-- 0~~I' -wS1396'. - POLICY RESEARCH WORKING PAPER 15 A Typology of Foreign T~~~~~~~~~~iransparerit policy rdes.ad cnuct-are o aaon Exchange Auction Markets monetohescsso in Sub-Saharan Africalbegechneutos. Janine Aron Ibrabim Elbadawi *A ;;=- . -v..- - . ". ' ' : ' ' ' 3 ' 0 ~~ ~ ~~~~~ ~ ~~~~~~~-.." '-.-.'9. - , The World Bank ---' Policy Reseach Department r Macroeconomis and Growdt Division Decemiber 1994 WI _w '- ' A OLICY RESEARCH WORKING PAPER 1395 Summary findings Aron and Elbadawi compare and contrast the design and Tle auctions in Nigeria and Zambia, on the other outcomes of differenr foreign exchange auctions in four hand, were subject to frequent policy interventions, countrics in Sub-Saharan Africa and present a typology rcsulting in unsustainable auctions, inefficient allocation of such auctions. of foreign exchange (through ad hoc disqualificadons), They identify two distinct sets of countries in terms of limited unification, and a rather volatile exchange rate. the auctions' fcaturcs, policy interventions, and The conclusions reached by Aron and Elbadawi are outcomes. broadly corroberated by a statistical analysis of weekly In Ghana and Uganda, the exchange rate auctions are micro-auction data for all four countries. judged to have been largely on target in exchange rate unification, exchange rate stabilization, and efficient allocation of foreign exchange. This paper -a product of die Macroeconomics and Growth Division, Policy Researh Dcpartment - is part of the departmental project "Foreign Exchange Auction Markets and Exchange Rate Unification in Sub-Saharan Africa" Copies of this paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Raquel Luz, room N11-053, extension 39059 (33 pages). December 1994. The PolkyR c Worng Paper Scre by thePocy Re fserch D in pras to encong ter cwhn of ;e abo psaperscarytb-ae xnsof ibLacwsnd shodd bcusa ox.decru~. ieabcgf, interprctatiozs,andw chLond arnedct azudxrs' own and shouldnot c&bei tod iWeortd B-,itsE%c tBoardof Dvrs, orany of its mcenbfrcoYtris Producd by the Policy Resach Dissemnation Center A Typology of Foreign Exchange Auction Markets in Sub-Saharan- Africa- Janine Aron Ibrabim Elbadawi The ahors are grateful for helpfi3l commets from Miel K;guel, John Muelbau, Stephen OConnIel, Rafil Tenorio, Katbryn Domingucz, Chad Leechor, and [int PiiDchet. Table of Contents 1. INTRODUCTION ............................. I 1.1 Multiple Foreign Exchange and Auction Markets in Sub-Saharan Africa .. . . I 1.2 The Rationale for Transitiona Foreip Exdcange Auctions in SSA ..... . 3 2. DESIGN CHARACTERISTCS OF AUCTIONS IN SSA ............. 4 2.1 Duraionand frequency oftheaucdons ......................... 5 2.2 Pricing ............................................. 6 2.3 Competitionand thenamueofbids ............................ 6 2.4 *Entry and alocative resictions ............................ 7 2.5 Informational rle ............................ 9 2.6 Transpawry ............ 10 3. A STATISTICAL ANALYSIS OF MICRO-AUCTION DATA IN SSA .10 3.1 Stndard Descriptive Statistics. 11 3.2 Normalty issues .12 3.3 Unit Root Structural Shts and Stationariy Issues .. 15 4. A MACRO-ECONOMIC OVERVIEW OF THE SSA FOREIGN EXCHANGE AUCTIONS .17.. - 5. CONCLUSIONS AND POLICY LESSONS ........................... 18 REFERENCES .............. 21 List of Tables TABLE 1: Some design characteristics of SSA foreign exc&ange auctions .... ....... 22 TABLE 2: Average weeldy size of auction variables by aL -on type ............ . 23 TABLE 3: Measures of normality for fr-differenced auction variables. .... ...... 25 list of Figures FIGURE 1 a,b,c,d: The equilibrium auction price and parallel market premium . ...... 26 FIGURE 2 a.b,r,d:. Skewed and kurtotic distnbutions of auction exchange race changes . . 28 FIGURE 3 a,b,c,d: Cochrane persistence test for the auction exchange rates. .. . 30 FIGURE 4 a,b,c,d: Macroeconomic policy indicators and the free exchange rate. .. 32 1. INTRODUCTION Following the economic crisis that impacted Sub-Saharan Africa (SSA) from the second half of the 1970s, the unification of exchange rates (official and parallel)' and integration of parallel markets into the official economy, has been a major policy objective for reforming African countries. The extent of macroeconomic imbalances (especially fiscal) that have prevailed in SSA and the rudimentary nature of economic institutions (such as the banking system), provide a strong case for a gdualistic approach for exchange rate unification in SSA. In this context, centralised multi-unit auctions for foreign exchange were introduced from the early 1980s in SSA as a transitonal medium towards a credible and sustainable unified regime (such as. an efficient, decentralised interbank market). A brief overview of the structre of foreign exchange markets in SSA (including auctions) is contained in sub-section 1. 1 below. Further, sub-sction 1.2 provides a discussion on the objecdves and rationale for the choice of foreign auction markets as an initial component of the exchange rate unification reform in SSA. The objective of this paper is to conduct a basic analysis of auction data to explain the different outcomes regarding the auction objectives (e.g. exchange rate unification) that have been observed in SSA This paper studies. foreign exchange auctions in four African counties: Ghana, Nigeria, Uganda and Zambia. The auctions' periods covered by the study are: 19 Sep. 1986 to 5 March l992 for Ghana, 26 Sep. 1986 to 24 Nov. 1988 for Nigeria, 31 Jan. 1992 to 7 May I993 for Uganda, and 11 Oct 1985 to 24 Jan. 1987 for Zambia. While auctions in Ghana and Uganda are generally considered to have been fairly sucessf those of Nigeria and Zambia proved unsustinable or produced considerable exchange rate volatility. We hope that the typology of auctions developed in this paper will help to inform the imroved design and conduct of auctions at the micro-leveL A further objective of this paper is to provide the background for a more rigorous analysis in a companion paper (Aron and Elbadawi, 1994), which models the micro-es of the auction rates for the above four countries, and tst some policy propositions motivated by aucton theory. 1.1 Multiple Foreign Exchange and Auction Markets in Sub-Saharan Africa For most SSA countries, exchange rate policy was static from Independence until the 1980s, playing no role in balance of paymen adjustments. The fixed exchange rate regimes of the 196Ds and early 1970s appeared to be viable, given consistent macroeconomic policy and a relatively favorable exal environment By the second half of the 1970s, however, hese economies had been weakened by public expenditure. booms (exhibiting ratcet effects) following the positive shocks in ol and commodities markets. Adverse and persistent-terms of trade (TOT) shocks from the mid-1970s fiurher drained foreign exchange reserves in many SSA countries, bringing their exchange rate regimes under pressure. In most cases, the overvalued exdhange rates were defended by the rationing of foreign mThe term unificafion in the SSA context refers to eradicaon of the parall nrket. However, since these counries ac likely to maintain capital contols in the mdimn term, there would remain a small role for the parl market in meeting portfolio demand- Our concept of unification in SSA is thus a substantial reduction of the panre markct so that it is no longer a major signal in the economy. -2- exchange, and more stringent capital and import controls were imposed. One consequence was a burgeoning of illegal, parallel markets in foreign exchange. Parallel markets in SSA have shown a tendency to expand rapidly. In contrast to most Latin American countries, where small quasi-legal black markets operate as a sideline to official multiple markets with premia typically below 60 percent, illegal black markets in SSA are major markets, with substantial premia. Of the ten SSA countries considered by Kiguel and O'Connell (1992), the average parallel prenium for 1970-89 exceeds 90 percent for six countries, and 150 percent for two countries. Since the 19S0s, the current and long-nm costs of multiple markets for foreign exchange have been increasingly emphasized by multi-lateral donors and governments. The negative macroeconomic implications of official multiple markets (the Latin Ameican case) largely hold for official/unofficial multiple markets (the SSA case) (Kiguel and O'Connell, 1992). However, the African case must be disdnguished by the prominent role for illegal markets. The large size of unofficial markets and of premia Leans there is considerable diversion of trade and productive activity into ilegal channels, with a narrowing of the tax base. Domestic prices may lagely be detrmined in parallel markets and strongly influenced by changes in the parallel premium. Flight from domestic currency reduces the savings base for domestic investment, presenting a severe constraint on long-term growth. Another consequence is wide-spread speculative activity- The is clearly a strong case for the elimination of parallel markets. The question remains how a sustinable unification of multiple markets might be achieved. Flexible exchange rar arragements to this end have chatrized most I.M.F. and World Bank- supported programs since 1983 (Quirck et al, 1987). These fal into two categories: interbank markets in foreign exchange,' and a number of innovative eximents using foreign exchange auction Markets.3 Potentially these mark-de d exchange rates hold a nUmber of advantageS for counties with thin foreign exchange markets over non-market alternatives, such as the crawling peg They wre efficient in rationing foreign exchange and reduce manipulation in dtin markets; can allow a more certain path for reserve management; avoid the choice of the appropriate exchange rate, and can serve to distance the government from the political costs of devaluation (Krumm, 1985). Auctions may hare advantages over interbank markets where there is insufficient institutional depth to allow effectiv functioning of a decentralized foreign exchange market, where a few commercial bankcs have historically been domiant and there is a danger of collusion, or where there. are linited sources of foreign exchange.4 Given the stmctural characteristics of Sub-aharan African countis, therefore, auctions potilly have an important role to play in exchange rate unification, Counties whih have used interba mark inlude Zai. Th Gmbi, Siera Leone and Nigeria (in tandem with an auction). 3 Courics which have estabihod various types of axchange rate auctions incude Bolvia (1985 onwards). Jamaica (198-89) and Afiican counies, Ghana (1986-92). Nigeria (1986-94), Guinea (1986-), Zambia (1985487). Sierra Leone (1982-13), Uganda (1982-15, 1992-93) and Ethiopa (1993-). 'There are disadvantages too i e incrased admin requiremas of a ctalized system, ic potetial for govenment mnulati of the excuhge rate or fr rest-seeking, and a Closer Lssociation of the governme to the politics of the cxchage rate.. primarily as a transitional device to an efficient interbanc nmarket.A Yet to date there is little understanding of the functioning of auction markets ia Sub-Saran Africa, and there has been virtually no research on the causes of freWent policy reversals or auction failure. Apart fron supply problems due to initial conditions of market thinness and vulnerability to terms of trade shocks and fluctuations in the disbursement of foreign aid, poteial causes of failure include macroeconomic laxity, inapropriate auction design and poor micro-magement of auctions. 1.2 The Rationale for Transitional Foreign Exchange Auctions in SSA The principal objective of exchange rate libc.-aisation in SSA is unificaion of multiple exchange markets into a single market, or at least to inrrow significandy the gap between the rates, so that the free/black market premium no longer constitutes a major signal in the economy. This objective of a realistic exchange rate which reflects market conditions, inially has far greater priority dt a second objective, that of stilizing the echange rate. Stabilization is crucial in some Latin American economies, where the parallel market is small relative to the official market, and where the fred-through coefficient from official devaluation to inflation in near hyper-inflationary conditions is very high (0.6 for Bolivia (Dominguez, 1991)). In SSA many prices already reflect the parallel exchange rate, so that official devaluation is unikely to have a significant effect on inflation (Aron and Elbadawi (1992)). Stabilization in SSA wil assume increasing m e as real depreciation is ahieved. Neverteless, reducing voaLility in the transitional period is also desirable for all the usua reasons: in SSA, exchange rate volatility has been shown to be the most important deerrent for intors. Volatility is in part related to a limited supply of foreign exchange, which is based on highly concenated exports of primary commodities, vulable to TOT shocks, and on donor aid, also subject to shocks.' A third important (and perhaps underesdmated) objecdve is the restructuing of domestic production dtrough a market-determied allocation of foreign exchange. Allocation accordig to competitive market valuations should allow scarce foreign exchange to be utilized in an efficient manner. Damaging distorons were introduced by the praice of ma-ua allocation, with fvoitism and incentives for wastefid rent-seeing activities. Moreover, the sem pepeated large, inefficient parasttals, which have proved a drain on scarce foreign exchang and on tb's budget Given the rudimentary nature of many markets in SSA, and of institutions such as the baning system, auctions potendally have an important role to play in the establishment of deeper and more stable markets in the transition to efficient, ised inbank markets. Auction markets have two advanges olrer interbunk markt in constaining volatility in thin mkets. Fist, they may be feasbly opeted in conjunction with a stabilizaion fimd, which can commit supply in the medium- term and still offer a market-based rate. Secondly, auctions are in ionly-rich, and this may be 5 This is how the reltdiy succei and lengthy uificato expmea in Ghan pgremsed 'Desabising capial flows a cnst id by d capital contols prese in tact SSA counties and arm lily to be maintained for the modium-erm. Thee ay be consideable outflow i cail fligti, howevcr (rea and Elbadtaw, 1992). -4 - exploited by the seller in two ways. By appropriately revealing information or adjusting the reserve price and supply, uncertainty may be reduced, speculative pressures stemmed and collusion limited. hence stemming volatility. On the other hand, auctions give the seller information about private agents' valuations; in signalling demand pressures auctions can be used to enhance monetary management.7 This- was found to be a principal advantage of the auction in Bolivia (Domigue, 1991). This macro-micro interaction resulting in improved monetary management also enhances the credibility of auctions. An auction system has significant advanges over the interbank system in allocation of foreign exchange in thin markets. There is no reason to assume that a fair and efficient allocation will occur where predominantly state-owned banks or a highly concentrated commercial banling sector allocate foreign exchange to final customers (after a wholesale auction of export proceeds, or a decentralized purchase of export receipts). Indeed, practice shows otherwise -in a concenatd sector, the banks possess considerable allocational autonomy, with favored customers and bank shareholders first in line. Auctions offer the possibility of restricting entry through a set of rules which reflect development priorities, but stl maintain market-allocation. These rules can be liberalized over time. Thus, for example, foreign exchange for lumiry goods can be limited to repatiated flown capital. The problem of an asymmetic set of importers, diffiated by credit-worthiness and production costs, could be addressed in a number of ways, for instance by clatve caps on allocation for very large users of the limited supply of foreign exchange, at least for a tansitional period. Thus, the significant advantage of auctions in SSA is that they offer an informaionaly-rich frework and the possibility of gradually liberalizing auction restictions in the progression toward deeper and more competiive financial markets. Unfortunately, associated with this strength is the license it offers goverments to manipulate an auction and to engage in ad hoc interventions. This is the main reason for the current disaffection with auctions amongst some within the donor community. This point has to be addressed squarely, and we believe that there is scope to design trsparent auctions which reduce manipulation by the seller- The structure of tAe rest of the pape is as follows. Section 2 gives a typology of foreign exchange auctions in the four countries. A detailed statistical analysis of micro-auction data from the above four auctions is contained in section 3. Section 4 conains a brief macroeconomic interpretation of the auctions' outcome. Section 5 concludes. 2. DESIGN CHARACUERISCS OF AUCFIONS IN SSA The design characteristics for the Zambian, Ugandan, Ghanaian and Nigerian auctions are summarized in Table 1. These case-studies present a wide specum of auction designs and outcomes 71t might be argued that a black marke or reign cxchange bureaux markt offes the sam signallig advanages as an auction. This is false: these markm are vey hin markts awd rec the influence of remainig tad restitions and exchangc conols, as well as vaous demand prssur not constmt wih development priorties, such as smuggling aivts. -5 - for cross-country comparisons. In three of the countries, different auction types follow consecutively, allowing within-country comparisons of auction design. in this section we compare and contrast auction designs and outcomes across countries, and motivate hypotheses concerning the effect of auction design on the achievement of policy objectives such as stabilization of the exchange rate, and unification of multiple exchange markets. Some of these hypotheses will be tested in a second companion paper (Aron and Elbadawi, 1994). 2.1 Duration and frequency of the aucfions The various auction experiments have spanned widely different periods. In Ghana, weekly auctions of varying design were held over almost 6 years, and then replaced by an interbank markiet, with decentralization of the sale of most export receipts. The same result has been achieved in Uganda, although after a considerably shorter period of auctions (about 21 monits). In both these countries the auctions proceeded fairly smoothly, and the rules were gradually liberalized over time. By contrast, an IS-month experiment in Zambia with 68 weekly auctions, saw restictions tightened over time and frequent rle changes. The auctions were suspended for three months; subsequently a further 6 auctions were held, but a speculative run on the kwacha saw the system abandoned, with a return to a fixed exchange rate regime. In Nigeria, auctions have similarly yielded mixed results, and bave been subject to considerable rule changes. Auctions were held over six years, weekly for the first 67 auctions, during 1986-88; then daily during 1989-90, in conjunction with an interbank market; a weekly system was again initiated early in 1991, and continued untdl early 1994, when the exchange raze was fixed. The competitiveness of the 1989-90 auctions in Nigeria is in doubt: participants were guarateed a mxmum aDlocation regardless of their bids. We therefore examine only the 198648 periodL There may well be a relationship between the duration of the auction regime and the success of the outcome, though this will obviously depend on the initial depth of the financial and foreign exchange markets. Admirers of the Ugandan and Ghanaian outcomes suggest that a slow and unambitious start, tgther with gradual expansion of the aactions through liberalization of the rdes, served both to facilitate institution-building and learning, and to enhance the credibility of the reform. However, auction stability in these countries is also importantly related to other design features, such as tansparency and the use of a reserve price, both of which were hallmarks of the Ugandan and Ghnaiaa auctions. A higher fteucn-x of auctions would seem desirable since this could smooth disbursement, make fims' production more flexible, and probably encoura investmenL Furte, the speculadve behavior or over-bidding that might occur in more sticky markets would be curmiled. On the other hand, a high equency auction market may be so tiin that the marginal price would probably not be a useful guide to the exchange rate, and would in any case be rather volatile between auctions. Further, there may be increased inentives for collusive behavior amongst bidders. For the SSA auctions under consideration, weekly or bi-weekly frequencies were employed, and in general the demand for foreign exchange exceeded the supply in fairly deep markets. -6- 2.2 PricIng Two types of pricing mechanisms have been used in these auctions. These are the discriminatory or Dutsh auctions, where bidders pay their own price for each uit; and comvstitive auctions, where bidders pay the lowest accepted bid price for each unit. The clearing rate for all the auctions was defined as the marginal rate at which the available supply is exhausted, when ordering the bids from the highest to the lowest price. All four countries used the Dutch action, although the first auction in Ghana used competitive pricing, and Zambia and Nigeria preceded their Dutch auctions with, respectively, 42 and 24 competitively-priced auctions. The stated rationale for the choice of Dutch over competitive pricing was two-fold: raising revenue through discriminatory pricing; and the disincentive toward depreciadon thought by some policy-makers to be embodied in the Dutch pricing. While there is no theoretical justification for the latter view, the former proposition has proved controversial in the auction literature (e.g. the U.S. Treasury Bill auction debate). There are also potenial disadvantages to Dutch pricing: first, if there is a large spread between bids, this may be constued as conitutng a multiple exchange rate system, with the attendant disadvantages (Quirckl 1987); and secondly, a smaller pool of bidders may ensue because the Dutch auction introduces a barrier to entry for risk-averse bidders who are poorly informed about market developments (Goldstein, 1962). Theory predicts that Dutch pricing lessens collusion (Robinson, 1985); but some authors are of the view that through a narrower range of bidders, Dutch pricing may also encourage collusion (Quirck et al, 1987). Given that Nigeria and Zambia included both types of auction, some of these issues could be subject to empirical verification. The official buy and sell exchange rates for all these countre were closely based on the clearing rate. Where an interbank market operated in oonjunction with the auction (e.g. Nigeria), the buylsell rates in this market were constained to lie within some margin of the auction rate.8 2.3 Competition and the nature of bids Two types of sale have been employed in the SSA foreign exchange auctions. In the eail aucdons, the bidders were private and public sector importers, and bids were channelled through commercial banh. In some cases the mues were liberalized to allow the banks to compete for their own requireme. This system was used in Zambia, and for 174 auctions in Ghama. The Central Bank opened sealed bids for each importer, specifying the quant of foreign exchange desired, the price that the importer was willing to pay for it and the intended usage of the foreign exchange. These auction markets attracted between 100 and 600 bids,th the potential for a competiively-determined exdchage rate (the numbers of bids are shown in Table 2). By cont, in the wholesale auctions used in Nigeria and Uganda, and later in Ghana, the bidders were registeed banks or foreign exchange dealers. The system adopted in Nigeria initially allowed an unc I secondary market for foreign exchange, so that banks competed freely in the action for later resale to importers in the interbank market (see Table 1). Each bank submitted l Ua rse price is usd (see sub-setion 2-5). a quantiy-weighted average of the reserve prie and the bids above k coud detemine the echange stte. This was the pratice in BolivaL -7- one bid. In Uganda and Ghana (1990-92), ttie banks were constrained to submit aggregated quantity bids at the various prices bid by importers (and possibly themselves). The number of bids opened by the Central Bank to determine the clearing rate in either case was thus far smaller than in the retail auctions (Table 2). The danger in the Nigerian scenario, is that allowing banks full autonomy in bidding for and then dispensing foreign exchange in a secondary market, could, in concentrated financial markets, lead to collusion. Banks could offer bids below their own value of the foreign exchange, but resell at the markt price, thus widening the spread between the buy rate (i.e. the auction rate) and secondary (interbank) market sell rate, with rents accruing to the banks. This appears to have occurred in Nigeria, obviously defeating the objective of unification. Moreover, it is not obvious that the subsequent allocation to final customers would be optimal: indeed anecdotal evidence from Nigeria suggests that the bank's shareholders and favoured customers had privileged access to foreign exchange in the secondary market'. Auction theory advises tha auction managers employ a reserve price (below which foreign exchange will not be sold) to combat suspected collusion (McAffee and McMillan, 1987). A reserve price was not employed in Nigeria. Under the strict monitoring of the wholesale system used in Ghana and Uganda, the low mnmber of composite bids need not have affected the potential for a competitively determined auction rate. Nevertheless, the possibilities for collusion do seem to be greater, given the reduced wanarency of the system. Both Uganda and Ghana used an unnounced, though fairly predictable, reserve price, around which the bids cluslered over time. Thus, the clearing rate was influenced with reference to some target, so that the system resembled a crawling peg, though with a mare (not manual) allocation of foreign exchange. This convergence of bids (see Table 2) oould either have represented the bidders learning the reserve price rule over time; or possibly collusion amongst bidders. The reserve price is discussed further in sub-section 2.5 below. 2.4 Entry and allocative restrictions A limited supply of auctionable funds, and significant excess demand has in practice been used to justify entry barriers on grounds of allocation. Typical entry barriers are: confining the use of foreign exchange to particular sectors; confining the use of foreign exchange to particular qtpes of imports (typically exduding luxry goods); and limiting participation by commercial banks, licensed foreign exchange bureau and the state-owned sector (e.g. with ceilings on their maximum allowable allocation, or complete exclusion from the auction). To enforce these various exclusions, bidders are monitored dirough documentation requirements, such as evidence of paid up tax, deposits with the commercial bank filing the application, and import and export invoices. An unintentional entry barrier has in some cases been posed by the varying credit-worthiness of bidders in imperfect financial markes. All the SSA auctions required a local currency cover for bids, and some smaller companies apparently found their access unfairly restricted (e.g. in the Zambian and Ugandan auctions). 'The exiacnr or considerable ms in te Nigeran aucdon is reflected in the prolifastion of licanced banks, which incased in nmber from 41 in 1986 to 112 by early 1991. -8 - One trade-off with entry barriers, Is that transactluns costs for bidders In terms of the time involved in preparing bids may impose a disincentive to use of the auction, and it is obviously important to streamline the. license and bank applications procedure. It also seems clear that efficiency would be enhanced if there were reduced sectoral constraints. On the other hand, the non-eligibilt of some consumer and luxury Items seems reasonable, at least in the early stages of an auction regime, when it may t.l vulnerable to speculative pressures. Another problem arising with entry restrictions is that where barriers are not clearly defined, they effectively decreae transparency in the auction, granting license to the authorities to m"nipulate the set of bidders. In Zambia, disqualifications -.ere ad hoc at various times in a deliberate attempt to suppress depreciation. Apart from the non-market pricing and allocation this may induce, the perceived credibility of the reform may be jeopardised (see also sub-section 2.6). Expanding an auction through the liberalization of various barriers to entry will increase competition, and therefore bid prices, as importers previously satisfying needs on the parallel market or buremau d change (or not at all), move into the auction. While the auction rate is expected to increase, the volatility of the rate over time may decrease, due to a more competitive and credible system. Auctions where entry restrictions were increased over time (e.g. Zambia and Nigeria) seem to have traded a less depreciated real rate for increased volatility and a larger black market premium. A paramount reason for the success of the Ghanaian auctions was a gradual widening of the auction system, combining trade and exchange rate liberalisation. This served to increase competition and foster confidence in the system, deepening fincial markets by allowing institution-building over time. A t-tier system operated for about six months: when these windows were unified in early 1987, all transactions were settled through the auction market. Access to the auction was widened thereafter through a step-wise inclusion of additional goods and services, while import resticto were relaxed. By early 1988 virtually all consumer goods were eligible, and import Licensing finally abolished by early 1989. Current invisible payments were liberalized over the same period. The foreign exchange bureaux were legalized early in 1988, and late in 1989 allowed to bid in the auction with the commercial banks, on behalf of final users. At the same time, decisions concerning the eligibility of auction bids was transferred to the commercial banks and eligible bureaux. Finally, the sale of non-cocoa exports were decentralized t the banking sector in early 1991. These moves paved the way for the transition to a less concentrated interbank market, less liable to collusive practices. Uganda similarly successfully liberalized the set of restrictions on impors and bidders over time. Finally, an important question is whether the state sector should be included in the auctions. It would be naive to ignore the political economy of public enterprises. With fixed exchange rates, and manual allocation of foreign exchange, parastatals are the principal beneficiaries of exchange rate rents. As large employers, parastatals have a powerful political voice; they may also impose intrnal party pressures on the government (as occurred in Zambia). This suggests that shock therapy resulting in the rapid demise or even bankruptcy of parastatals is to be avoided. In anycase the negative impact on the budget of such an event could be significant. On. the other hand, there are significant allocational gains to be had from not perpetuating such a system. In Uganda, Ghana and Nigeria, state enterprises competed in the auction and received no prior allocations. With the change of the Nigerian -9- auction system in 1989, however, public enterprises again had direct access to foreign exchange though the Central Banol In Zambia, the largest parastatal by far, the state copper nmining company, received allocations outside the auction. Oher parastatals were included from February, 1986, but the heavy users of foreign exchan, the oil and airways ste companies, bid competitively for only a few auctions, and were then allocated on average 40 percent of the auctioned foreign exchange at the marginal rate. Thus, a crucial determinant of supply to the private sector was the privieged allocation to the state sector: only about 25 percent of total foreign exchange inflows were actually auctioned in Zambia during 1985-87. 2.5 Informational rules There are three ways in which information could be revealed by the auction managers: first, the use of a (fairly predictable, or possibly pre-announced) reserve price, which reflects the authorities value of the auctioned foreign exchange; secondly, pre-announcement of the supply of auctionable foreign exchange, or commiment of supply over the medium-term; thirdly, publication of the nature and/or number of biddersfcomposite bids. These are considered in tiU. Auction theory recommends the use and pre-announcement of a gay_pr_ydm particularly where bidders' valuations of foreign exchange are "affiat (where one bidders's high value makes more likely a high valuation in other bidders)- There is probably an element af affiliated values in foreign exchange auctions, and it is then to the seller's advantage to provide information about the value of the auctioned item The authorities' value of the auctioned item may be fairly easy to assess if there are well-developed secondary markets for it (e.g. as for some Treasury Bill auction markets). Parallel foreign exchange markets Oegal or illegal) were present for all the SSA auctions, but except in the case of Nigena these were not resale markets. hes parallel markets, at least initialy, were typically thin and restricted markers: the black rate, in particular, being su bject to various influences making it an undesirable anchor (see Aron and Elbdawi, 1992). Uganda and Ghana leglsed the parallel market (as the bureaux market) and encouraged its deepening over time. nterviews have suggested that these bureaux markets were used as a guide in reserve pricing. Both tee counties used unannounced reserve prices, while Zambia and Nigera did not use reserve pricing. The stabiizing effect of such a 'crawling peg" system is apparent from the Figures la-d, and contrasts with erratic behavior of the exchauge rate in the other two coutries. The dramatic fail-off of the bid-spread over time inthese two counties (Table 2) suggests that bidders may have learnthte reserve price policy rule, and bid more efficiently. If a reserve price is used, there may be a transpany advantage in announcing it each week just before opening the bids.'° There may be similar stabilizing advantages to pr-ouncing supply and sticking to it. Pre- announcemet Gust before opening the bids) occurred for all auctions in Nigeria. In Zambia, a geneal Twis s the practic in Bolivia: the tmvec price and supply were announced only after te sealed bids were in- kt is important to note though tha only in 1071107 days did demand exceed supply (the avenge dal aucton supply over fivc ycarn was $SA million). This rate contasts with the SSA auctions, where excess demand was nmsly the order of the day. Further, with smal parallel markets and near hyper-inllation gabIztion of the rate. ratha than unification, was Bolivias primary objective. 10- ceiling of $5 milion was declared for auctions 1-19, raised to $9 million for auctions 20-52, and dropped to $5.5 million for auzcdons 53-68; but except for auctions 55-68, this bore little relation to actual supply (Table 1). Both Zambia and Nigeria were subject to sharp appreciations through periodic massive disqualifications of bids. Supply was not announced in Ghana, and was volatile (Table 2). Supply was naounced only just prior to opening the bids in Uganda, but while pre-announced supply was fairly constant, actual supply was fairy volatile (Table 2). It seems likely that an uncerain supply could foster uncertainty and place a premium on the amount bid, in the absence of a predictable reserve price- It would thus be interesting to test the impact of supply volatility on the behavior of the marginal auction rate. The publication of bids occurred weekly in Zambia, listng each importer's bid, by quantity, price and use, for auctions 37-68. Uganda published the list of composite bids by price and quantity, idiucating disqualifications. Nigeria and Ghana did not publish details of the auctions. Whether the spectrum of bids should be published has no clearcut answer: it may crexte confidence thrugh trsparency. aad enhance learning, which were the intentions in Zambia. On the other hand, when the number and nature of bidders becomes known, this may foster collusion in repeated, sequentil auctions. Collusion is less likely to manter in an auction with a large number of bidders, however. -2.6 Tnsparency An impornt lessonfrom the Zambian auction is ta if there isn't transparency inthe auction, it is probably poiless holding one. The degree of trmsparency of rules and management differed across the auctions. 'ransparency was poor in Zambia and Niria Ad hoc intrventions and frequent rule changes by auction managers were an important cause of instability, fosering uncertainty, incredibility aDd resulting in poor (non-market) allocation. The auction rules in Zambia actuly stated that the Central Bank had license to disqualify any bid without having to give any reasons: this was a serious design fault." Auctions in Ghana and Uganda were operted with tanarently, with clear ndes, and the public opening of the bids. The management committee in Ghana did not only consist of Central Bank personnel, but included representatives from the private sector. There was fairly broad consultation on changes in the Cental Bank's regulabons for the auction, which were published as amendments. 3. A STATISTICAL ANALYSIS OF MICRO-AUCTION DATA IN SSA This section examines the evidence on the evolution of the auction rate and other auction variables for each of the four African countries. The objective of this section is to use h cs "This was exercised in auction 41, for instace, wher all bids largW than 8 kwachaiS were abhrarily declared to be d fied '2n interiew at the I.M.F. in I993 revcaled thl foreign xchange auctns in Ronaniad Klaaktn were at hat time less thn anspar, and tbat this ws pving debitadng to the fuctioing of the aucions. - ~~~~~~~~- 11 - of auctions data to draw some empirical regulaities: (i) to account for the behavior of the auction rate L terwm of the impact of various fudamentls, as well as policy intvention, policy credibility and uncruinty; (ii) to atempt to link particular data characteristics to different outcomes, regarding the auction objectives (e.g. exchange rate unification); and (iii) to inform a futher and more rigorous analysis on testing theoretical propositions firom auction theory (Aron and Elbadawi, 1994b). The analysis of this section is centered around three broad methodological issues. First, basic summry statistics for each country will be analyzed for the overall auction period as well as for the episodes identified in section 2 above. Second, the distributional properties of the auction variables will be studied, focusing on the extent to which innovations from these variables exhibit departures from non- normality. Third, the section condudes with a discussion of stationarity issue of individual auction or data series. 3.1 Standard Descriptive Statistics. Table 2 provides basic summary statistics of auction data from Ghana, Nigeria, Uganda and Zmbia, on prices, price spreads, foreign exchange quani,ties1 number of bids and bureaux or black/auction prema. The results broadly corroborate the analysis of the typology of auction designs and policies in the four countries. A summary of the evidence follows. Auction prices (auction rate, maxmum and minimum bids) have been fairly stable in Ghana and Uganda, while in contrast they have been rather volatile in Nigeria and Zambia. For example, the range of auction rates (in terms of domestic currency per US$) was 970-1042 in Uganda and 317- 395 in Ghana (in the second auction regime),'3 compared to a range of 3-5 for Nigeria and 5-25 for Zambia (also see Figure 1). Stability of auction prices has improved around a depreciating tend in Ghana as the auction progressed from a retail to a wholesale, auction while volatiit increased in Zambia with the change from competitive to Dutch pricing. In Nigeria, however, auction rae volatiity has declined, pehaps due to the stabilizig effect of the steady foreign exchange supply. The average max-min bid spread for example declined in Ghana and Nigeria, in contrast to Zambia. Except for Nigeria, for which the foreign exchange supply has been between US$50120 million, foreign exchange supplies have been rather volatilel4 ranging from minima of USS (million) 2.60, 0.33 and 0.85, to ma xima of USS (million) 20.80, 3.48 and 18.48, in Zambia, Uganda and Ghana, rspectively. in all of the four countries foreign exchange supply has been almost always less than demand, but the evolution of this ratio over time has been yet another dcractrstc of the foreign exchange auction seting Ghana and Ugmada apart from Nigeria and Zambia. The supply demand ratio in Ghana increased from an average of 0.85 during the retail auction to an average of 0.87 during the wholesale, while the ratio remained quite high at an average rate of 090 in Uganda. On the other 13 In fact the auction rate was quite volaLi in Ghan during the fist rtail auction. where the maximum sct was 314.0O relave to a minimum of only 128.00. 4"The level non-sftionarisy of the variable involved, especially the aucion prics. imis options for the use of standard devation or coefficient of varanon as measues of volatilty. For te kvel vanbles in Tabk 2. the volatiliy measure i-s the diflfrenee betwen. the aveage minimum ad mamum viabes. hand, demand rationing increased in the other two countries as their auctions changed from competitive to Dutch pricing, with supply ratios declining from 0.90 0.87 in Nigeria, and from 0.68 to 0.45 in Zambia. For the case of Zambia, the same story can be told for the winning to total bids ratio, which declined dramatically frm an average of 0.65 to 0.42 between tde two auction regimes. This is a direct result of the increasing frequency of interventions and bids disqualifications that took place in the Zambian auction. Interventions/disqualifications have also taken place in Nigeria; unlike Zambia, the bid total excludes these disqualifications, so that the successful ratio is high at its average of 0.86 in the competitive auction, and 0.89 in the Dutch. On the other hand, the same ratio remained high at 0.92 on average for Uganda, and was lower * 0.78'5 over the retail auction in Ghana. The blacklauction premium rose from 1.33 to 1.59 bween the two auction pricing regimes in Zambia, while it almost remained constant ([.36 to 1.33) between the two similar regimes in the Nigerian auction. On the contrary, the buremux/auction premium declined dramatically in the Ghanaian aucion from an average of 1.31 for the retail regime to only 1.06 for the wholesale. For Uganda the premium rates averaged 1.19 o 1.24 overall. To conclude, it is important to emphasize that the increasing (or constant) and high premia that prevailed in the Nigerian and Zambian auctions, are the uldmate outcome of the above auction design!policy mistkes in these two counries. The frequent and some imes arbirary interventionsldisqualifications have fostered an atmosphere of policy incredibility, lack of transparency and collusive behavior. In addition, the absence of a reserve price policy in the two, countries fiurder compromised stability of auction prices. For a reserve price to be sustainable, however, it has to be consisten with a stable relevant macroeconomic signal, such as the parallel or bureaux rate. Given the rather unsable macoonomic environmet dtat prevailed in these two countries during the auction experment, both the bureaux and the parallel market could not constte efficient secondary markets. These are exacdy the problems that the Ghanaian and Ugandan auctions seemed to, have avoided, hence their success at achieving exchange rate unification-the key objective of foreign exchange auction in SSA. 3.2 Normality iss Given the results on the non-stionarity of most of the auction data (see sub-section 3.3 below), the analysis of ihis sub-section will be based on the rates of changes of the variables in question (A;xt = logX4-logX. I) The departure from normality is another characteristic freqently observed in economic data that has important implications for the analysis of variability and uncertainty. In a recent paper, Pritchett (1991) found that the time series data developing countries real exchange rates tend to be characterized by excess krtosis (relatively large probabilities of exeme observations) and skewne (tendency of extreme observations in one tail). Pritchett's results suggest that when there are significant higher order moments (skewness and/or kurtosis), the standard t Data on the number of bids as not available for the Gha whoksale auction. - 13 - deviation measure of variability may not be an adequate proxy for uncertainty. As we will see below, the analysis of this subœection will lend firther evidence on the differentiated data characteristics between Ghana and Uganda on one hand and Nigeria and Zambia on the other. Furthermore, given the importance of uncertainty (of foreign exchange supply, for example) for the analysis of bidding behavior in the auction (Aron and Elbadawi 1994b), the ts for normality is crucial for this study. Separate skewness and kurtosis tests for the rates of changes in auction variables are provided in Table 3, two for skewness (b3 and s) and one for kautsis (b'16 The fi skewness ts is the scaled third moment of the distribution about the mean scaled by the standard error, the second is the standad deviation normalized difference of the maximum rate of change minms the absolute value of the rate of change. Values less (arger) than zero indicates left (right) skewness. The excess huois coefficient is given by the ratio of the fourth moment to the square of the second moment minus 3. A value significandy dfferent from zero for this statistic indicates a kurtotic underlying data generating process. Critical values for each of the three tests are derived under a null of a normal distnrbution. Table 3 also contains a column on standard deviations and two more columns (s2 and s3) giving the ratio of the value of the maximum (and the absolute value of the minimum) rate of change relative to the absolute value of the average change. The results of Table 3 show that the distribution of the auction rate is markedly asymmetric in all of the four countries, with a pronounced tendency towards large depreciation (right skewness) in the cases of Ghana and Uganda, while in the case of Nigeria and Zambia a significant tendency towards appreciation (eft skewness) is observed (see also Fig. 2). For the cases of Ghana and Uganda (Nigeria and Zambia), the null hypothesis of a symmetric distbution against the alternative of right (left) skewness could easily be rejected at conventional levels, using both of the two skewness tests. This findin on the differentiated distributional cteristcs of the auction rate between Ghana and Uganda on one hand and Nigeria and Zambia on the other, is fairly consistent with the auction designs and outcomes in each of the two groups of countries. In the first two counries the reserve price effectively ruled out exchange rate appreciation, while on the last two the absence of a rese pnce and the frequent disqualification of bidders and the subsequent reduced effective demand for foreign exchange within the auction have fosed a tadency towards exchange rate appreciation. Also in terms of auction outcomes, the tendency towards depreciation in Ghana and Uganda appeas consistent with the success of these two countries in closing the auction/bureaux rate premium (unification), while the opposite happened in Nigeria and Zambia. As with the auction rate, te distributions of the bureaux excange rate and premia in Ghana and Uganda also exhlbits departure from symmetry, albeit witi a tendency towards appreciation (left skewness). This behavior is consistent with a fiequendy, sometimes significantly, declHning premium, Pritfchct noted that in smal sampes it is difficut to distinguish betwe skewnss and kurtosis. given tat a hi kinodc, the criiad levels genated for skewne tess assu5ming a N(0,1) will t be corrt and the test wil have a tendency to rica the null of zero skewness in this case. Pritchtt prvides critical values lbr skewness tests under the null of a moze geneal symmetric, but kuhtotic distribution (the student t- distributn). The tes of table 3 abstact from this compfication. However, the number of obserirtios avaiable to us is much lrger than do Pntchs. - 14 - as the depreciation in the bureaux rates began to fall short of the auction rates depreciation. Futhermore, the bureaux razes seem to be frequently bat only incrementally depreciating, while adjustment to previous overshooting (excess lagged depreciations) or movements towards a lower longer-m equilibria have caused substantial and not infrequent appreciations.17 On the other hand, the two variables (black market) have bad a significant but rightly skewed distribution in the case of NigeriL The tendency towards a rising premium and depreciatng black rates could be explained by a combination of two factrs: an unstable macreconomic environment leading to sustained black rate's depreciation, coupled with auction policies that limit depreciation of the auction rate. For the case of Zambia, no significant departues from symmetry could be esublished for either the premium or the black ra-"e' In Ghana and Uganda, the distributions of supply variables are characAized by significant right skewness, while demands have been negatively skewed. This implies that in order to ensure convergence of the auction rate towards a unified target rate, the auction foreign exchange supply has been successively adjusted upwards, in the face of declning demand, as the auction rate becomes more deprecated. As with other auction variables, the story is different for the cases of Nigeria and Zambia. The supply and demand variables do not show signficant depure from symmetry for Zmbeia, while only die demand variable exhlbits sigaificant left skewness in the case of Nigeria. The latter appears consistent with the frequent recourse to the policy of disqualification of bidders noted for Nigeria (section 2). The final issue on symmetry relates to last two column of Table 3, which gives the ratio of the maximum (and the minimum absolute value) of the rates of change relative to the average of the absolute values of the rates of chane. For the aution rate, for example, this indicates how much larger the biggest depreciation (or the absolute value of the biggest apprecation) is than e typical raze of dange. The results show rather dramatic orders of magnitud on the deparre from symmetry, especially in the cases where deviations fiom symmetry were found to be significant For exampe, the maximum appreciation (si) of the auction rates are more than seven tmes the average absolute rate of change in Zambia, and more than six times the average in Nigeria. On the other band, the maximum depreciations (si) of the auction rates are more than eleven fimes the average absolute rate of chage in Ghana and about nine times the average in Uganda. The other aspect of deparue from normlity is provided by excess kurtosis, where the underlying distibution is more peaked and fatter tailed dtan the norma, implying a larger relative pobabiity of extreme observations in both tails. The evidence on this score is even more overwhelming. Except for the ditributions of demand and the premium in Zambia, the null hypothesis of a zero excess kurtosis could be rejected in all other auction variables in all of the four -7 S'c the burmux (black) rtc is subs influenc by aicution of futum auction systen libealisatiR changes may be appat in the rate befdc they actuaUy ccwur. Thbis may be cxplained by dte faict only monly black market dat was avalable for Zambia- lDh reason is ta total demand is repotd, which cludes bids that tl disquarlified - 15 - countries. These results provide additional evidence on the prevalence of departure from normality in foreign exchange auction data in Sub-Saharan Africa. However, in this case auction data could not be differentiated between the two groups of countries. Concerning variability, it has been pointed out by Pritchett that when there is significant deparure from normality, the standard deviation may not be the appropriae indicator of uncerinty. For example the rates of changes in the auction rate and the premium in Zambia have almost similar standard deviations at 0.10 and 0.09, respectvely. However, they have two very different distributions, the auction rate has a non-symmetric and substantially left skewed kurtotic distribution, while the premium had a kurtotic but symmetric distnbution. Another point mentioned by Pritchet is that under departure from normality, there may be several but very contadictory measure of variability, none of which has a clear advantage over die others. This point is validated in our case by comparing the standard deviations of the auction rates in Nigeria and Zambia which are very smal at 0.08 and 0.10, respectively, to the ones for Uganda at 1.98 and Ghana at 9.91. This, however, -conzradicts Figure 1 which shows that exacdy the opposite happens.? 3.3 Unit Roots, Sructural Shifts and Stationarity Issues. The hypothesis of the unit root (loosely speaking, this refer to a time series wi an infinite variance at the limiit) in the individual economic time series data has geeratd considerable attendon and acceptability in recent empirical and theoretic,' economic research. The seminal paper by Nelson and Plosser (1982) which found that most macroeconomic varibles have a univara time series structure with a unit root, and the more recent path-breaking work on cointegracion theory9 (Engle and Granger, 1987) has provided the impetus for a burgeoning research programn that has already started to have far-reaching implications for both theory and empincal work in economics. The most important economic implication of the unit ro hypothesis is that random shocks have permanent effects on economic variables. In our context this means that the fluctataions that chracterize the auction variables are not ranitory. Therefore, it is important to conduct careful testing of the unit D Perhaps this contradiction coud bc resolved by using the coefficent of variation rahe than the staniard deviation. 'Formally. Id y= TDt, + Z, bc an econonic series composed of a deterministic trnd TD1 and a 5ochastic component. For simplicity assune that 4 ca be desctibed by an autoregrcssiv_moving averagc process: A(L), = B(L) cj, whe A(LJ and B(L) arc polynomials in the lag operator fr and e, is a sequence of iLd. innovaton. The noise finction Z7 is asumed to have mea zero, the moving avenage polnomial is also assumed to have roots srictly outside the ni circl. Then Z, has a unit root if A(L) has one unit root and all ado roots stiey outside the unit circle In this case (I-L)Z, = AZ, is a stationmy procss and (1-Lby, = Ay, is staionary around a fixed man, If on the other hand A(L) has all its roots outside the unit circle, the Z is a statinr process and y, is stationary aound a trend- -Thc idca of conte_tio basically states that even though individual series may have a unit root, thcre may cxst9 various linear combinations of variables which are statay. Stated more formally in the context of te definigion of footnote 1, kt the n-vector y, be comnposed of (yb ... yj. whcem Ya i dcfined in footnote I above. Then y, is sai to be coinegmted if there exists at ast one n-elmnt vector 0 such that 'y, is trd sttonay. This is a mider definition of cointegrtion (Campbell and Pern, 1991), which is more suited to anablsis of eoonomic date since it permits the inion of ddernmntic components (such as tends and structural break dummics) in the cointegration model along wiith other z-stuuionary stochastic vanabls. - 16- root hypothesis for each of the auction variables of interest These tests are presented in Aron and Elbadawi (1994): in all four countries, the data series for each of the thtee exchange rate variables (the auction, bureaux rates and the premium) was found to have a unit root non-ationary process. Noting that the distinction between a random walk (unit root) and a trend-stationary series is extreme, Cochrane (1988) argues that there are many cases where a stationary but a slow trend- reverting series could be confounded with unit root series. To address this problem with unit root tests, he proposed a variance ratio test that basically asks the following question: How large is the variance of shocks to the random walk or permanent component of the series compared with the variance of the growth rates of -the series? Or, equivalently, How big is the random walk in the series? The ratio is given by l " X . For a stationary series the ratio will eventually converge to zero as k goes to infinity, the speed of this convergence reflects the extet to which the stationary time series stucture is doser to either of the two polar cases: the stationary (zero root) or the random walk (unit root). On the other hand a unit root series will have a ratio that settle down to the share of the variance of the shock to the random walk component in the total variance (> 0). Figure 3 shows the plots of the Cochrane ratio for the auction equilibrium exchange rates against k for each of the four auctions. The auction rate is much closer to the unit root case in both the Ghana and Uganda, where the shares of the random walk in total variance settle down to approximately 77 percent in Ghana and 96 percent in Uganda24 Particularly in the case of Ghana, the ratios have in fact diverged for a prolonged time before sarting to converge at k = 62. On the other hand, random walk components account for about 54 percent of total variance of the auction rate in Zambia, and only about 6 percent of total variance in Nigeria. Furtermore, convergence of the varimae ratios have strted rather rapidly (at k = 2) in both countries. This finding is again fairly consistent with the stochastic reserve price3 that has had a significant effect on the auction rates in Ghana and UgaXlda, as opposed to the frequent and direct policy inteventions that characterized the Nigerian and Zambian auctions. Therefore it is obvious that the auction rate is chacterized by high persistence (random walk component), which agrees with the unit root tests in Aron and Elbadawi (1994). But Figure?v also shows the presence of substantial temporary components in the innovation v=arices of these series, especially in the case of Nigeria and Zambia. This is also in general agreement with the prevalence of detrministic trends and regime shifts.in auction rates and auction data (Figure 1). The lawge sarmpl standard error for the ratio is given by ' tarQC_)1d) where T is the number of observations., 7'The cboice of the temminal k is rathcr arbiray. but following Cochmne we chose the maximunm k to be the ine of one Iounh to a third of ihe total number of observations. - The reserve prc closely fosilow the evolution of the burauz or paradel rates in Ghana and Ugmada which are shown to be l(l) scries Table 4). - 17 - 4. A MACRO-ECONOMIC OVERVIEW OF THE SSA FOREIGN EXCHANGE AUCTIONS. It is not our inteton to exami in any detail the macro-determinants of the aucton rate in this paper?' However, we present in this secton a brief overview of dte macro-econmic environment in dte four countries, with the objective of explaining the observed differentiated outcomes of foreign excdange auctions in the fur countries. The mac-economic sty is provided by Figures 1 and 4, fucusing on exchange rate prenia and monetary emission. The broad conclusion that can be derived from dise graphs points to the importance of a stable macro-economic envirnment for the achievement of a steady and susainable progress towards a unified exchange rate. Ghana and Uganda seem Do be largely on arget regarding the above objective. Figures 1 b and I c show a steadiy depreciatng and sable auction rate over sime, while a significantly declinin bureaxlaucion pmium is observed. Thus, tese countries have been able to achieve a reasonable degree of unification and exchange rate stability. At the heart of this outcome, as shown by Figures 4 b and 4 % is a sutined decline in mo gow This is reflected in a decelemaion in inflatio and a stable buream rate in both Uganda and Ghana, where in Ghana this folowed an episode of deep depreciation towards what appears to be an eqauibrium path Th figues also show a clear leadllag. effect going fom monetay emsion to iatim and bureaux rat depreciato This indica that not only has macro-policy been consistent, but also a it has been credible in these two countrie In contrast, in Nigeia and Zambia Figures 1 a and 1 d also show a deprecating maction rate, but with far greater volatility. More importantly, unlike the othe two countries, the black/auction premium was steadily rising. Thi oumers both the objectives of stabiit and exchange rate unification. This could be a direct outcome of utaiable macro-economic policy and the collapse of policy credity. This is crborated by Fgures 4 a and 4 . Very volatie and significently rising rates of growth of money suppy over dme are observed in both coutis. This is reflected in high inflation and blacklauction premiavariability, with exteded episodes of sharply rising prmia. The relevance of macroeconomic poliy in - -if between acdon outcomes in Ghana and Uganda on one hand and Nigeria and Zambia on the othe, is fairly consistent and closely linked with the eviden on the microeconomic dermina of die auction rate (sections 2 and 3). For example, die reserve price policy in Ghana and Uganda has been credited with fostering auction rate stability. For tese two conries to manage a scssful reser pruice policy, a susable secondary market signal is necsay. This was provided by a fairly stable bur rate, made possible%y a credible maco-enviroment These results are also corroboratd by the more fomal analrysis of Aron and Elbadawi (1994). YHowcvc. in a subsequ pmr we PIa to anays =ac-coooic nies in the coun of ueon md ccbmge rae aifimtios in SSA- - 18 - 5. CONCLUSIONS AND POLICY LESSONS. The experience of foreign exchange auctions in SSA has been one of rich and diversified design features, policy interventions and outcomes. This paper has compared and contrasted auction designs and outcomes across four SSA countries and produced a typology of SSA foreign exchange auctions. Two distnct sets of countries have been identified in temLs of design features, auction policies and outcomes. Ghana and Uganda represent a set where auctions have been judged to be largely on taget in terms of the three policy objectives of exchange rate unification, stabilisation of the exchange rate and an efficient allocation of foreign exchange. On the other hand, the auctions in Zambia and Nigeria were subject to frquent policy interventions, with the consequence of unsustainable auctions, inefficient allocation through ad hoc disqualifications (at least in Zambia), lImited unification, and a rather volatile exchange rate. Ihe condusions of this analysis are broadly corroborated by a statistical analysis of the micro-auction data for all four countries. A summary of the main stylised facts and their possible explanations foilow- First, auction prices have been fairly stable in Ghana and Uganda, while in contrast they have been rathr volatile in Zambia and Nlgeria. Stability of prices improved around a deprecating trend in Ghana as the auction progresed from a retail to a wholesale auction. In Zambia, price stability decreased from the competitive to the Dutch auction; while in Nigeria the reverse was true, perhaps due to the effect of stable supply. Empirical evidence on the depature from normality of hie differenced auction rate is provided by skewness test on auction data All fbur countries show marely asymmetric auction rates; Nigeria and Zambia have a significa tendency toward left skewness (apreciation), while Ghanaand Ugandadisplay apronounced tendency towards depreciation (right skewness). For all countries, appreciation or depreciation tends to show a very hi,gh order of magnitude for the absolute value of highest appreciation or depreciadon relative to the average rate of change. Thus, departure from symmetry is not trivial when it it is statsticaUy significant. Further, excess kurtosis is significant in all data in all auctions, which implies a high probability of extem observatio in both tails of the distrbution of auton data Kurtosis measures do not offer any criteria for differendating between the countries, but reinforce the point that the standard deviation is not a useful measure of exchange rate variability given these deparures from normality. An explanation for increased stability in our first set of countries is that Ghana and Uganda used (unannounced) reserve prices, while Zambia and Nigeria did not This "crawling peg" type system prevented appreciation in the presence of disqualifications. At the same time the depreciation path of the exchange rate was smoothed as bidders learnt the reserve price rle, which is perhaps evidenced by the dramatic fal-off of dhe bid-spread over time in Ghana and Uganda. Furhermore, in Ghana and Uganda, not only were the paralel markets legal (bureaux markets), but the macro- policies followed were relatively consistent: ese two facts allowed the bureaux rate to constitute a credible and efficient secondary market signal for the reserve price rule. It would be interesting to test the observed differences in exchange rate depreciation across regimes, such as Dutch and competitive pricing, or retail and wholesale auctions. - 19- Second foreign exchange supply was found to be rather volatile in all countries save Nigeria, where a pre-announced supply rule was adhered to. Differenced foreign exchange supply in Ghana and Uganda is characterised by significant zight skewness, suggesting at supply was successively adjusted upwards to allow convergence of the auction rate to a unified target rate, as the auctions were liberalised. The opposite holds for Zambia and Nigeria. It would be useful to ascertain whether this volatility influenced the path of thie exchange rate. The implication of the observed non-normality of the supply variable is important for its use as a proxy for uncertainty. The preceding discussion suggests that where supply volatility is high, which may raise uncertainty in thin markets, there may be considerable virtue in the use of a reserve price. -ird, foreign exchange rationing reflected in the total to satisfied demand ratio was at a high level in the Zambian and Nigerian auctions, while it remained constant and low in Ghana and Uganda. Similarly, the winning to total bids ratio declined dramatically in Zambia, while remaining at a very high average level in Uganda (data are missing for the other two countries). In Ghana and Uganda the auction systems were gradually widened, increasing competition, fostering confidence in the system and importantly deepening financial markeas by allowing insttuion-building over ime. This gradualism allowed the transition to a less concentrated interbank market, less liable to collusive practices. By cont, auctions in Zambia and Nigeria saw restrictions tightened over time in an aempt to suppress depreciation. It would be interesting to assess the impact of increased competition through an enlarged set of bidders and allowable imports on thfe level and stability of the exchange -rate. jFourh, the parallel market premium was high for Zambia and Nigeria, whilst the bureaux premium experienced a sustained and sharp decline for Ghana and Uganda over ime. In Ghana and Uganda the bureaux rates and premia exhibit departures from symmetry, with a tendency towards appreciaton (left skewness). This is consistent with a steady and incremental depreciation, and infrequent but somedmes significant appreciation as a response to past overshooting, or adjustment towards a longer-term equiium On the other hand, Nigeria displays right skewness; while in Zambia, where data only of lower frequency was available for the black market (montily) there is no evidence of skewness. The tendency towards a rising premium and depreciating black rates can be explained by a combination of two factors: an unstable macro-economic environment, coupled with micro-auction policies that atempted to curtail depreciation. The brief macro-analysis in this paper suggest that the unsustanable macro-economic policies in Nigeria and Zaribia had a major impact on credibility and the size of the parallel premia. The key macro-policy lesson is that these two counties did not make the commitment to switch the nominal anchor from the exchange rate to the money supply, despite their clmed policy objective of exchange rate unification. Onthemicro-side, the frequentand sometimes arbitary interventions/disqualificationsftered an atmosphere of policy incredibility, lack of transparency and collusive behavior. A major lesson from the Zambian and Nigerim auctions is that if there isn't transparency in the auction, it is probably pointless holding one. Ghana achieved transparency in a very effective manner: the management commtee did not only consist of Central Bank personnel, but included repr ve from the -20 - commercial banking sector and private sectr, and was chaired by a commercial banker. There was fairly broad consultation on changes in the Central Bank's regulations for the auction, and rules and amendments to rules were published. Uganda similarly operated with a high degree of trasparency, with public opening of the bids. Fijh. one final set of empirical analyses reveals the presence of considerable non-stationarity and regime shifts for most of the auction data from the four countries. This is unsurprising in view of the above stylised facts which emphasise nrle changes, the non-normality of auction data and the importance of anticipations ofpolicy changes. The non-parametric variance ratio test CCochrane, 1988) showed that the share of the random walk component in the total variance of the rate of change in the auction rate, while large for all of the counties, is noticeably smaller for Nigeria and to some extent, Zambia. The rate of convergence also strted rather more rapidly in these two countries. This finding is again consistent with the use of a stochastic reserve prioe in Ghana and Uganda (i.e. based on the non-stationary bureaux series), as opposed to the frequent and direct policy interventions that charactsed the Zambian and Nigerian auctions. These results on non-stationarity have profound implications for the way in which auction data are analysed: random shocks have permanent effects on economic variables, and consistent ing of the behavioural propositions of auction theory requires modelling to account for these data characteristics. In our companion paper (Axon and Elbadawi, 1994b) we buid on the results of this paper, and estimate dynamic models influenced by auction dteory for the determination of the auction rae in these four countries, using weekly micro-auction data. Our modelling methodology aes account of the non-stationarity of the auction data as well as regime changes and policy interventions. As motivated above, we use dtese models to test the impact on the level of the auction rate of increased competition amongst bidders, the effect of uncertainty (proxied by a volatile supply of foreign exchange to tie auction), and dift pricing mechaims (Dutch and marginal pricing). -21 - REFERENCES Aron, J. and Elbadawi, 1. (1992). Parallel markets, the foreign exchange auction and exchange rate unification in Zambia, PR Workin Paers No. I992-909, The World Bank. Aron, J. and Elbadawi, 1. (1994). Foreign exchange auction markets in Sub-Saharan Africa: Par Dynamic models for the auction exchange rates, Mimeo, PRDTM, Tbe World Bank. Campbell, J. and Peron, P. (1991). Pitfalls and opportunities: what maroeconomists should know about unit roots, NBER Macroeconomics Conference, Cambridge MA, March 8-9, 1991. Cochrane, J.H. (1998). How big is the Random Walk in GNP7, Joumal of Political Economy, 96:893-920. Dominguez, K. (1990). Do exchange auctions work? An examination of the Bolivian experience, NEER Universities research conference on exchange rate regimes. Eagle and Granger (1987). Cointegration and error correction: represention, estimation and tesng, Eco2o22meri, 55, 251-276. Goldstein, H. (1962). The Friedman proposal for auctioning Treasury bills. Joumnal of Political Economy, 70, 4, 386-92. ICiguel, M. and O'Connell, S. (1992). Parallel foreign exchange markets in developing countries: experience and lessons, mimeo, The World Bank. Krumm, K. (1985). Exchange auctions: a review of experiences, CPD Discussion paper, No. 1985- 22. The World Bank. Nelson, C. and Plosser, C. (1982). Trends and random walks in macroeconomic time series: some evidence and implications, Joula ofMpntaryLanQmnta, 10, 139-162. Perron, P. (1989). The great crash, the oil prce shock and the unit root hypothesis, 57, 1361-1401. Prtchett, L. (1991). Measuring real exchange raze instability in developing countries: empirical evidence and implications, Policy Research Working Paper #791, The World Bank. Quirk, P., Christensen, B., Huh, K. and Sasai, T. (1987). Floating exchange rates in developing countries. Experience with auc;ions and inter-bank mrkets, IMF, Occasional Paper 53. Robinson, M. (1985). Collusion and the choice of an auction, Rand Journal of Economics, 16. 141- 143. -22 - TAUI I Sowe rme. es 1_ 0 of dSA te.elgn muaaug _adum. ZAMBIA UGANDA NIOIRIA GHANA duraion 1t Oot'S0 lo 31 Jun.'92 we 20 u.p.'fl ID 10 Sept'6O to 24 Jan.'57 7 May'23 24 Nov.'15 6 MaeohWS2 numbs s 67 07 270 fraqumy wekly wekly wbldy1 -241 W"ek fonlhtlZB*57l pricingl oompeptitill42l Dutchlt7) eopeltlvmll-241 omplhel I Dulmh143-681 Du_ohI25-W71 Dulhl2-2701 male raS 'wholeel 'du 11 -1 74) whollaull(l7C-) biddn Imnportm bunk. gid to bm wh ra sl 114-30 Impornm (1-174) Imorntersn 11- bank. id to .beaII 11S1-1741 *buraux 110-1 mwora. 130-671 banlm (17112701 eubmlan indual bhde conolte bide dwMdusl hid Irdldsu bid bvy Im by bunk (igg.gatln by bank by Imr (1-174) Import ble by prfloe by brnk (l75-270) entry} reutrlotlone yes InchredInitia removed yes Inresd halt gradualv baring madon during auction durin auction -.movud totall allacative cape ri Y- Flo manltoulng .trlngent Stringmnt varlble(I1-691 e autwod pply fo aid aid oil ireei aid,oooe recelpts rppr re- no r tt eXP. publihed bi. ye: Inividual bids y- compoel no no P,aU. 137-68) bId. by bank P,CA ree_we p- no ya: not no y": not announcedanoce pr- ouwcmd yew a calling, ye: js prior yae Jw prior no - uppi not adh eed to. lo open t openng - except 1iEi -6 bid. bids trapaeny poor high poo high noonday malt Ulegal. black bma legal .kegel black ilegal blcik(l-741 bureaux.lglm(74-) SOUNOt CeAl Da guldene; Iteve Would Bank *nd LM.F.. 1. Zamblb had O *uiier two-tier aclon MauchApril. 1987 (not ended dus to ncoplete daa, and hi Aprl 1987 revrted to lbaid raft regih. Ugandaes action wee ongoig at di time of endyg e deW but we replid by an bntwnk mabe In ate 193. Nieiaeublhed an nkw maretwdurJ J Day aton r bd, butwreot ie ba were guaranted foegn exchangeeocuion baed on netitonal alm mwdlm of hekr bide GOhm ened dhe maction nd established an Ikneb mae in Marci 1992. .2. The del"nt f pri - a follws (H competitive Icng : biddera paV tie margiW raft (onrerg bids by d eaan price) whch uOauma the aWume fogn exchange. sgerdlm of VOW bide (U Dutch pricing: bdere pay the pric te bid. 3. Sddere refe to lhcee who su*uSfjg bids to eCenrl Ba i the heS aruow hnpom uAnittd ir own bide whih wr chnneld ough nanpetoplg bait). in Nigri* whcieeela u*nc at tat maeoned foten exc cued be tV rsd bind ta mallet teraer bank.woe requd tD he non-baiuetei to cover tir bi. Neverdheless k wa li bh bide alone St detarm-nad the clearIng r.im in Nigeria. 4, The c 'llge on maxinum she of offered uppl or biddin banta w re fquendy te in Nlgeri 5. P. C U: prc. qu_tt requead. ae of forein exae. The mm. for hti mialunet of action anprency - discurned In the texL 23- TABLE 2t Aveage wereely uel sofe,o m vadb by aumsde typ.1 Z&MDIA COMPPETITIVE (1 421 DUTCH AUCTION (43468 OVERALL 1141 Mean Mh,mt. axmum Mtdhan Mom Milmtmu Maximum Mean Mhtlmum Maxdmurn auctonra t 6.56 5.01 8.07 10.11 5.01 15.35 7.92 5.01 15.26 maxim bid 7.97 6.50 15.00 11.54 6.85 18.00 9.34 6.50 19.00 mimemn bid 5.00 2.44 7.10 6.47 4.01 6.00 5.11 2.44 7.10 max-nil spread 2.97 0.55 12.25 607 1.05 12.00 4.16 0.55 12.26 max-oar spread 1.41 -0.01 0.99 1.43 0.26 5.90 1.42 0.01 9.99 nin-o spread 1.56 0.03 4.6e 4.84 0.00 9.25 2.74 0.00 9.25 supply 1* mar 5.67 3.10 8.80 6.0% 2.80 20.80 5.90 2.60 20.80 demand CO mRI) 9.19 3.50 21.70 14.25 5.0B 24.00 11.12 3.50 24.00 suppivldanand 0.08 0.20 1.00 0.46 0.12 0.91 0.69 0.12 1.00 muts of Mdte totslno. bkd 289 101 6o0 462 192 769 3S5 101 760 wkwing/otaI bids 0.65 0.21 0.9 0.42 0.17 0.98 0.56 0.17 0.19 bleak madket aeclu/auc pr.mmum 1.33 1.15 1.90 1.59 1.27 1.93 1.43 1.16 1.93 UGAND)A - OvERALL AUCTKIO 11-621 Mean Mhmau Maxbnwn n: auction rate 1022.21 970.00 1042.0O madmum bid -1043.73 1010.00 1200.0 miirdmu bI 101 9.45 900.00 1042.00 :pmb max-mih npd 24.27 1.00 200.00 max-oar spread 21.52 1.00 176.00 mincer spoad 2.76 0.00 80.00 suppjv1$ InSl 1.80 0.33 3.48 demand It rmlW - 2.02 048 4.0D wupply/demand 0.90 0.17 1AID numbe al bids! ttial no. bids 14.32 7.00 30.00 wnvlnrgftotal bids 0.92 0.36 1.00 humessm mainet burma rate tel. T.Cheque) 1216 1192 1264 buresoc bid-ak speed 0.96 0.90 - 0.97 btrlhuc priedum (T.Clque) .1.109 1.16 1.30 burhsuc preniurn IcashI 1.24 1.20 1.33 -24 - TABLE 2 iContd.l GHANA RETAIL (1-1761 WHOLESALE 1176-2701 OVERALL 41-270W Mean minhnum Maximum Mean Mlninum Maxinnum ean Mlninum Ma ximum aution rate 216.89 128.00 314.00 368.25 317.00 395.00 288.35 128.00 306.00 mxilmum bid 223.15 152.00 316.00 362.03 318.00 420.00 272.37 152.00 420.00 minimum bid 213.56 P0.00 314.00 3S7.63 317.00 396.00 264.62 90.00 395.00 mmemin spred 9.60 2.0D 68.00 4.40 0.00 30.00 7.76 0.00 68.00 max-or spread 7.26 1.00 51.00 3.78 0.00 27.00 6.02 0.00 51.00 mhar spread 2.34 0.00 45.00 quendtlls: supply P mull 5.46 1.06 12.38 7.37 0.86 18.46 6.18 0.86 18.48 demand l* mIll) 6.78 1.35 17.91 8.82 0.85 19.28 7.47 0.86 19.28 supply/demand 0.85 0.1B 1.00 0.87 0.16 1.00 0.86 0.15 1.11 number of bid: wotl no. bids 141 32 468 n na no na no na winnkinltot lbx 0378 0.15 1.00 na na no us na - n bunaux merlkt' bur. bkidal spread 0.S5 0.90 0.98 0.37 0.94 0.99 0.98 0.90 0.99 bureewc rate teal 363.21 317.00 377.50 378.76 347.50 440.00 372.44 317.00 440.00 bur/ c. prmnhum 1.31 1.14 1.51 1.06 1.03 1.14 1.18 1.03 1.51 NIGERIA COMPETITIVE 1-42) DUTCH AUCTION (43-48) OVERALL (1483 Mean minimum Ma,dnum Mean MWnmum Maximum Mean Mhiimun Maximum puloe. autio rate 3.71 3.00 4.90 4.25 3.50 5.28 4.06 3.00 5.28 maximum bid 4.33 3.10 5.60 4.41 4.02 5.30 4.38 3.150 5.60 minimum bid 3.38 2.50 3.90 4.21 3.50 5.28 3.91 2.50 5.28 mnaxmn sprea 0.95 0.30 2.64 0.21 0.00 0.90 0.47 0.00 2.64 mar-or- pread 0.62 0.25 1.85 0.16 0.01 0.90 0.33 0.01 1.86 minoer Bpread 0.33 0.00 1.90 0.04 0.00 0.36 0.15 0.00 1.90 supply (J mill 63.17 50.00 86.00 108.83 70.00 120.00 91.08 50.00 120.00 demand IJ mili 71.77 38.39 100.10 123.80 64.39 150.00 105.16 38.39 160.00 supplyfdumand 0.90 0.50 1.30 0.87 0.67 1.10 0.88 0.50 1.30 number of ldsc tots no. bid 38 27 44 48 30 53 45 27 53 wiwnng/ot bids 0.86 0.34 1.00 0.89 O.SB 1.00 0.88 0.34 1.00 bureauxrn 5.00 4.70 5.70 5.70 4.20 S.o6 5.45 4.20 8.60 burw/ucpiemiun 1.386 1.10 1.67 1.33 1.04 1.75 1.34 1.04 1.7S SOURCE Autho- calcuaions 1. The ucton varable a- mostly non-stationery (Aoand Elbdrwl. 19941. so that convmenonal meaLe of variabilit lIrd deviotio are not define. 2. Athough aN Zamblan parastattb wer iclded in ebs suction tram aucton 20. the largest ot them continued to renwv subtantil foreign exchange with nmrcompetlv bid - tw myhal rare. A cornred value is th use for wsppy in lbs Zambian regrsons. 3. Blac market data or Zambia me monthly. from Picke Cumernc Yearbook, Weekl Nigwisn blck markat dat -r compled from ith World Bauk reident mission satin ad Paici. Forign exchange buaux began functioning in Ghan from ApdrL 19OE. Regresidons ue some black medet data. - 25 - TAILE 3: Masuren of wmaflty for flpt-ulNfrenod auoden vadabis. b, b or0 5 S. S, N Skewnss Kurtosis Std. Dev. X..gC X.. .Ixi a~~~~~~~~~~~~~~~~~~~~~~~~~~~a 2AMB A-' ' ' - aloulerl 67 -2.38 10.22 0.10 -2.87 3.25 7.90 logtlO) 67 -0.16 1.44 0.41 -0.76 3.64 4.74 Ala(R0a.I - E7 0.81 2.37 0.46 -1.E6 3.18 5.36 aIog(Odl 67 0.47 0.32 0.27 0.84 3.72 2.86 *1OQ(02Idl 67 -0.78 1.71 0.42 -1.30 3.00 4.78 Alogprnmil 57 0.22 -0.78 0.04 0.43 2.01 2.11 aIgipromnl 67 0.12 6.16 0.09 0.68 9.59 7.92 UMANDBA og(o0r) 86 3.31 12.45 1.99 6.18 9.09 0.00 AlogoIla) 66 0.42 2.86 0.70 0.24 4.79 4.45 AloogOdi 66 -0.06 2.01 0.59 0.31 4.39 3.96 nlog(OaIO 86 0.68 9.73 0.36 .1.03 8.76 6.81 iloogberc) 61 -1.95 8.34 3.58D-3 -1.88 3.48 6.66 aogilpra, 61 -1.46 2.09 4.29 -1.91 2.31 6.26 aIo(psam0l S1 -0.80 2.74 4.38 -2.24 2.58 5.53 GRANA a&logoied 269 2.64 16.30 9.91 2.48 11.81 7.20 AbqglCs) 269 -0.09 4.78 0.41 -0.86 6.47 7.86 l&iCodl 269 -0.9a 5.16 0.41 -2.74 4.44 8.1 ogiIOdM) 269 0.80 5.62 0.29 2.25 9.81 5.87 aloqbar) 196 -1.32 14.77 0.01 -2.08 .06 12.12 Abolpreml 196 -0.85 9.74 0.02 -2.23 5.99 9.72 NVIGMIA Mogioer) 66 -1.33 6.71 0.08 -089 5.37 6.84 aloglOal 66 0.38 4.78 0.14 0.45 7.64 6.59 Mlog(gdl Go -1.25 5.47 0.1 9 -2.08 3.79 7.29 oig(Os/Qd) 66 1.95 6.86 0.16 2.14 8.06 4.10 &bgogbrl 6e -0.01 2.55 0.04 0.20 5.08 4.77 Alogpron) es6 1.30 6.8S 0.09 1.69 7.35 4.58 SOURCE: Authom calcultions. NOTES: 1. The followin abbreviations are used: oer bauetion exhange rte. leal currercltl): Gs IS suppled): Cs. (I uppliad, lass ora- competitve parasal bids at the maginal rata Gld It demaededi: prem Ipwradl Car bureiuc)Iauction rate): prest. rmtwpaatad monthly promunl: bar, (bureau soll rate for travels cheques): bor0 (bureau sell rate for asuh). -26 - FIGURE I a,b,c,d: The equilibrium auction price and parallel market premium. ZAMBIA 19 85 -87. 16.0- ~~~~~~~~~~~~2.00 14.0C 1.90 COMPETiTK DUTCH - ~~~1.80 12.0- 1.70 - 1.50 00 !1.50 3 1020 \ eES . -1.40 1.30 -1.20 480 ................. /S-.10 1 5 9 13 17 21 25 29 33 37 41 45 49 535761 55 3 7 11 15 19 23 27 31 35 39 43 47 51 55 596367 oaction ntmer | L- ciuJDn role - blu k/auclion rale UGANDA 1992-93 1050- p1.32 1040 1.0 1030- 12 1020 - ~~~~~~~~~~~~-1.26 X 1010. ~~~~~~~~~~~~-1.24E 1090. i~~~~~~~~~-.22 rL S90 - 1~~~~~~~~~~~~1.20 970 - ----------- ~ ~~~~~1.16 1 5 9 13 17 21 25 29 33 3741 4549 53 57 61 3 7 11 15 19 23 27 31 35 39 43 47 51 55 59 actionr mmnber ~Smacion role -bureau TCfwzcli-on - 27 - GHANA: 1986-92. 400- 1.60 350 - 1.50 200 . J -1.40 250 t1.30 150. . l4 1 27 53 79 105 131 157 183 209 235 261 aucuion nwmbe - dion role - bureau A/el/cudon NIGERIA 1986-88. 5.5 . 1.80 . COLTmVr DUTCH' 1.70 1.50 ~* 5 ! j+ 1.5°0 E 4- 1.40 a. 3.5 ~~~~~~~~~~~~~~1.20 1.10 3 ......... .00 1 5 9 13 1721 25-29333741 4549535761 65 3 7 11 15 19 23 27 31 35 39 43 47 51 55 59 63 67 auclion number | auLion role block/-uction rote| -28 - FIGURE 2 a,b,c,d: Skewed and kurtotic distributions of auction exchange rate changes. ZAMBIA UGANDA - 29 - GHANA .'11 u~~~~~~~~~~~~ tMI 4 - - 30 - FIGURE 3 a,b,c.d: Codhmne persistence test for the auction exchange rates. ZAMBIA: 68 auctions. 1.4 - 1.26 02 5 9 13 17 21 -3 7 1 1 15 19 23 |-* vcrionce ralc - std errors l .2. *. 2 . .. 0 I 0.25 1 5 9 13 17 21 3 7 11 i5 I 19 23 k -|- vcriknce retlio std errors UGANDA: 67 auctions.~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - 31 - GHANA: 270 ouclions. 2.5 1.5 :i .. 1 5 9 13 17 2125233341449356166 3 7 11 15 19 23 27 31 35 39 43 47 51 55 59 63 67 71 wariae,. rol- sid errors MNGERIA: 67 ouctions. 1.4 _ 1.2. 1 0.8 - *j0.6 0.4 0.2 - 0 -0.2 1 5 g 13 17 21 3 7 11 15 19 23 e t ~~~~~-- | rjI: ratio std errorsl -32 - FIGUR1E 4 a,b,c,d: Macroeconomic policy indicators and the free excuange rate. ZAMBIA: monthly 1 9B5-87. 14 7 70.00 2- -60.00 10 50.00 -40.00 6~~~~~~~~~~~~~~ 30.00 . 20.00 o J 2 10.00 0 U -. 0.00 -4' q10.00 -6 -20.00 N4ov-85 mor'-55 Jii-86 NOV-186 Jan-86 "-ay-6 Sep-836 -i-M2 --ClI - black rote UGANDA: monthly 11992/93.. 12' 10 .6 0i C Z- Jan-9 2 May-9 2 Sep-92 Jan-9 3 Maor-92 Jul-9 2 Nov-92 Mor-93 --quosi-money 'IEI CPI -bureaux rote - 33 - GHANA: quarterly 1986/92. 16 40.00 14 -35.00 i12 a30.00 10. 25.00 Om.20.00. 2' . -2\/V .0 13604 8704. 8804 8904 9004 9104 8702 8502 8902 9002 9102 9202 NIGERIA: monthly 1986/88. 10- 20.00 8 -15.00 6- 10.00 0- -~~~~~~~~~~~~5.00 Oct-86 Feb-87 Jun-87 Oct-B? Feb-88 Jun-88 Oci-88 Dec-86 Apr-87 Aug-537 Dec-87 Apr-88 Aug-88 Dec-88 --n- M2 -2- cmI black rate Policy Research Working Paper Series Contact Title Author Date for paper WPS1371 The Evolution of Trade Treaties and Sarath Rajapatlrana October 1994 J. Troncoso Trade Creation: Lessons for Latin America 37826 WPS1372 Administrative Charges in Pensions Salvador Vald6s-Prieto October 1994 E. Khine in Chile, Malaysia, Zambia, and the 37471 United States WPS1373 Firm Behavior and the Labor Market Simon Commander October 1994 B. Keller in the Hungarian Transition Janos Kollo 35195 Cecilia Ugaz WPS1374 Infrastructure Finance: Issues, Anand Chandavarkar November1994 M. Geller Institutions, and Policies 31393 WPS1375 Policy Lessons from a Simple Shantayanan Devarajan November 1994 C. Jones Open-Economy Model Delfin S. Go 37754 Jeffrey D. Lewis Shemian Robinson Pekka Sinko WPS1376 How Trade, Aid, and Remittances Maurice Schiff November 1994 J. Ngaine Affect International Migralion 37947 WPS1377 Macroeconomic Adjustment to Vittorio Corbo November 1994 R. Vo CapitW Inflows Latin American Style Leonardo Hemhndez 33722 versus East Asian Style WPS1378 Mexico after the Debt Crisis: Daniel Oks November1994 S. Harbi Is Growth Sustainable? Sweder van Wijnbergen 37143 WPS1379 Fnancing Infrastructure in Developing Barry Eichengreen November 1994 WDR Countres: Lessons from the Railway 31393 Age WPS1 380 Transfers and the Transition from Katie Krumm November 1994 K Krumm Socialism: Key.Tradeoffs Branko Milanovic 34263 Michael Walton WPS1381 Welfare Economics, Political RaviKanbur November1994 M- Yousset Economy and Policy Reform in Ghana 34614 WPS1382 Saving, Investment and Growth in Klaus Schmidt-Hebbel November1994 E. Khine Developing Countries: An Overview Luis Serv6n 37471 Andr6s Solimano WPS1383 Rural Demand for Drought insurance MadhurGautam November1994 C. Spooner Peter Hazell 30464 Harold Alderman Policy Research Working Paper Series Contact Title Author Date for paper WPS1384 Fiscal Decentralizatlon and David Sewall November 1994 G. LangLon Intergovemmental Finances in the Christine I. Wailich 38392 Republic of Albanla WPS1385 Fiscal Federalism Dimensions of Tax Robin Boadway November 1994 C. Jones Reform in Developing Countries Sandra Roberts 37764 Anwar Shah WPS1386 EU Bananarama iII Brent Borrell December 1994 G. Iiogon 33732 WPS1387 Fiscal Decentralization and the Size Jsber Ehdale December 1994 C. Jones of Government: An Extension with 37699 Evidence from Cross-Country Data WPS1388 Does Voice Matter? For Public Samuel Paul December 1994 B. Moore Accountability, Yes 35261 WPS1389 Ruble Overhang and Ruble Shortage: Patrick Conway December 1994 L. Suld Were They the Same Thing? 33974 WPS1390 Regional Integration and the Baitics: Piritta Sorsa December 1994 J. Ngaine Which Way? 37947 WPS1391 Where in the World Is Population Jeff Kling December 1994 S. Fallon Growth Bad? Lant Pritchett 38009 WPS1392 Some Economic Consequences of Jean-Paul Azam December 1994 C. Jones the Transition from Civil War to Peace David Bevan 37699 Paul Collier Stefan Dercon Jan Gunning Sanjay Pradhan WPS1393 The Interf,ace of Trade, Investment, J. Luis Guasch December 1994 J. Troncoso and Competition Porlcies Issues and Sarath Rajapatirana 37826 Challenges for Latin Ameica WPS1394 Macroeconomic Reform and Growth Lawrence Bouton December 1994 R. Martin in Africa: Acgustment in Afdca Christine Jones 39026 Revisited Miguel Kiguel WPS1395 A Typology of Foreign Exchange Janine Aron December 1994 R. Luz Auction Markets in Sub-Saharan Ibrahim Elbadawi 39059 Africa WPS1396 Foreign Exchange Auction Markets Janine Aron December 1994 R. Luz In Sub-Saharan Africa: Dynamic lbrahim Elbadawi 39059 Models for Aucbon Exchange Rates