined rate. and an illegal unofficial market with a freely floating rate. During the auction period, a quasi-legalization of the unofficial market occurred with the introduction of "own funds" import licenses. The quarterly black market premium for Zambia is shown in Figure 1 where the considerable regime shifts of the last two decades are clearly reflected in the evolution of the premium. 21n fact Pinto and van Wijnbergen (1987) argued that the fail- e to adjust policy for the existence of par&llel markets may be the main cause behind recent macroeconomic instability in SSA. - 3 - A statistical analys's of the premium and black rate capturing broad changes in size and variability across exchange rate episodes is given in Table 1 (b). Black market rates are available from mid-1970 which omits the period of the copper boom (1964-74). It is likely that the premium from 1964 to 1971 would reflect incentives for capital flight due to political uncertainty post Independence, the presence of exchange controls and an increased demand for foreign assets during the copper boom. The premium rose after the negative copper price shocks of 1971 and 1974, thereafter following an upward trend in tandem with the incroasingly over-valued fixed exchange rate. In the controlled float episode from July, 1976, a continuous dcvnward trend in the premium ensued which was maintained under the crawling peg, after 1983. A sharp rise for 1985 was due to the break with I.M.F. late in 1984, but the auction in its early stages saw a low premium. The premium fluctuated considerably during the auction, but the trend was to higher values after August, 1986. Across exchange rate episodes the average premium has declined, though rose by a factor of ten in the final episode. The black market rate has continuously depreciated across the five episodes. Both the premium and the black market rate exhibit decreasing stability across exchange rate episodes (though variability declined during the auction). In part this may be related to bubbles and crashes reflecting agents' reactions to accumulated mismanagements and to speculation. Speculative aspects and the distribution of the premium are analyzed in Section V. Three key aspects will be examined in this paper. First, the origin of the parallel market and the statistical properties of the premium are analyzed. The factors influencing the premium in various episodes are -4- considered and the premium is stimated using annual data. Secondly, the effects of the premium on inflation and on recorded avd unrecorded (illegal) trade fli w are considered. There in a paucity of empirical research on the role of smuggling and misinvoicing of exports and imports in response to the premium and trade policy. An attempt is made here to quantify these responses and to link them to changes in policy and the premiun. A premium on foreign exchange serves as a tax on exports and a subsidy on imports. The fiscal impact of the premium in Zambia is important in the uniflcation of official and unofficial exchange rates, where a lose of revenue implicit in the demise of the unofficial market may have induzed inflation tax to be substituted for it in the early stages of the auction (Pinto, op cit). Finally, Zambia has attempted unification of official and unofficial rates of exchange in two distinct episodes: first, by pre-announced monthly devaluations in the crawling peg episode (1983-85), and secondly, using a market defermined rate in the foreign exchange auction (1985-87). What distinguishes the two episodes is the timing of unification, sequencing of accompanying measures of liberalization and portrayal of commitment to liberalization (i.e. credibility aspects). The episodes may usefully be contrasted and lessons drawn from the aborted auction episode. The paper is organized as follows. A detailed narrative tracing the development, structure and operation of multiple foreign exchange markets is contained in Section II. In Sec*ton III the link between the premium and the macro-economy is examined, and in Sections IV and V, a model for the premium is specified and estimated. Section VI analyses the effects of the premium on recorded and unrecorded trade flows using econometric models for key foreign trade aggregates, and In the case of unrecorded imports, data was constructed from estimates of overinvoicing ratios computed in this section. Yn Sectlon VII, the impact of the premium on Lnflation was eatimated using a simple, empirical inflation model$ also, the fiscal lmpact of the premium, *epecial'y exchange rats unifLcation a la Plnto (op. cit.), is estimated and analyzed. Finally, Section VIll concludes. -6- 1. *- DEVELOPMNT OF MULTIPLE MAURKT FOR FOREIGN EXCHANGE SN ZAMEZA 1. A brief survey of official currency arrangmeants Erom 1964. Parity changes since 1964 are disc_ajed with reference to Table 2 (a). At Independence in 1964, Zambia's currency was pegged to the pound rnd was fully convertible. The kwacha replaced the Zambian pound on 16th January, _968, halving the official rate (to KO.714/$ or Xi.7/ST) whilst retaining the link to the pound sterling. From 1964 until the early 19703 this fixed exchange rate was ataintained. On 3rd December, 1971, the kwacha's ties to the pound were severed and it was linked to the U.S. dollar at the rate KO.714/$. This represented a devaluation, given the kwacha's appreciation againat the dollar following a de facto devaluation of the American uni, on 15th August, 1971. The gold content of the kwacha was reduced to 7.89 per cent on 22nd December, which paralleled the U.S. dollar's devaluation of 18th Decemi-z, thus leaving the official rate of exchange unchanged. The kwacha did not follow the February, 1973 devaluation of the U.S. dollar, and was realigned to KO.643/$ on 15th February, based on an unchanged gold content. Thereafter, with the increase in the dollar's value in consequence of floating and the devaluation of various currencies, the kwacha appreciated against the currencies of all major trading partners (save the Deutsche mark). By mid-1976, the value of the nominal effective exchange rate (trade-weighted) had risen over 25 per cent. On 8th July, 1976, ties with the dollar were severed and the kwacha linked to the SDR at SDR1.08479 on a controlled, floating basis, a de facto devaluation of 20 per cent. This rate was devalued by 10 per cent on 17th MArch, 1978 to SDRO.976311 and by a further 20 per cent on 7th January, to SDRO.78125. The link to the SDR was broken on 6th July, -7- 1983: a crawling peg based on a basket of currencies of maior trading partners, rather than the basket determining the value of the SDR, was introduced. A controlled devaluation was to be achieved of 1 per cent per monvh; the rate was increased to 2.5 per cent in 184. Dissatiefaction with the downw_rd rate of ad4untment and the inefficiencies of the accompanying manual exchange allocation system led to the adoption of a foreign exchange auction system in October, 1985. The rate to the dollar on 3rd October was K2.2, reaching KS.01 in the first of the weekly auctions, on 11th October, and K8.30 a year later on 11th October, 1986 in the 53rd auction. Thereafter the rate declined sharply reaching K15.25 in the 60th auction on 29th November, 1986 (an 86 per cent depreciation of the kwacha from the inception of the system). From the 41st auction, new documentation was required to accompany bids, and from the 43rd auction (2nd August, 1986) a "Dutch Auction" replaced the former system.3 The Central Bank increased the amount of foreign exchange for auction three-fold from 2nd August, but this policy was not oustainable. These modification. failed to arrest the kwacha rate of depreciation. Following riots on the Copperbelt and in Lusaka on 5th December, 1986, with the rate at K14.92 in the 68th auction, the auction was temporarily suspended (28th January, 1987). The rate was set at K9.00 . Thereafter six auctions were held, beginning on the 28th March, 1987, under a two-tier auction system: an official window at !.9.00 was restricted to debt service, essential imports and receipts of loans and grants, whilst the auction rate was allowed to fluctuate between this rate and an upper ceiling of K15.00 3In the "Dutch Auction", successful bidde..s paid the full amount of their bid regardless of the excess over the marginal bid, and this sum was surrendered as government revenue. (this restriction was abandoned after four weeks). In the fifth of theme auctions on the 24th April, the kwacha rate fell to its lowest value, K21.02; President Kaunda abolished the system on 1st May, after a final auction which served only to allocate foreign exchange at the rate of K15.00. From 5th May, a fixed rate of K8.00 was installed. Requests for foreign exchange were filed through the commercial banks, and allocation of foreign exchange and isbue of import licenses decided simultaneously by a Foreign Exchange Management Committee (FEMAC). Non-traditional exporters were allowed to retain 50 percent of their export earnings. Free trade of title to the reta'ned foreign exchange occurred at a premium on the official rate, but required prior approval from FEMAC. A third, legal foreign exchange market was limited to larger non-traditional exporters and allowed the use of export earnings to purchase Zambia's commercial (pipeline) debt at a discount on international financial markets, then exchanged at the Bank of Zambia .'or kwacha equal to the official rate times the face value of the debt inatrument. A devaluation to X16.00 occurred in June, 1989, and a further devaluation to K24.00 in December of that year. As of February, 1990, a two-tier currency was re-introduced: an official rate of exchange was initiated at K27.80/$, with the "market exchange rate" at K40.00/S. The Bank of Zambia sold foreign exchange through the First Window (the official rate) under existing FEMAC procedures for imports, and purchased all foreign exchange earned by ZCCM. The Second Window operated at the market rate and was used for those goods eligible for foreign exchange allocation under the Open General License System (OGL). The sources to fund this window were non-traditional exports earnings and remittances received from donors. only 10 per cent of Imports by volurtie were eligible for OGL allocation at -9- the inception of the system. By the 31st March, 1991, most imports had been transferred from the First to the Second Window, due to the gradual expansion of the OGL to cover over 92 per cent of imports. Donor funding accounted for 54 per cent of the funding of the Second Window in 1990 and all of it in 1991. In April, 1991, the rates were unified at K58/$. 2. The emergence of an unofficial market in foreign ex-hange. The origins of Zambia's unofficial market in foreign exchange lie in the control regime introduced and extended from Independence in 1964. Political uncertainty in the decade of nationalizations post 1964 and deterioration of the economy after the copper price shocks in the early 1970c provided further incentives for capital flight. At the dissolution of the Federation of Northern Rhodesia, Southern Rhodesia and Nyasaland (1953-63), Zambia had been the only country of the three to relax exchange control restrictions imposed in 1961; however, these liberalization measures were minor. After a negative copper price shock in 1971, foreign exchange reserves were protected by tightening exchange controls. The overvalued exchange rate and foreign exchange controls created excess demand in the market for foreign exchange, and thu parallel market burgeoned. Import licensing, originally devised to avoid congestion after the closure of transport routes in the 19606, increasingly served to control import levels. In 1972, a restrictive import licensing system was introduced. Certain categories of imports were banned, but capital and intermediate goods imports were liberally licensed. Import licensing was further tightened early in 1975 following the precipitous decline in copper prices in 1974. These - 10 - measures continued to be strictly enforced until the liberalizing reforms of the early 1980s. Price controls had been applied from Independence to producer prices of agricultural goods, the prices of "essential commodities" and some parastatal company products. Interest rate controls rendered real interest rates negative from 1966 onwards, reaching below -10 percent between 1976 and 1980, and just preceding and following the auction. Evasion of price and trade controls by smuggling and misinvoicing, and consequences for the size of the parallel market and the premium are discussed in the section on unrecorded trade (Section VI.2). Various liberalization measures were implemented preceding the foreign exchange auction and during its operation. Most price controls were lifted in December, 1982 and interest rates decontrolled late in 1985. The trade account was liberalized in October, 1985 concomitant with the floating of the exchange rate. External capital controls remained in place throughout the nineteen month auction period. Quasi-legalization of the black market was also effected just prior to the auction with the introduction of export retention schemes and "own funds" import licenses.' At its termination most of the liberalizing measures were reversed. The impact of the liberalizing reforms on the premium is considered in Sections III and VI, and quantified in Section VII.2. 3. An evaluation of exchange rate policy in Zambia. Six distinct exchange rate episodes have been identified for Zambia since Independence in 1964 (see Table 1 (a)). Exchange rate policy is analyzed with ' Remittance of foreign exchange held abroad, or payment for imports of goods and services out of funds held abroad, was allowed without restriction or declaration as to the source of the funds. - 11 - referencs to Figure 2 (a) where nominal and real effective exchange rates for the official rate are contrasted with the parallel real effective exchange rate.5 The official REER and NEER are set to 100 in 1980 while the parallel REER is 60 in 198G to reflect the actual premium. The premium is the distance between the two REER indices divided by the height of the parallel REER index. As they move closer together, so the premium declines. The lower the parallel REER and the higher the official REER indices, the higher the premium. The parallel rate is expected to display more short-term volatility (this is discussed in Section III), but discounting for the risk of illegality, may serve as a proxy for a market rate. Thus the effective rates in Figure 2 (a) may be used both descriptively to contrast the official REER and NEER, and to assess the misalignment of the official REER from a "market rate" (parallel REER). In episode 1 (1964-1976Q2), the exchange rate was fixed, initially to the posind sterling, and subsequently to the dollar. Exchange rate policy was inactive in this period. From 1971 the nominal rate became increasingly overvalued, culminating in severance from the appreciating dollar. The appreciation contributed to the deterioration in trade and current account balances following the severe negative shock to copper prices in 1974 and rise in oil prices after 1973. At the same time wage costs were rising and mining 5 The real effective exchange rate (REER) La the geometric average of trade-weighted real exchange rates (RERi) ior the six major trading partners, excluding South Africa, where RERi=NERi.CPI(Zambia)/CPIi , and NERi is the nominal bilateral exchange rate in foreign currency per kwacha. The weights, Wi, are import weights, based on partner-country recorded exports, to avoid the misinvoicing bias (South Africa does not record exports to Zambia). Weights were redefined for 1968-72, 1973-77, 1978-82 and 1983-88. The CPI rather than WPI was used as there are less likely to be definitional discrepancies across countries (IFS, I.M.F.). - 12 - profitability falling with increased production costs and transport dLfficultis. The overvalued rate further encouraged capital intensive production methods towards which a bias already existed given the nature of tariff policy. Little variation is apparent in the official REER (inflation averaged 6 per cent for 1970-73) save for a muted response to the shocks of 1971 and 1974. The parallel REER is considerably devalued relative to the official REER and tracks the terms of trade shocks closely. Progressive devaluations of the kwacha from 1976 until the auction of 1985, had their origins in conditionality associated with International Monetary Fund credit agreements with the government. Their aim was to redress external imbalance in the Zambian economy by enhancing the competitiveness and profitability of the export sector and encouraging the development of new exports and import substitutes. The IMF's involvement with economic policy management in Zambia began in 1971 with the negotiation of a compensatory financing facility for SDRl9 million, following a negative copper price shock and the flooding of Zambia's second largest min. Financial flows between Zambia and the IMF have since been considerable. The first stand-by agreement was negotiated in 1973 covering one year and was accompanied by mild conditions. The post-1975 stabilization programs with the IMF in 1976, 1978, 1981, 1983 and 1984, with the exception of the 1976 program, all involved higher conditionality (Ndulo and Sakala, 1986).7 In episode 2 (1976Q3-1983Q3) the rate was pegged to the SDR and two 6The Rhodesian border with Zambia was closed in 1973 while civil war in Angola rendered the Benguela Railway route unusable. 7 The exchange rate components of the agreements are summarized in Appendix A. - 13 - devaluations enacted, of 10 per cent in March, 1978 and a maxi-devaluation in January, 1983 of 20 per cent. The NEER fell steadily until 1980, in tandem with the official REER (inflation from 1976-80 averages 15 per cent, but reached 10 per cent by 1980). The parallel REER appreciated in 1979 reflecting the more realistic official REER and a positive terms of trade shock in 1979/80. In the aftermath of the 1979/80 copper boom, inflation rose (it averaged 13 percent, 1980-83) and nominal rate depreciation was insufficient to prevent misalignment of the official REER and NEER. The premium narrowed after 1979 as a function of devaluation, terms of trade changes and increased foreign borrowing, reaching 30 percent in 1983. The large devaluation of January, 1983 heralded a new exchange rate episode, episode 3 (1983Q4-1985Q3), where from mid-1983 the SDR link was substituted by a basket of currencies of major trading partners and a crawling peg employed. Given high levels of inflation (averaging 26 percent from 1983- 85), the official REER continued to be overvalued, but the premium narrowed reaching 40 percent in the month preceding the auction. During 1984 and 1985 the importance of role of the exchange rate as an economic policy instrument to induce the required structural adjustment was increasingly emphasized, culminating in the foreign exchange auction (episode 4 (1985Q4-1987Q2)). The stated objectives of the foreLgn exchange system and its accompanying measures (Bank of Zambia, 1985) were to ensure the responsiveness of the rate to changes in the demand for and supply of foreign exchange and to provide foreign exchange in a timely manner. Foreign exchange would be attracted into the banking system while the parallel market would have reduced importance. Subjectivity and the reliance on administrative mechanisms in the allocation of foreign exchange would be eliminated. A larger - 14 - volume and broader range of exports was expected to result. Further, foreign investment would be encouraged by facllitatic if the remittance of dividends and profits. The principal feature of the weekly foreign exchange auction was that the auction rate applied to all foreign exchange transactions. Predetermined amounts of foreign exchange were allocated outside the systeme but at the auction rate, for the governmentls imports, ZCCM and non-traditional exporters' export earnings retention quotas, crude oil imports and related port charges and IATA payments by Zambia Airways.' Foreign exchange for all other imports and service payments, including the remittance of profits, dividends and transfers were obtained through the auction. The source of funds for the auction was foreign exchange from the commercial banks arising from exports of goods and services and external loans and grants. Wide support was received from the international community, especially the World Bank, the IMF and bilateral donors. Both the parallel and official REER fell during the auction, and the gap between them narrowed.9 Thus the premium fell to about 30 percent in 1986, a level considered compatible with the maintenance of exchange controls. Nevertheless, high levels of inflation (see Section 7 on the Pinto Effect) induced overvaluation in the official REER. The reversion to a fixed rate with periodic devaluations for over two years after the auction constituted episode S (1987Q3-1989Q4) with persistently high inflation, the official REER was progressively more s Petroleum and IATA payments were included in the auction from February, 1986. 9 This is seen more clearly with monthly data (discussed in Section 7). - 15 - overvalued and the real depreciation gains of the auction episode were eroded. Meanwhile the premium soared, exceeding 400 percent in 1988. This regime was superimposed by an official multiple rate *ystem from early 1990, which was unified in the first quarter of 1991. - 1r. - III. TEE PARALLtL PR36U1 AND TEE NACRO-ECONOWY XN ZAMBIA: 1965-1989 The premium and the etock of black market currency are generally considered to be jointly determined by the interaction of stock and flow conditions in the black market for foreign exchange (Dornbusch et al, 1983). Portfolio balance considerations, given expected yields on domestically and foreign-held assets, will govern the stock of assets willingly held; that is, there is a level of the premium that w!'l establish stock equilibrium. Current and expected changes in econom - jolicy or changes in exogenous variables, such as terms of trade and foreign aid, will affect the portfolio composition through the wealth effect. Flow considerations are especially emphasized in black market models: the level of the premium will influence the flows into and out of foreign currency by altering profit opportunities for smuggling, the faked invoicing of trade, tourism and the remittances of migrants. Factors affecting flow conditions, and therefore the spread, include trade controls and the risks and costs of engaging in illicit activity. In the Introduction, the considerable changes seen in the annual black market premium were briefly discussed with reference to Figure 1. Statistics for the premium by exchange rate episode (Table 1 (b)) showed a decline in the mean premium until episode 5, while variability tended to increase with the exception of the auction perlod. Examination of monthly data organized by exchange rate episode in Figure 3 (a) shows that some periods have experienced more volatility than others. The increased variability reflects the short-run impact of "news" on the premium. The black market is integrated with forward looking asset markets so that the assumption of rational expectations in conjunction with portfolio - 17 - decisions means that expectations about political events and changes in economic policy may be apparent in the premium before they occur, while political and economic shocks have an immediate impact. Whether this variation in the premium will be positive or negative depends upon the change induced or expected to be induced in the private sectors' desired holdings of foreign exchange. Dornbusch et al (1983) find expectations of major devaluations, in addition to variables capturing changes in asset stocks and the flow of 2oreign exchange, to be significant in explaining the Brazilian black market premium, while politics and interest rates were major determinants of the Argentinian premium (Dornbusch and Moura Silva, 1987). In the short-run, portfolio considerations and expectations are likely to be the major determinants of the premium. The long-run steady state premium has been shown in models employing the above theoretical framework to depend on such fundamentals ae the terms of trade, the budget deficit, capital flows and the trade regime (see for example Lizondo (1987b), Pinto (1988a) and Section V of this paper). In what follows, first, the Zambian experience is considered in the light of the above cited literature. Linkages between the premium and the macro-economy are examined for each exchange rate episode, with recourse to macroeconomic indicators in Table 3 (a) and monthly premium trends in Figure 3 (a). Secondly, we specify and estimate a simple monthly model of expectation of future devaluation, drawing on the discussion of episodes. This model accommodates the reactions to anticipated economic policy as well as to political and economic events, and can be used to study the evidence on the determinants of expectations in Zambia. Furthermore, if expectations of - 18 - future devaluations are important for the determination of the premium and our model of expectation is correctly specified, the derived series of expected devaluation should closely approximate the actual behavior of the premium. 1 * acroeconoamic linkag e by excange rate opisode. 1.1 Eoiggde I: 1964-197602 Zambia attained independence from Britain in 1964, the first year of the copper boom, at the dissolution of the Wederation of Northern Rhodesia, Southern Rhodesia and Nyasaland (1953-63). A very small industrial base was inherited, with manufacturing contributing les than 7 per cent of GDP and 10 per cent of formal sector employment in 1964. Agriculture contributed 12 per cent of GDP in 1964, due largely to the output of the 1200-1300 expatriate, commercial farms along the line of rail to the copper belt, most agriculture being rural and subsistence-based. Dominant in the economy from the 1920s was the copper sector, owned and managed by foreign interests and providing the major impetus for the development and settlement of the economy. At Independence copper contributed 40 per cent of GDP, 20 per cent of formal sector employment, 90 per cent of goods exports and over 60 per cent of tax revenues. Zambia's early economic development was characterized by an increasing role for government in the economy and the imposition of economic controls. The origin of these controls lies in the practical implementation of the Zambian philosophy of "Humanium", a version of African Socialism developed in the pre- Independence political struggle by Kaunda and the United National Independence Party (articulated ln Kaunda (1974)). It was considered a political necessity to address pre-Independence health and education (racial) - 19 - differentials and to provide housing, to "Zambianise" employment, and gain from expatriates economic :ontrol of agriculture, mining and industry, creating Zambian entrepreneurs. Diversification of the economy away from copper was accorded priority, a long-held view, the first allusione to which were in the 1930. (King (1987)). Agricultural and industrial development were therefore emphasized. The 1968 Mulungushi Reforms expanded the government holding company for industry, finance and real estate by means of a 51 per cent state participation in 26 existing private enterprises, and it grew rapidly thereafter due to further take-overs and new investment. In 1976, the public share in manufacturing was over 50 per cent. Nationalization of the two major mining companies (a 51 per cent equity share) followed in 1969 with the compensatory issue of 6 per cent bonds, unconditionally guaranteed by the government and serviced out of dividends. The financial sector proved unamenable to nationalization, but various state financial intermediaries have since been created. State involvement in agriculture was via State farms and subsidized statutory monopolies dealing with the largest marketed trade. Substantial subsidies and loans were extended to parastatAls while food subsidies guaranteed a quiescent urban population (politically Important, comprising 50 per cent of the population). Economic controls were introduced and steadily extended post Independence, particularly after the first negative copper price shocks of the early 19709. These included interest rate controls, wage controls, price controls in agriculture and industry, exchange controls and quantitative restrictions on trade. The pattern of copper prices from 1950 to 1990 is shown in Figure 3 (b). A positive shock to copper prices from 1964 induced a trade surplus. Foreign - 20 - reserves were accumulated reaching a maximum in 1970 while capital inflow was modest. Government savings were positive until 1970 and amounted to about a third of national savings. Public sector capital and current expenditures rose strongly in line with the ambitious First National Development Plan (1964-71), but the budgetary position was sustainable until 1970, given high mineral revenues. From 1965-76 real income grew at an annual average rate of 2.4 per cent. The aggregate figure reflects a decline in mining growth over the period due to rising costs and transport difficulties (-2 per cent), and slow growth in transport and communications and construction ((1.8 per cent and 1.4 per cent, respectively). Although emphasis had been given to agricultural development, output growth was low with continued reliance on Imported food and agricultural raw materials. The domestic terms of trade were against agriculture with producer prices set below border prices. Services expanded rapidly at 6.7 per cent with government demand while the manufacturing sector achieved the highest rate of growth from a low base at 7.7 per cent. However, wage increases, low interest rates, generous investment and depreciation allowances, low tariffs on capital goods and an over-valued exchange rate encouraged an import-substituting, capital intensive mode of production. By 1976, the manufacturing sector received almoot 40 per cent of visible imports of which about 15 per cent were capital imports. Thus despite an increasing government role and stated commitment tc diversification, the negative copper price shock in 1974 impinged upon an economy still highly dependent upon copper (see Table 3 (b)). The first negative shocks of the early 1970s had induced balance of payments and budget - 21 - crises. The copp r price recovered from 1973, but after the crash Ln mid-1974 persistent budget and trade deficits were the norm. Theme were sustained by decumulation of reserves, borrowing arl the appreciation of the currency. Zambia's link with the I.M.F. was initiated with extended facility agreements in 1971-72 followed by a standby agreement in 1973. Zambian borrowing appeared to be pro-cyclical with nominal copper pricess borrowing was heavy in the 1970-75 period while Zambia wac rationed by creditors during 1976-77. The imposition of more stringent in-port and exchange controls was discussed in Section 11.2. Black market rates are available from mid-1970 (averages of monthly rates). The trends for misinvoicing in Tables 6 (b) and (c), however, suggest considerable over-invoicing from 1964. The premium from 1964-70 is likely to reflect this increased de'uand for foreign assets during the copper boom as well as the effects of stringent exchange controls and political uncertainty in the first decade after Independence. The flow of illegal foreign exchange is responsive to the terms of trade, inducing a rise in the premium when the terms of trade deteriorate. In 1971 following a negative terms of trade shock the level of the premium was over 50 per cont. Thereafter it responded appreciably to changes in the terms of trade, falling with the 1972-74 improvement and rising after the 1974 crash. From 1973 the exchange rate fixed to an appreciating dollar became increasingly over-valued which exacerbated the premiumse sharp rise to a maximum in 1977, only exceeded in the late 1980s after the auction. - 22 - 1.2 Eoisode 2: 197603-198303 In the aftermath of the precipitous decline in copper prices from 1974, the government entered into several conditional credit agroements with the I.M.F. In line with the 1976-77 I.M.F. stand-by facility the tax base was altered, wages were frozen, subsidies cut by 30 per cent and development expenditure decreased by a third. The budget deficit was reduced but three-quarters of it was funded domestically, which added to inflationary pressures, as evidenced in Table 3 (a). With a highly restrictive import policy and a small rise in the price of copper the current account balance showed some improvement. Thereafter the pattern of large budget and trade deficits continued, mitigated only by the rise in copper prices from 1979. The 1978-80 I.M.F. stand-by facility reduced credit and inflationary pressures, but crowded out the private sector, most credit being absorbed by the government and mining companies. From 1975, there was little cap.tal inflow, but payments arrears built up (by 1980 amounting to $600 million) and net foreign assets declined. Foreign reserves had by thie time largely been eroded. The foreign exchange constraints forced a contraction in the manufacturing sector, services declined with government demand and agriculture continued to stagnate. Although a 1981-83 extended facillty was successful in reducing the budget deficit in 1981, it became inoperative in 1982, in which year the attempt to maintain government services and employment in the face of falling revenue induced a deficit of 18 per cent of GDP. The payments deficlt was financed mainly by the accumulation of arrears. The black market premium exhibits two phases of behavior within this episode: in the first, until mid-1979, the level remains on average above 150 percent; In the second, the premium falls by 70 percent and thereafter - 23 - gradually declines, with decreased volatility. The first phase is characterized by sharp rises in inflation and large budget deficits. Zambia was rationed by foreign creditors, reserves were eroded and the terms of trade continued to decline. Tight import and foreign exchange controls compounded these effects in keeping the premium high. The second phase begins with a recovery of the copper price in 1979, resulting in a relaxation of borrowing constraints. More reotrictive macroeconomic policies were peraued which together with inflows of foreign borrowing sustained the low level of the premium, even when copper prices fell in 1981 (though volatility briefly increased). Exchange rate policy was passive during the entire period, save for two devaluations. The devaluation of March, 1978 has a negligible impact but in January, 1983 a maxi-devaluation of 20 per cent provides a classic instance of expectational response in the premium (this is the only example of a one-off maxi-devaluation throughout). The theoretically predicted time path for the adjustment of the black market to current expectations of a future devaluation of the nominal exchange rate is shown in Figure 3 (c) (Dornbusch et al (1983)). Speculators expect that in the long-run a convergence to the initial steady state level of the premium will ensue; this would necessitate a future depreciation of the black market rate. The profit opportunities this presents induces an immediate increase in the demand for foreign exchange and hence a jump in the premium from A to A'. Thereafter, the premium rises over time to A'', in combination with a growing stock of foreign currency. When the official devaluation is realized, an immediate fall in premium from A'' to A''' occurs, due entirely to the fall in the official rate. T,&e black rate does not alter concomitantly with the official rate since this change was - 24 - anticipated in the initial jump to A'. Thereafter, the transitory accumulation of foreign currency is gradually depleted, and the premium returns to its initial steady - state level. The behavior of the premium in Zambia around 1983 is shown in Figure 3 (c) to follow the theoretically predicted path. The black market rate did not alter at the time of the devaluation, but depreciated subsequently. The derived expected devaluation series covering the period before and after the January 1983 devaluation is depicted in Figure 3 (e) (discussed in III.d). It is noteworthy that both series confirm the prediction of the theoretical model. This supports two assertions mentioned at the beginning of this section. First, expectations are important as determinant of the free rate premium; and secondly, our empirical model of expectation seems to be well specified, at least for the period in question. '° The non-normal distribution of the premium suggested by such speculative responses is discussed further in Section V. 1.3 Episode 3: 198304-198503 This episode coincides with the onset of liberalizing reforms in connection with the 1983-84 IMF stand-by agreement, later intensified in the auction period. Following the January 1983 devaluation more emphasis was placed on active exchange rate policy to facilitate balance of payments adjustments. The Fund recommended delinkage from the SDR, given that its value is determined by a basket of currencies with little relevance for Zambia's trading patterns. A mechanism for gradual, systemative exchange rate "3This demonstration has employed monthly black market rates which therefore embody lags in reaction time. They were derived from Pick's Currency Yearbook, concerning which data cautionary comments have already been made. - 25 - adjustment was installed. Most of the performance criteria under the program -were met. The budget deficit was reduced from 19 to 8 per cent of GDP, in particular by cutting subsidies and personal emoluments. However GDP continued to decline and inflation began to accelerate. The final purchase in the agreement was disallowed due to postponement of paying arrears. Most measures associated with the stand-by facility of 1984-86 were implemented. However, GDP fell by 2.7 per cent and inflation rose above 20 per cent. The facility was inactive from September, 1984 when criteria for the reduction of external payments arrears were not met, and it was canceled in 1986. The black market premium showed a steady fall in this period from 30 per cent to about 17 per cent in September, 1984. An instance of the short-term response of the premium to news is provided by the increase of 10-fold on termination of the IMF program. The steady decline persisted after this jump, but from the higher level and at a reduced rate, despite the increased rate of crawl. 1.4 Evisode 4: 198504-198702 As Figure 8 (a) and Table 1 (b) show, the auction period displays an increased volatility in the premium together with a spectacular trough in April, 1987. The auction raises a number of interesting issues for the analysis of the role of the black market premium. These include the extent to which unification was achieved and the rate which ensued, the short-run impact of shocks and signals on the premium, and finally, the fiscal effects of the unification (Pinto, 1987). Some models predict that anticipation of unification will induce the parallel rate to move to the level expected to obtain on unification. This - 26 - appears to hold for Zambia, with overshooting; Figure 3 (d) shows the extent to which depreciation of the parallel rate tracked the official rate around the auction episode. The premium follows a downward trend within the first quarter of the auction. Thereafter short-run influences shape its path. An important aspect of expectations formulation in the context of structural adjustment is the private sectors' assessment of the sustainability of liberalization measures and the government's commitment to them. These issues of credibility have increasingly been emphasized (Calvo, 1987). Sachs (1989) further suggests that past experience of aborted liberalization attempts makes players particularly sensitive to official signals of commitment and credibility. Zambia's foreign exchange auction provides an interesting illustration of these phenomena. During the auction there were several mechanisms for the transmission of official signals. The amount of foreign exchange put up for auction each week was a visible official control variable: fluctuations induced increasing excesses of dollars requested over those allocated (see Appendix Table A.3(a)). The degree and nature of intervention in the running of the auction affected credibility. Intervention included disqualification of bids, more stringent documentation requirements, suspension temporarily or permanently of the auction and alteration of the type of auction system. Table 3 (c) details shocks and signals influencing the black market premium during the nineteen month auction period. The short-run influence of shocks and signals appear to dominate the behavior of the premium during this period, which also shows sensitivity to "news" about the auction. Official actions signalling commitment and sustainability decrease the premium, and visa versa. The response to political news is evident on Dr. Chivuno's - 27 - appointment an Governor of the Central Banks he was known to be opposed to the auction. Efficiency in implementing the system would be observed at the micro-levelt the large back-log in allocated dollars that built up from August, 1986 was not sustainable (disbursement delays averaged 10 weeks and were not cleared until May, 1987) and the premium is seen to rise thereafter. The implication of speculative responses on the part of firms to official signals was further to impair economic sustainability of the liberalization. Visible speculative behavior implied an uncertain macro-economic environment which decreased business confidence. very rapid devaluation in the final stages, due to speculation by firms, actually rendered the premium negative. With an imminent end to the auction and a clamp-down on illegal activity in prospect, the risk aspect of black market activity is revealed in a deficit wLth respect to the official rate being paid in kwacha for the purchase of dollars. This phenomenon was observed for Brazil (Dornbusch et al, 1983)."1 The auction can be divided into two phases, a stable period from its inception to August, 1986, and an unstable period thereafter, until its demise in April, 1987. This division is based on whether or not there is an appropriate response in the official exchange rate to the quantity of dollars allocated. Although the advent of Chivuno saw a trebling of allocated dollars, a perverse response in the exchange rate can only be observed from Auction 47, on the last day of August (see Appendix A.3(a)). The fiscal position in these two phases of the auction is analyzed in Section 7.2. "We hope in the future to study the impact on the premium of signals and speculative responses during the auction period in the context of a rigorous empirical model using weekly auction data. - 28 - 2 * Model of Expected Devaluation The derivation of a forecast equation for expectations of future devaluation starts with the following assumed government reaction function: r1 (3 ) Alog E0 - bo + E b1L log (RERb/RERO)t_i t+l Lao where E P 0p RE%a b /p and RER 0 p This reaction function implies that if the official real exchange rate RERb was overvalued (smaller) relative to the black market real exchange rate there will be a pressure on the authorities to devalue. Note that given the particular definitions of the real exchange rates adopted, the reaction function is actually governed by previous periods' levels of the premium (since RERb/RER. - q). Now we also assume that agents in the economy know this reaction function and that this influences their considerations; they also take into consideration the record of previous devaluations and the history of monetary emissions. Furthermore, given the various regime shifts over the period considered (see above) a set of dummy variables is included to account for this effect on expectations. Based on this, the following linear forecast equation is used to generate expectations: - 29 - r1 (3.2) Alog Eo 0 bo + E b i log (RERb/RERO)ti t+1 i-0 r ~~~r r + E2b log E0 + E b3 Alog "t_L +$ b4i RSHtL imo 2 t-i i-0 i= 4 - + a t+l where M is the stock of broad domestic money, RSH stands for regime shifts, and c is a stationary disturbance term. The above model was estimated using monthly Zambian data for the period Jan.1980 to June 1987. The results of the estimation is contained in equation (3.2') below. (3.2') AlogE,011= -2.97 .0.17l0gqt-a (-2.73) (1.80) -0. 18 1 ogE0, (2.38) +0.4210gM20 -0.12 JAN83 (2.74) (-0.88) -0.35 OC285 (-2.45) -0.23 JtIY6 +0.25SEP86 (-1.71) (1.82) +0.60 JAN87 +0.66MAR87 (-4.32) (4.46) -1.24 APR87 (-7.92) R2 = 0.63, R2 = 0.58,DW = 2.11 where t-statistics are in parentheses and the method of estimation is OLS. - 30 - The results show significant effects for the role of the authorities reaction to past overvaluation of the official real rate relative to the free real rate given by log REPR - log REN (or simply log q). This factor however, only takes effect with a lag. Overvaluation two months earlier has a positive effect on expected devaluation. The effect of expansive fiscal or monetary policy is reflected by the statistically significant and appreciable positive coefficient of the log of the stock of nominal broad domestic money. This effect is more immediate than that of real overvaluation currently influencing expectations of next period devaluation. This evidence shows the importance of fiscal and monetary policy in shaping agents' expectation about future devaluations. The model also shows, as expected, that starting from an initial conditiorn of high official exchange rate is likely to reduce expectations of future devaluations. This is reflected by the large negative and statistically significant elasticity obtained for logE,_ in the model. The remaining influences on the model are accounted for by a variety of past economic and political events (see Table 3 (c)). In all cases the obtained results are consistent with the interpretation of the model and the discussions of the various episodes in ths coming sections. In January, 1983 a maxi-devaluation was effected which would of course reduce expectations of an immediate future devaluation. The dummy proves to be slightly significant, given the degrees of feedom, and displays the expected negative sign. All the other dummies serve to capture signals of credibility in the course of the auction episode. The auction was initiated in October, 1985. - 31 - In the first of the weekly auctions, the rate fell from K2.2 to K5.O1/$ and thereafter stability was maintained Ln the rate, generating no expectation of an immediate devaluation. From June, 1986 weekly bids were published in an effort to stabilize the auction, a credibility signal likely to decrease speculation. But from September, 1986, after the institution of a 'Dutch Auction', the foreign exchange pipeline began to build up, indicating possible unsustainability of the auction and inducing speculation. In January, 1987 after riots on the Copper Belt, the auction was suspended for nine weeks. Expectations of devaluation were clearly negative; this was dramatically reversed in March, 1987, when a two-tier auction was reinstated. In April, 1987 just prior to the abolition of the auction experiment, tremendous speculation drove down the rate and repressive reaction from the authorities actually rendered the premium negative. A fixed rate was again in prospect. - 32 - IV . A MODEL OP THE PARALLEL PREMIUM The preliminary analysis presented above indicate that the parallel market premium in Zambia is influenced by (or jointly determined with) a variety of factors including macroeconomlc and trade policies, agents expectations, as well as exogenous determinants such as TOT and foreign aid.'2 In this section we develop first a basic model of dual exchange rates in which the premium will be specified as a function of the stock of money, and the uncovered interest parity differential, reflecting the asset market condition; as well as the flow determinants which include the official real exchange rate (E,/PN), permanent real wealth (or stock of money), and the parameters of trade policy. This basic model which does not endogenize the official RER is more suited for estimation with finely sampled data such as monthly data. The main emphasis in such estimation should be to distinguish the long-run flow determinants of the premium from the short-run (dynamic) asset market influences. In this case a formal analysis of stationary and cointegration can be quite useful, because it will resolve the problem of endogeniety and provide a justification for an error-currection dynamic framework -- which allows the distinction between short-run and long-run effect. The second objective of th's section is to extend the basic model by endogenizing REP, as a function of its fundamentals as well as allowing for other exogenous factors such as foreign aid. This extended model will be estimated in section (V) using annual data. In this model the obvious 12 It is also evident that in the very short run a combination of market thinness and agents reactions to frequently changing policies might have caused the premium to follow a skewed or non-stationary process (see appendix table A.5(a)). - 33 - emphasis will be on a more detailed view of the determinants of the premium. 1. The Basic Model We assume a dual exchange rate system where an officially sanctioned share, z, of total exports, are required to be surrendered to the authorities at the official rate. Also a share of imports, v, of total private sector imports, I, is allowed through the official market at the rate E, along with the exogenously given government imports I.. The rest of the commercial transactions are settled at the black (free) market according to the freely floating dual rate, Eb. The effectiveness of the above arrangements depende on the extent of leakages between the two markets which in turn depends on the tax adjusted level of the premium (q = g- = , and enforcement efforts, 1ENF* Therefore we define export smuggling and under-invoicing ratio, , relative to the z exports (zX) as: (4.1) ox =x (1-t I 1ENF)'3 x (+) (-) As for imports we abstract from smuggling and concentrate on *lodelling 'where l is an indicator function in V. - 34 - the over-invoicing ratio of offLcial imports (vl), Ot14 (4.2) r a 0I (t' -AN- ) (+) (-) Equatlon (4.1) sets an arbltrage conditLon for the decision to smuggle or under-Lnvolce exports accordlng to the zxc,ss of Eb over tho tax adjusted nominal official rate E0(1-t). In eq. (4.2) also, the hlgher tha ratio Eb/Ed(l+t1) the hLgher the LncentLve to over-invoice the set of $mports (vi) allowed through the offlcial rarket. On the other hand enforcement efforts will raise the cost of amuggllng or mis-invoicing and therefore both 0k and #I will decllne as a result of intensified enforcement efforts. Key relative price. needed for the specification of total private sector exports, x, and imports, I, are the real exchange rates for exports and imports, which we define as follows: (4.3) RER, p [8Z.*xeb + (1-Z) (1-tX) Eb + Z (1-*x) (1-tC))E0 PM1 = . P, [z1 z x (l-t) (1-tx) q + Z (3-#x) (1-tx)] whers, e., the officLal RER L equal to so PM' Similarly RER, is given by: - (official over invoiced imports/actual official imports) - 1. - 35 - (4.4) PER, PI. 1[(bz+) V'eO(1+ty+tOR) + (1-V) Eb (1+t)] /PN - eo, * Pr' f (40+2+) v(l+t,+C,,,) + (1-V) (1 | -r) q] where tQg is the implicit tax rate that reflects scarcity rent in imports channeled through the official market. Now write total exports as a positive function of RER1: (4. 5) -X = X(RERX) Next we defLne total private imports, (4.6) 1 = (RZ, M + EbF ) VEO + ( 1 -V) Eb (-_) (.0.) where M+E?F stands for private sector financial wealth with M being the stock of domestic money, and F the stock of foreign assets held by the private sector. Nominal wealth is deflated by vEZ (l-v)Eb, the average unit cost of imports. Now we state the following flow equilħbrium condit.iono of the model: (4.7) F[4xZ+(1-Z)] X + v'*I -(1-v) I - 36 - (4.8) A = (X-I) - *+ NkF where is the change in official reserves, and the last term in the right hand side of (4.8) is official net capital flows. The second part of the model is a portfolio equilibrium condition which gives the relati%e stock of foreign and domestic monies as a function of the uncovered interest parity differential. Letting X+, denote conditional expectation of X,+ conditional on information available at time t, the portfolio balance condition can be stated as: me.9 - = g(iet+Alog Eb,.l it) g-<° M49 C+EbFt We also have (4.10) * Alog Eb,., =rAlog qt.1 +*Alog EHoc. where A Is the difference operator. Expectations of future devaluations are based on the expectations model given in equation (3.2) of section III above. As we mentioned in section III this model accounts for the impact on exrectation of past economic policy as well as past economic and political events. By imposing the flow equilibrium condition on the unreported current account equation (i.e. =0 in (4.7)), we obtain the following expression - 37 - for the stationary level of F. (4.11) F = F (RERX, RERy, m, q, tx, t; z, v. 1No) ' (+) (+) (-) (+) (+)(+) (? (?) (-) where m = E- is the stock of real domestic money, clearly for given values EO of the right hand side (especially (RER,, RER, and m), equation (4.11) is a steady state solution for F that provides a long-run (non-dynamic) expression for the premium depending on the fundamentals of the flow equilibrium condition. Equation (4.9) on the other hand is a dynamic specification that governs the dy.amics of the premium, and hence its short-run determination. Putting (4.11) in (4.9) and using (4.10), we obtain the following basic model of the premium in a linear logarithmic form: log qe -1jog qte = do+b (it, +,Alog EO,-it) +82 log mt - 63 log RERX, -64 log fiREr (4.12) + 65 log (l-tx) t +66 log(1+ t1) t + 87 l.MWV.8j This equation provides a basis for estimating a basic cointegrated relationship for the premium." I7 Specifically, if a forward-looking solution of this martingale exists, and if the variables involved (or a subset of them) are cointegrated, then a cointegrated specification and a " Actually the sign on RER, is ambiguous, unless v * > 1-v (see equation 4.7). 16 Refer to the discussion in the second paragraph of this section. 17 The potential endogeneity of RERX, RER, and m, for example, is not a problem under individual nonstationarity and cointegration. - 38 - corresponding error-correction dynamic equation can be obtained as well. 2. SndAoenizinathe URR As we mentioned above, equation (4.12) contains at least four endogenous variablecs the premium, q; the real exchange rates, REP, and RER; the stock of money, ml and the domestic interest rate, i. In this subsection, we will endogenize the real exchange rates by deriving an expression for the real rates by solving for the equilibrium condition in the nontraded goods market. I Following Rodriguez (1989, 1991), the internal balance condition can be stated as follows: (4.13) DN (pg, Px PN) .A - SN(pN, Pz, PX) ' 0 +) (+) () (+) ( ( ) "We abstract here from the potential endogeniety problems of m and L; but under monetary rule A=O for example, the steady state level of m, is given by M= g-t where 7: is the steady state rate of crawl of the o"icial exchange rate, hence the steady state stock of money does not depend on q (see Elbadawi :1991) for details). Regarding the interest rate, no data on the potentially endogenous market determined rate is available for Zambia in a consistent basis over the period considered; we are therefore left with the official rate which can be considered predetermined, at least for the period before the auction. This, however, may call into question the suitability of the official rate for reflecting the opportunity cost of holding foreign money as required in specification (4.9). - 39 - where A is nominal absorption, Y is nominal income, and the definitLon of Pm is extended here to include a weighted average of purely domestic goods, as well as of those imported items where scarcity rent (i.e. t1 o) exists. This leaves P, and PI to be determined by foreign prices, exchange rates and explicit taxes. Now abstracting from misinvoicing we can state the following two basic definitions for P3 and P,. (4.14) P. = PI (1-Z) (1-tX)Eb + Z(1-t:tx) Eol (4.15) p . pz (1-V) (1+td) Eb + V (l+t1) ol Now imposing the nominal price homogeneity condition on the hom good equilibrium condition (4.13), we obtaint DN(. 1PZJ [5Px] pz (1 + td) (4.16) where td = is the resource balance deficit normalized by GDP. The solution of (4.16' and definition (4.14) and (4.15) lead to the following expressions for PX/PN and P,/PM: - 40 - (4.17 ) pR pN -xp td TN OX |. t,,, tr, v, z, q td ( I'r (_) (_) (+) (-) (?) ( ) (4. 18) fx _ PX fX P X'*z, ' i4. 18 p = $ s,td | p t p , #t, tz, v, z, g td From the definitions of RERx and RERN in (4.3) and (4.4) and from equations (4.14) and (4.15) that give the expressions for Px and P1, we can define RERx and RER, in the following generic forms: (4.19) RRx n-' gx (z, 1.Q.? (4.20) 1pm - V, low (?) (?) Now let us define an economy wide real exchange rate as a weighted geometric average of RERx and RER, (4.21) RER RER RERZ * hf (z P) () gPN) where 0 < a < 1 is the share of exports in traded goods. Flnally using (4.17) and (4.18) in (4.21) we can obtain the reduced form - 41 - for RERt (4.22) RER ' RER PX'* I-t=x 4t q, td; v, z, 1lw P( () (-) M) where we assumed the more direct offeet of ;. on RER through px to be stronger than lts influence through -' , but in general the coefficient of PM TOT could not be signed apriori. Now substituting (4.21) and (4.22) in (4.12) gives the following extended expression for the premium presented in equation (4.23) below. The right hand side of the specification includes net capital flow term NKP. NRK enters into (4.23) directly through the official balance effect (equation (4.8)), and as a proxy for the total current account deficit GDP ratio (td). - 42 - log qe-A8log qe.1 - 60 + 81 (it +tAlog Eot4, - it) + 62 log mt + 83 log (1-tx) t + 8, log (1+t.) c - a5 NKFt (4.23) - 6, (a. + a, log ] + aa log (1-t.) c - a3 log (l+tZ) t -a4 eF + 51 ( V, Z, ) ) (?) (?) (?) log q, - 1,log qe, - YO + y,(I, +t A log Hot, -it + y2 log mt + y3 log (1-t1d) t + ySlog(U+t1) c - y5 log P . 4.24) TM t - Y6 NKFC + y71 (v, z, EIUF) (?) (?) (?) where a before dlrect effects are assumed to dominate indirect effects (for the cases of log (11t) and NKF). - 43 - V. MPIRICAL DETZRMINATION OP TIM PRMIUM Estimation of this model with monthly (or even quarterly data) proved to be quite a 'ormidable task because the regime shifts - which tend to average out in the annual estimation - are important in this case. Therefore, we elected to undertake this exercise in a rigorous and detailed fashion in a separate frture work. In section III above however, we discussed the role of expectations as a determinant of the premium using monthly data. In this paper therefore, only the extended model of equation (4.24) will be estimated using annual data. Tables A.S(a) and A.S(b) contain some indirect diagnostic tests, based on the statistical properties of the annual premium. It is clear from the tables that both the unconditional distributions of the premium and the rate of change of the premium exhibit departurec from normality, while no departure from normality is found for the conditional distribution of the premium rate of change (the residual of (5.24'). The potential causes of the departures from normality are mentioned in footnote (9) above; but the important economic implication of the above simple tests, is that the determination of the premium, at least in the long-run, is solely accounted for by economic fundamentals. 1 The -An -a.L.M2ls An eclectic error-correction model based on equation (4.24) of the previous section is estimated using annual data from the period 1970-1987. The variables involved in the estimation are reported in table A.5(c) and equation (S.24') contains the results of OLS estimation. To conserve on the degrees of freedom; we impose the condition that the RER fundamentals (TOT, taxes, and NKF) enter only in levels, ie. they are restricted to influence - 44 - only the long-run *pecification of the premium. (23') Alogq 1.69 - 0.40 log TOTt (1.50) (-1.58) - 0.03 log GRANTt -0.21 log m t (-.91) (-1.33) + 0.42 Alog mt (1.63) + 0.31 ($t 1+ t 1 Alo°Et it_ ) (2.20) -0.45 logqel +0.24DUM(70-77) -0.15 DUM(82-87) (-2.35) (2.45) (-1.30) 2 -2 R - 0.82, R - 0.62, DW - 1.60, Q(7) - 3.4 where t-statistics are in parentheses and the method of estimation is OLS, and Q (7) is the Box-Pierce statistic for residual autocorrelation to 7th order. Given the paucity of the data and the short period available for estimation, the above estimating equation is fairly succossful. The explanatory power of the model is quite satisfactory and no evidence of mis-specification is present. Most of the variables are statistically significant and the signs of the coefficients are all consistent with the predictions of the theoretical model. The asset market determinants: the interest parity differential and the change in the stock of real domestic money, have significant and positive "The expected devaluation sries used in the construction of the interest parity differential term of the above equation, is based on the annualized predicted values obtained from the appendix estimating equation (A.1'). This equation is a quarterly version of equation (3.2) discussed in section III above; see also table A.5(c). - 45 - influence on the premium. Interestingly enough the level of the stock of real domestic money (reflecting the real wealth effect) was not found to be significant. The explanation for this is that the TOT and GRANT (proxy for NKF) terms provide for better indicators of the flow effect of permanent wealth. In an economy highly dependent on the export of a single commodity and on foreign aid like the Zambian economy, these two xactors may be major determinants of permanent real wealth. The effects due to the levels of log TOT and log GRANT on the premium are both negative. No satisfactory measure for the relative tax variable is available, and hence we dropped it from the regression.2' The effect due to the TOT operates through at least three channels. Two indirect channels through the real wealth and real exchange rate flow effects. An improvement in TOT leads to appreciation in real wealth and hence a rise in the premium. The effect of TOT improvements through the RER channel depends on whether the change in TOT leads to RER depreciation or appreciation. The third, and what appears to be the dominant, channel is the direct effect of TOT on the supply of foreign exchange which will have the effect of reducing the level of the premium. The effect of foreign aid (GRANT) on the premium operates through two channels. A reduction in foreign aid (directly) worsens the official current account (equation (10)), as well as the unofficial current account balance through a reduction in v (the share of official foreign exchange sale to mThe effects due to TOT and GRANT are statistically significant at the levels 0.15 and 0.40 respectively. TOT and GRANT appears to be correlated, however, and re-estimation of the above equation with GRANT dropped improves the significance of TOT coefficient. 2'In fact we included a term given by import tax revenue to GDP, but the results worsened substantially. - 46 - imports) A reduction in foreign aid however, will ceterius-paribus lead to real depreciation and hence a reduction in the premium. The results show that this last channel is dominated by the dire.t effect on the flow balance. The above evidence indicates that the recent sharp decline. in the copper price. as well as the increasing difficulties of external finance, might have compromised economic reforms aimed at reducing the premium and achieving ultimate exchange rate unification. The negative significant effect of the lagged log of the premium, shows that a high premium in the previous period is likely to lead to less increase in the rate of growth of the premium in the following period. Also the negativity of the coefficient of logq.1 implies that in the long run, the premium depends negatively on TOT and GRANT. Finally DUM (70-77) and DUM (82-87) represent two different exchange rate and trade regimes where the premium rose over the first period and steadily declined during the second, see figure 1 of section 1 above. - 47 - VI. TIE EFFECTS OF TEB PREMIUM ON TRADE. In this section the importance of the premium's link to reported and unreported (illegal) foreign trade transactions will be assessed. We will estimate econometric specifications (based on the reduced forms of section IV above) for reported import demand and reported copper export supply,! while the extent of misinvoicing in foreign trade will be quantified using various statistical techniques. 1. Reported Trade Flows. Reported Total Private Imports: According to equations (4.1) and (4.6) of section (IV) above, we can write reported private imports (I, = V * . I + (1-v) I) in the following generic form: (6.1) Ir 2( V + (1-V)) (e q, m, F; v,l (+) ( ) ( + ( ) According to the above equation e. - EO /PN has a negative impact on recorded imported, while I, is positively influenced by real wealth (m and F). The effect due to the premium could not, however, be signed a priori: a rise in q discourage demand for aggregate imports I, but it also lead-for a given tax rate-to an increase in the over-invoicing ratio, ˘X. Furthermore, clearly both I and q are jointly determined; hence the endogeneity of q must be corrected for in the estimation of the I, specification. I, is also t Zambian exports other than copper are very marginal. Also the technology and arrangements of copper exports in Zambia do nmt allow for misinvoicing or smuggling of copper. -48 - influenced by discretionary government policy: positively by the ratio of official foreign exchange sale, v, and negatively by government enforcement efforts. Based on equation (6.1), we estimated an empirical linear logarithmic dynamic equation for 1, using quarterly data from 1970:3-1987:4. The empirical specification is based on the framework developed in Moran (1989). d Moran (1989) specified a demand for imports Ir and a linear relatir-;ship linking the long-run volume of imports I with long-term reserves r* given by (6.2) r=;a + Ireo, oSa1 S1. And another specification linking actual change in reserve to foreign exchange availability, f, and actual imports: (6.3) Ar = -I d Subject to these two conditions and a specification for the demai,d for imports Ir he posed the problem as choosing I, to minimize: (6.4) Cc b1(Iz-X;˘)2 + + b3(I + b4 (I-I In our case we use a linear logarithmic specification for 1d, based on - 49 _ (6.1) with real wealth (m and F) represented by real output* (6.5) log Id = a0 + a, log qt + a2 log e. + a3 log ye + a4 log (l+t) e +asltZHFv] where 1p,] represent indicator variables for enforcement and commercial policy. Minimizing (6.4) above and using (6.2), (6.3), and (6.5), we obtain the following eotimating equation for reported imports, where we impose the condition that 1; = f - constant. (6.6) log IT. = P + log qt + P2 log eO + P3 log Yt + P4 log (1+td) t + P5 log *r- + P6 log r1-1 The estimation results of (6.6) using quarterly data from 1970:3 - 1987:4 is preuented below. (6.6') log Ir = -6.06 + 0.09 log qt-0.18 log GO (-1.53) (1.30) (-2.98) . 0.S3 log yt + 0.22 log Rt_ (1.21) (5.36) + 0.72 log Ir (11.88) -2 R 0.87, DW - 1.94, Q(24) - 27.2 t - statistics are in parentheses and the method of estimation is OLS.m "We also used M2/Eo for real wealth but it was found to be insignificant. 24Attempts to correct for the potential endogeneity of the premium, q, by running a 2SLS were not successful in producing a meaningful empirical specification for I,. - 50 - Other potential determinants of I, such as foreign trade taxes, and the indicator variables for foreign trade policy such as own resources or enforcement efforts were all not found to be significant. The above results seem to provide broad support for the predictions of the model. The effect due to e, is appreciable and significant. Of'icial real appreciation will led to increased demand for official or recorded imports. The effect due to the premium is significant at 10 percent significant level and with positive though small short-run elasticity at 0.09; the derived long-run elasticity, however, is high at 0.32. This result shows that the effect of the premium through the over-invoicing ratio dominates its negative influence on aggregate demand for import. The effect due to real income is quantitatively high, with a 0.63 short-run elasticity and a long-run elasticity bigger than 2; but the income effect is only marginally significant. The results also show the presence of considerable inertia in the demand for reported importss the elasticity of lagged imports is equal to 0.72 and is significant at any reasonable significant level. Finally it is also clear from the positive and significant effect due to official reserve that the effect of foreign exchange constraint is an important factor in the determination of the level of reported imports. Copper Exports: Four main considerations figure in our empirical estimation of copper export specifications in Zambia. First, we decided to model it as export suppiy with foreign prices for copper taken as exogenous. The exogeneity of foreign prices is justified by the slowing international demand for copper and - 51 - the declining share of Zambian copper exports.5 In this case, the main relative price factor should reflect the relative domestic structure of incentives for copper production. specifically, we considered a relative price, RER,&, given by the ratio of the domestic price of copper to the building price index, the latter being a proxy for the domestic price of non- traded goods. Second, the technology of copper export supply in Zambia effectively precludes smuggling or mis-invoicing of copper exports; thus the parallel premium is not likely to have a direct disincentive impact on copper exAort supply. The premium, however, is a good proxy for foreign exchange sho-tages, which during 1975/76-1989 became the most important constraint, with a direct impact on copper production and exports. The third consideration in modelling copper exports in Zambia is the issue of regime shifts. There is clear evidence of a regime shift in the production &nd export of copper over the 1960-1990 period. From 1960-1975, transport presented the principal constraint for the copper companies. Zambian production fell 75,000 tons in 1966, due to transport problems and fuel shortages associated with the unilateral declaration of independence in Rhodesia (1965). Another one-shot event that impacted on copper exports was the Mufulira mining disaster in late 1969. The completion of the TAZARA railway line late in 1975 saw a surge in exports in 1976 as accumulated stocks were cleared. From 1975/76-1989, foreign exchange and skills shortages predominated. Increased government control after nationalization in 1970 made nThe Zambian share of world copper production fell from 13 per cent in 1961 to 6.7 per cent 1982-84 (World Bank Commodity Staff Working Papers, No. 15, 1987). - 52 - the companies susceptible to multiple and sometimes conflicting objectiveg..6 During the 1980s, the major factors responsible for a 30 percent fall in production were skills and foreign exchange shortages, transport problems and poor mineralogy (a declining ore grade over time). Further, there was considerable government diversion of the company's resources into non-mining expenditure. These constraints outweighed the effect of a more favorable RER, . .27 There were two episodes of export revenue taxations a dual regime of royalties from 1960 and an export tax on sales from mid-1966 up to mid-1969, and a mineral export tax on sales from 1984-88. The precise impact of taxation on the copper sector requires a more detailed study, but it can be argued that the taxation regime of the 1960. (aimed at windfall gains during the 1964-74 boom) constituted no serious disincentive to supply in the current and immediate future, but rather to new mining investment (Garnaut and Ross, 1985). The taxation episode in the 1980s, however, proved damaging to the company. The fourth and final consideration is that the exportable volume of copper will be affected by capacity constraints; increasing production in the short-run can be quite costly, hence considerable inertia may be present in the production and export of copper. The above considerations motivated us to model the export of copper (in thousand of tonnes) as a linear logarithmic function of the following X The copper companies were 50 percent nationalized in 1970, but continued to be managed by the previous owners until early 1975. The state's share was increased to 60 percent in 1979, and the companies were merged to form a singlo copper company in 1982. " ThG constraints during the 1980o are discussed in detail in Aron (1992). - 53 - variablest the real exchange rate for copper as defined above, expected to have a positive effect since real depreciation should improve the external competitiveneas and the domestic incentive structure for copper; the rate of change in the parallel rate premium as a proxy for foreign exchange shortages, which should have a direct negative impact, and also possibly, though of lesser importance, an indirect negative effect through its influence on the domestic price of nontraded goods; the volume of copper exports in the previous year to account for the inertia and capacity constraints; finally, a host of dummies reflecting the various regime shifts we discussed above. The following regressions were estimated using annual data from 1960-89 (t-statistics are in parenthesis). (6.7) log EKP,,., =(4 76) + (3.00) log RERcu, x DUM(1976-89=0) *(392) lg XPcut_l (-2 46) A log q. (-2 30) DUM(1966) 03.920 0.246 -0.14) (2 74) DUM(1972) + (3 426 DUM(1976) (-3. 02) DUM(1985-88) R2 = 0.87, R2 = 0.83, Durbin H Statistics - -1.56, s.e. = 0.07 (6.8) log EXPCu= (3 26 + 0 12 log RER0.b, + 0449) log EXP':1 (4.52) (2.93) og (4.37) - 0.15 A log qc - (°i. 53 DUM(1966) (-2.63) - -2.03)DU(66 + 0 18 DU0(1972) + . 24 DUM(1976) (2.46) DU3(197341) 0.18 DUM(1985-88) (-3.82) - 54 - R2 = 0.87, ? 0.83, Durbin H Statistics = -2.13, o.e. = 0.07 Equation (6.7) estimates the joint influence of the REP;* and the regime shift for 1976-89. Comparing the coefficient of the interaction term log REFt" x DUM(1976-89) in (6.7) to the coefficient of log RER;., in (6.8) confirms the presence of this regime shift. Furthermore, the relative stabilicy of estimated coefficients of other variables in the two equations indicates that the effect on copper exports of the regimes shifts operates through the structure of incentives for copper i.e. the RERob effect. Overall both equations lend strong support to the model, with individual effects statistically significant and consistent with prior expectations. The model explains over 80 percent of the variations in the dependent variable. Equation (6.7), which captures the regime shift, will be considered for further analysis. Equation (6.7) shows export supply to be strongly influenced by the relative structure of incentive in the domestic economy as reflected by the 21 percent short-run elasticity for the REPR* effect, which also reflects the influence of regime shifts operating through the RER.^. In the long-run when full adjustment and substitution take place, the estimated effect due to RER,, will be much higher with an elasticity close to 0.5 (equal to 0.21/0.45). This result shows that copper export, despite its technology of production and export arrangement, is strongly linked to the reet of the economy. Also according to equation (6.7), a 100 percent rise in the premium will lead to about 14 percent decline in the export of copper in the short-run, with the corresponding decline in the long-run as high as 30 percent. As we suggested earlier, the premium is likely to be reflecting the direct effect of foreign exchange shortages. The estimated combined effects for the RERc, and the - 55 - premium on Zambia's copper exports thus provides a strong argument for unification based on its macro-level resource allocation effects (Pinto (1989) makes a similar argument for Ghana). Finally, the estimated effects due to the dummies are all consistent with the likely influences of the episodes they represent. The dummy for 1972 captures the surge in exports due to recovery from the 1969/71 Mufulira mining disaster. The transport dummies, D66 and D76, capture increased constraints after U.D.I., and the export surge with the opening of TAZARA. Lastly, the results show that the mineral export tax on sales (DUM(1985-88) had a serious disincentive effect on copper exports. The export supply is characterized by considerable inertia, with the lagged export supply having an elasticity of 0.45. This result is fairly consistent with the prevailing technology in the sector. 2. The misinvoicing of foreign trade The misinvoicing of foreign trade arises in response to excess demand created by the use of exchange controls and trade restrictions. An overvalued exchange rate and foreign exchange controls create excess demand in the market for foreign exchange, leading to the formation of a black market with a premium on the official rate of exchange. Profit opportunities created by the existence of dual rates induce additional supply and demand in the market. Importers, using legal trade channels, may over-invoice their imports to obtain extra currency and sell the balance on the black market. Similarly, exporters have an incentive to under-invoice exports.3 The illegal proceeds 2Where there are restrictions on the private holdings of foreign assets, flight from domestic currency via over-invoiced imports or under-invoiced exports could be illegally directed abroad. - 56 - of smuggling activities channelled through the black market provide a further source of supply. Additional foreign exchange requirements arising from under-invoiced imports or over-invoiced exports would be also met by the unofficial market for foreign exchange. This type of misinvoicing of trade is due to evasion of trade and price controls. In Cooper (1974) the blanket term "smuggling" is used to encompass four types of evasion of duties and taxes: the under-invoicing of exports and imports, under-assessment of duties, the mis-classification of imports, and failure to record trade, in which case it is either of the "ships in the night" variety using illegal ports of entry, or passes covertly through legal channels. Over-invoiced exports may also arise where export subsidies exceed the discount obtainable on the unofficial market.9 Finally, where price controls imposed on exports or importables create a premium with respect to border prices there is an ince tive to evade these controls by smuggling through illegal channels. The successful pursuit of "smuggling" and capital flight is constrained by the vigilance of customs officials and the degree of complicity in illicit trade. Low penalties, laxity in enforcing penalties and diffieulties in policing large borders will facilitate illegal trade. The combinations of incentives arising in consequence of various controls serve to drive misinvoicing in different directions." A mix of smuggling types and modes of capital flight may be present at any one time. 'Foreign exchange retention schemes aimed at promoting non-traditional exports constitute a de facto subsidization. tThere are a number of legitimate reasons for inconsistent trade data apart from illegal trade activities, and these are discussed in Section 2.1. - 57 - Penalties for different kinds of illicit trade and the degree of enforcement may differ. In the aggregate therefore, whilst an order of magnitude for unrecorded trade may be achievable from discrepancies in reported exports and imports when matched against partner-country trade data, attribution to a particular form of illegal activity is not straightforward, and may be impossible. Cancellation effects may well result in an understatement of the magnitude of unrecorded trade. For an individual importer who faces dual incentives to under-invoice for evasion of tariffs or quotas and to over-invoice in the face of a substantial black market premium, the direction of misinvoicing will be determined by the difference between the black market premium and the rent due to an import license or the rate of tariff, further discounting for the risk entailed in engaging in illegal dealings. By its nature, illicit trade is difficult to detect and to quantify. However, where strong incentives have been created for these activities, detection and an estimate of magnitude is important. Reported statistics will not reflect the actual extent of trade nor the nature of capital flows. Under-invoicing of imports and exports and smuggling results in the recorded volume of trade being smaller than the "truew volume, while over-invoicing of imports and exports, will exaggerate the volume of trade. Where illicit activities are extensive, reported trade statistics will be inaccurate and policy prescriptions relying on them may be compromised. In the last two decades a growing literature has attempted to integrate illegal trade phenomena into the pure theory of internatlonal trade. Statistical methods originating from the Morgenstern (1950) study continue to be explored to assess the accuracy of international trade statistlcs and to - 58 - detect and quantify illicit trade activities. In Yeats (1990), two decades of Sub-Saharan export data were examined for trade discrepancies with developed countries and other African countries. Major disparities were found, and these were substantially larger for intra-African trade. The study concludes that these data are unreliable as regards the level, composition, direction and trends of African trade. In what follows, methods from the empirical literature are briefly outlined and are applied to Zambian import statistics. 2.1 Statistical methods for the detection of unrecorded trade. Direct methods of detection such as apprehending smugglers, comparing individual invoices with world prices and interviewing traders, are clearly limited in scope. Indirect methods rely on comparisons of quantity, price and value data, both between partner countries and within the country. Table 6 (a) summarizes expected price, quantity and value discrepancies for different kinds of illicit trade (adapted from Bhagwati (1981)). This classification encompasses most reported statistical methods for the detection and quantification of faked invoicing and smuggling, and indicates the poebibilities for distingu.ishing amongst different kinds of illicit trade. The indirect methods are applicable at various levels of disaggregation. Available data may only permit aggregate comparisons of quantity, value and price data. Aggregate comparisons provide a lower bound for illicit activity, but where individual classes of commodities can be examined, more information will be provided about incentives for illegal trade and the nature of the activity. Discrepancies in value data between countries may be due either to the misrepresentation of prices or of quantities of Imports. Partner-country-data - 59 - comparisons of quantity or value were first explored by Morgenstern (1950). Discrepancies between the imports recorded by a country and exports to that country recorded by itu trading partners (or between recorded exports and partner-country imports) may be due to a number of factor. besides false invoicing, some of which can be corrected for.3" But once allowances have been made for these factors, a positive or negative percentage difference indicates mis-invoicing. In the case of imports, a positive percentage difference implies net over-invoicing (PfQf - PcQc < 0 in column 5 of Table 6 (a)), while a negative figure indicates net under-invoicing. The reverse holds true in the case of own-country exports.32 2.2 An ao2roximate guantifigation of misinvoicing in Zambia. The empirical analysis is confined to unrecorded imports. Zambia has a "'(a) International transport and insurance coets may be captured by converting imports from a c.i.f. to an f.o.b. basis. Bhagwati et al (1974) assume this discrepancy to be 10 per cent on average. Yeats (1990) has calculated transport correction factors for trade with the U.S.A.. (b) Lags in recording imports which have already been recorded as exported by the partner country leads to an understatement of own country imports. Averaging successive years,before comparing data may smooth this discrepancy (Raya and Morgan, 1974). (c) Under-invoicing by partner countri$e will understate their exports and in the comparison show up as overinvoicing of own country imports. The incentives for trading partners to fake invoicing can be assessed and the comparison be conducted across several countries. (d) Other error sources are less easily adjusted for: errors of commodity classification, different recording definitions and simple errors in counting and recording. Multiple exchange rates may also cause trade discrepancies where rates alter during the period reported, or where the wrong rate is used for conversion. NWhere the data is available, partner-country-data comparisons of traded volumes may establish discrepancies, as in Richter (1974) for Indonesian rubber exports. Even without comprehensive quantity data, it is posesible to distinguish between value discrepancies arising from quantity an opposed to unit price differences, because they generate different inequality signs for the disparity between the declared landed import price and the domestic price of the import (column 1 of Table 6 (a)). _ 60 - high import-dependence: imports as a percentage of CDP averaged over 30 per cent from 1964 to the present day. The principal imports are capital and intermediate goods which display product differentiation. Under-invoicing of exports in Zambia is unlikely to be a significant component of capital flight, given the predominance of metals in the composition of exports trading at known world prices. This has been confirmed by Yeats (1990) where export discrepancies for Zambia in 1982-83 with industrial countries were in the range of transport correction factors.0 Export smuggling in the emerald industry, on the other hand, is reputedly rifo. The report of the 1979 Commission of Enquiry into the emerald industry claims 35-40 per cent of the world market share for Zambia and sUggests that 80 per cent of the emeralds produced are mined and exported illegally. The report recognized the role of a subatantial black market premium in stimulating emerald smuggling. In Table 6 (b), the recorded exports of Zambia's trading partners are compared with Zambian recorded imports from 1968 to 1985. The ratios R - 100.(Mz-Xf)/Xf, where Mz are Zambian-recorded f.o.b. imports and Xf, partner country-recorded f.o.b. exports, are shown for the U.K. and the U.S.A. (which comprise on average 22 per cent and 10 per cent of Zambian imports, respectively), for the six main trading partners, excluding South Africa and Saudi Arabia, and for the O.E.C.D. as a whole (comprising 60 per cent of "Export trade with Malawi in 1982-83 (largely non-copper), however, showed clear evidence of under-invoicing (poosibly exacerbated by import over- invoicing on the part of Malawians). 3'South African exports to ZambLa are contained within an aggregate for trade with Africa and comparison was unfortunately not possible: in the last decade South Africa has provided about one fifth of Zambia's imports. Saudi Arabian trade is predominantly in oil, which analogous to copper is unlikely to be misinvoiced; but other studies have found significant discrepancies due - 61 - imports). Outside of the above-mentioned countries, trade is erratic and limited, and the data patchy; so further comparisons were unlikely to be meaningful. Increasing product differentiation probably facilitates misinvoicing as departures from standard prices make confirmation of prices more difficult. Disaggregating imports to focus upon manufactured imports (SITC 5-8) in the comparison should amplify disparities (Bhagwati et al, 1974). However, the majority of Zambian imports already falls into SITC categories 5-8: these account for 77 per cent of imports in 1970 and 1975, and 70 per cent in the early 1980s. M Trends shown for the U.K. and the U.S.A., the largest non-oil trading partners apart from South Africa, are similar to those for the six main partners and the O.E.C.D. (R-values are larger for the individual countries, in part because cancellation effects are diminished). All four series suggest that capital flight through overinvoicing was rife during the copper boom, 1966-1974. After the 1971 copper price shock exchange controls and import levels became more restrictive. The premium doubled in 1972 and over-invoicing was sustained; but the terms of trade recovery from .473 saw a diminution of the premium and misinvoicing. A precipitous fall in the copper price in 1974 induced a further tightening of trade policy and exchange controls. The premium more than doubled from 1975-77, further influenced by an increasingly overvalued exchange rate and the persistence of poor terms of trade. In tandem, misinvoicing grew (this is most apparent in the U.S.A. figures). to confusion over the ownership of refineries and petroleum (Bhagwati, 1974). 35Categories SITC 1-4 comprise food, beverages and tobacco, crude materials, animal fats and mineral fuels. The exclusion of Saudi Arabia then serves to remove the mineral fuels category. - 62 - From a peak of 190 per cent in 1977, the premium fell steadily during the second and third exchange rate policy episodes until the auction. Between 1978 and 1979 alone, the fall amounted to almost 40 percentage points, induced primarily by favorable copper prices. Misinvoicing was negligible until 1981 when a sharp deterioration in the terms of trade saw a revival of over- invoicing. In the two years preceding the auction of October, 1985, the premium averaged 30 percent, a low level compatible with the maintenance of oŽxchange controls. Under-invoicing appears to have taken place in 1984, while misinvoicing falls in 1985 with the onset of the auction. Given the paucity of data in the 1980s, the complementary nature of import unit value comparisons proves useful. A rough aggregate measure of the misinvoicing trend for imports is achieved by the comparison of import unit values between countries with a similar composition of imports. The difference in their respective tendencies to over invoice may be captured by the difference in reported CIF import unit values, converted to a common currency (Bevan et al, 1990). In Table 6 (c), the Kenyan import unit value for manufacturing converted to Kwacha is compared with the constructed value for Zambia. Disaggregated import unit value indices for manufacturing were chosen to ensure that a similar composition of imports was being compared, but also because product differentiation in manufactured goods make them the most liable 1' misinvoicing. The trends for misinvoicing compare well with the partner- country-data trade comparisons. What is clearly apparent in the series is a tendency to over-invoicing when copper prices fall (after 1974 and 1981) and 36The Kenyan black market premium was not pronounced over this period averaging 10 per cent for 1975-80 and 20 per cent for 1981-84t in 1985 and 1986 it virtually disappeared. - 63 - under-invoicing whsere terms of trade improve (1973/4 and 1979/80). An interesting feature is the high level of over-invoicing seen during the early 1980s, which fell back prior to the foreign exchange auction, when under- invoicing predominated 37 This correspondence between the R-values and changes in the premium and trade regime may be shown more formally by regressing the trade ratios on a constructed series for the smuggling incentive. Following McDonald (1985), the incentive is defined as I=(q/(l-ti))-1, where q is the premium expressed as a ratio, and ti is the import tax rate. McDonald (1985) examined exports only, where tx was proxied by the average tax rate for the export aggregate. For import restrictions in Zambia, quotas are more important than tariffs. A measure of implicit trade policy is given by the differential between the CIF value of imports (pre-trade policy prices) and the retail price of domestic import substitutes (post-trade policy). This is shown in Figure 6 (a) and ia used here as a proxy for ti, capturing the endogenous trade liberalization of the boom period and subsequent tightening of restrictions. The initial figure is set at 15 per cent, the average tariff value for non-luxury goods in 1969. Statistical characteristics are given in Table 6 (d) for I and the smoothed series of R-values (expressed as a ratio of reported imports to partner exports). The mean of I is fairly large (1.27) with wide variation about the mean. A positive correlation between the trade discrepancies and the smuggling incentive is suggested in Figure 6 (b)38, and supported by the simple correlation coefficient of 0.75 (1969-80) (Table 6 (d)). 37 The fiscal implications of this altered trend in misinvoicing on exchange rate unification are examined in Section VII, 2. fThis is shown only for the U.S.A. and the O.E.C.D.: other series showed a similar correspondence. - 64 - The trade ratios were regressed on I and the lagged dependent variable and dummy variables used to capture regime shifts and exogenous shocks: TR = a + bI + cTR + D. t t t-1 The results are shown in Table 6 (a) and compared with McDonald's results for Zairean exports. The constant should display the level of the trade ratio in the absence of smuggling if trade is constant; otherwise the steady-state level (a/(l-c)) should approximate 1 if transport costs have been accounted for. All three import regressions incorporating the lagged dependent variable display the latter feature. In the McDonald etudy, f,r the nine countries examined, fifteen export regressions were run and in only four of these did I prove significant. For Zambian O.E.C.D. imports too, the coefficient on the smuggling incentive for the first two regressions is positive but not significantly different from zero. Previous misinvoicing exerts a significant influence on current misinvoicing. The 1974 and 1976 dummies have the expected signs: the 1973/74 terms of trade recovery saw the premium fall and with it misinvoicing; after 1975, poor terms of trade combined with tightened controls produced the opposite effect. Inclusion of the lagged incentive in the third regression produces a combined positive coefficient for I with improved significance and the dummy has the expected sign. Some limited support is provided by these results for the trade ratio and smuggling incentive link. However, the results are sensitive to the period examined and the dummies included, while poor data and institutional and regiae changes exacerbate the problems of equation instability. - 65 - VII. THE EFFECTS OF THE PREMIUM ON DOMESTIC INFLATION AND FISCAL BALANCE. In this section we will assess the importance of the premium's link to domestic inflation and investigate the impact of exchange rate unification on the fiscal balance. We will estimate a simple econometric model of inflation in which the premium affects domestic inflation directly through the traded goods prices and indirectly through disequilibrium in the money market. We estimate a seigniorage Laffer Curve for the auction period and establish empirically the location on this curve of sub-periods of the crawling peg and auction episodes. The government is shown to be a net buyer of foreign exchange and fiscal revenue implications are drawn. 1. Domestic Inflation To study the relevance of the parallel rate (or premium) to domestic inflation we consider the following simple model. Let the domestic price of tradeables, PT, be given by (7.1) PTEEb-P 3 where P* stands for foreign price and OSasl. Let the pri'e cf nontraded goods, P.N be specified as a mark-up over PT. (7.2) PNd=P+(l+1) and the rate of change in the mark-up function p is made to depend on excesa 39 A more general specification than (7.1) can be considered where P+Eo could be adjusted for taxes and P*Eb for the unit cost of smuggling (e.g. Chhibber and Shafik (1990). These extensions will not influence the final inflation 'quation however, if we assume zero rates of changes in smuggling cost and tax rate. - 66 - demand for nontradeables. (7.3) ft-flc.-Y.) Excess demand for nontradables (CN - YN) can in turn be linked to disequilibriums in the balance of payment and the monetary sector. A balance of payment surplus (deficit) must be accompanied by an excess demand (supply) for nontraded goodi of equal magnitude, measured in domestic prices. Given the importance of copper prices to both private sector incomes and public sector revenues, and the limited effectiveness of the authorities in Zambia at shielding the domestic economy from balance of payment surpluses brought about by the price of copper; we specify the excess demand for nontradables to depend on the real price of copper, P. i.e. (7.4) CN - YM = g (Pw) Equations (7.3) and (7.4) permit a direct link between inflation and the price of copper -- reflecting the usual Dutch disease phenomenon. The income-expenditure model focuses on the goods markots and therefore it neglects the link between the aggregate market for goods and the money market. In an economy such as the one in Zambia where a sizable and tolerated dual market coexist with a highly regulated official sector, monetary disequilibrium (W>P.m') can lead to nontraded goods price increases, depreciation in the black market exchange rate, and worsening in both the regulated official as well as the unreported private current accounts. To the extent that the black market is allowed to float freely, the impact of monetary disequilibrium is likely to be more reflected in terms of nontradables price inflation and depreciating black market rate (or rising - 67 - premium). Based on the above discussion, we add to the mark-up function the change in the rate of depreciation of the black exchange market rate, AA'b b to reflect monetary disequilibrium: (7.5) M5-Pmd =A b) Now using (7.4) and (7.5) on f (CN-YN,MM-Pmd) ) we write the following expression for the rate of change in the mark-up functic..: (7.6) a1 = at log PCUp + a2 ^£b Now we define the aggregate price equation, P, as a weighted geometric mean of PT and PN. (7.7) P=Pr PL_A and O