Exchange reform, parallel markets, and inflation in Africa : the case of Ghana

Policy, Research, and Extemal Affairs LA) If) S 1 WORKING PAPERS Office of the Vice President Development Economics The World Bank May 1990 WPS 427 Exchange Reform, Parallel Markets, and Inflation in Africa The Case of Ghana Ajay Chhibber and vv Nemat Shafik This model using Ghanaian data shows that in the r)resence of an active parallel market, official devaluation does not cause infla- tion because prices have already adjusted to the parallel ex- change rate. Although structural factors were important in the past, inflation in Ghana has been primarily a monetary phenomiie- non and the product of weakness in the financial system in recent years. The Polcy. Research, and hxtemal Affairs Complex disinhutes PRE Working Papers to disseminate the findings of work in prmgrss and to encourage the exchange of ideas among Hank staff and all oLhers nterested in developmnent issucs 1he-se papers carr) the names of the auLhors reflect only their viesw and should he used and citrd accordingly 1'he findings, interpretations, and conclusions am Lhe authors' own. 'hey should not he aitnuiited t the Wirld lank, is Board of l)Dirciors. iLs managanent, or anr of ii. memher cisintncs  Policy, Research, and External Affairs Office of the Vice President This paper- a product of the Office of the Vice President, Development Economics -is part of a larger effort in PRE to understand the dynamics of inflation, exchange reform, and price decontrol in Africa. Copies are available free from the World Bank, 1818 H Street NW, Washington DC 20433. Please contact Maureen Colinet, room S9-041, extension 33490 (67 pages with figures and tables). Adjustment programs typically involve a and simulations using a macroeconomic model complex policy package that includes price estimated with Ghanaian data. liberalization, devaluation, and trade policy re- forns - together with public enterprise and The model rcsults show that therc is no fiscal reform, including reduccd subsidies and direct rclationship bctween thc official exchange rationalization of public spending. rate and inflation; prices had already adjusted to the exchangc rate prevailing in parallel markets. A common concem in these reform pack- ages is the potential inflationary effects of the The results also show that official devalu- combination of devaluation, trade liberalization, ation had a positive effect on Ghana's budget. subsidy reduction, and price decontrol. This Revenue improvements came from three chan- issue is critical in Africa, where inflation has nels: the higher grant aid disbursed at a more accelerated in several countrics, particularly depreciated exchange rate, a reduction in the those undergoing adjustment. subsidies that had accrued to importers through an overvalued excharige rate, and an increase in Some argue that devaluation - an important export taxes as cocoa farmners increasingly instrumenit in IMF and World Bank adjustment marketed their output through official channels. programs - is not necessarily the best instru- ment for real exchange rate devaluation, given The oflicial devaluation therefore did not its inflationary effccLs and taking into account produce higher budget deficits, demand pressure the structure of African economies. Some have did not spill onto the parallel market, and the criticized INIF and World Bank programs as exchange premium narrowed considerably. The leading to teiC so-called "Latin Americanization" key to the success of the program was the of Africa. adequate level of loreign financing, combined with a coherent set of liscal policies. In Ghana, which has carried ouL one of the most thorouglh structural adjustment programs in Chhibbcr and ShaFlk argue that although Africa, an increasingly high inflation rate has inflation had structural causes in the past, the ac- been attributed to major devaluations of the celeration in recent years is primarily a monetary official exchange rate. Chhipbersand Shafik phenomenon. It also reflects weakness in the fi- dispute this conclusion based on careful esting nancial system that must be tackled to sustain reform. The l'RE Working Paper Series disscminates the findings of work under wvay in the Bank's Policy, Rcscarch, and External Affairs Comiiplex. An ohjectivc of the series is to get thesc findings out quickly, even if prcsentations arc lcss than fully polished. The findings, interprctatik-ns. and conclusions in these papers do not necessarily represent official Bank policy. I'roduced at the I'RE Dissemination Ccnter  (i) TABLE OF CONTENTS ThXI EAQ 1. Introduction ..................................... 1 2. The Model ...................... .............. 3 2.1. The Theoretical Model .......................... 6 3. The Empirical Estimation of the Model . .................. 13 3.1. Block A: Inflation ................. 13 Specifying Excess Demand . ...................... 19 Wage-Push Inflation ............................ 20 Econometric Estimation: Inflation ................... 21 Inflation and Exchange Rates ...................... 25 3.2. Block B: Exchange Rate Premium ................... 27 3.3. Block C: Monetary-Fiscal Block .................... 35 Money Supply ..... ..... ................ 35 The Fiscal Deficit and its Components ................ 36 3.4. Block D: The Real Side: Investment and Output ... ....... 40 4. Policy Issues ..................................... 44 4.1. Exchange Rate Policy ............................ 44 4.2. Monetary and Credit Policy ........................ 47 4.3. Food Prices, Subsidies and Inflation .................. 50 4.4. Prospects for the Future .......................... 52 5. Conclusions ...................................... 53 Bibliography ....... ............. ................... 57 This paper is part of a larger research project on inflation and fiscal adjustment in Africa. We are grateful to Vikram Nehru for his help and encouragement, but absolve him of any responsibility for the interpretations. Thanks are also due to the participants in a seminar sponsored by the Ministry of Finance in Accra on November 16, 1989.  Table 1 Ghana - Economic Indicators ...................... 14 Table 2 The Model .................................. 16 Table 3 Inflation Equations with Wage Variables, 1965-87 ... ...... 22 Table 4 Inflation Equations, 1965-88 ....................... 24 Table S Exchange Rate Premium Equations, 1965-88 ............. 31 FIGURES Figure I Mark-Up Model of Inflation Determination ............. 8 Figure 2a Equilibrium Inflation and Parallel Exchange Rates ... ..... 11 Figure 2b Comparative Statics With Devaluation of the Official Exchange Rate 11 Figure 3 Parallel and COfficial Exchange Rates ................. 33 Figure 4 Growth in Exchange Rate Premiunm Underlying Factors ..... 34 Figure 5 Government Transfers as a Share of GDP .............. 39 Figure 6a CPI With Slower Devaluation .. 46 Figure 6b Parallel Market Exchange Rate with Slower Devaluation ..... 46 Figure 7 CPI Inflation With and Without Drought .............. 52 APPENDIX FIGURESi PAGE Appendix Figure I Inflation in Consumer and Import Prifes ... ..... 60 Appendix Figure 2 Nominal and Real Six Month Deposit Rate ... .... 61 Appendix Figure 3 Inflation and Real Output Growth ... ......... 62 Appendix Figure 4 Inflation and the Government Deficit ... ....... 63 Appendix Figure 5 Index of Real Wages ..................... 64 Appendix Figure 6 M2 as a Share of GDP .................... 65 Appendix Figure 7 Money Multiplier ....................... 66 Appendix Figure 8 Real Private Investment ................... 67  1. IntrductiG., Adjustment programs typically i. olve a complex policy package which include some combination of price liberalisation, devaluation and trade policy reforms, as well as public enterprise and fiscal reform, including reduction in subsidies and rationalisation of the public expenditure program. A common concern in these reform programs is the potential inflationary effects of a combination of devaluation, trade liberalization, subsidy reduction and price decontrol. This issue has bewome critically important in the African context where inflation has been accelerating in a number of countries, and in particular in those undergoing adjustment programs. This is particularly evident in countries which are not members of the CFA Franc zone. In a recent paper Guillaumont and Guillaumont (1989) argue that devaluation - an important instrument in IMF and World Bank adjustment programs - is not necessarily the best instrument to bring about real exchange rate devaluation given its infla:ionary effects and taking into account the structure of African economies. IMF and World Bank programs have been criticised by others as leading to the so called "Latin Americanisation" of Africa.1 This paper presents a theoretical framework to address the question of exchange rate devaluation in the presence of parallel markets in Section 2. It then uses the Ghanaian data to empirically estimate a model in Section 3 which is applied to analyze policy trade-offs for Ghana in Section 4. The last section draws together the lessons from the Ghanaian experience. Ghana has carried out one of the most thorough adjustment programs in Africa. Its inflation rate is high and in recent years 'The most recent critique has come from the Economic Commission for Africa (ECA).  2 has been rising, although it is lower than during the worst crisis years. Ghana's average annual inflation rate2 has declined considerably since 1984 as against the average for the p.eriod 1976-83. However, tie inflation raie has risen from under 10 percent per annum in 1985 to almost 40 percent per annum in 1987. Although inflation declined to a little over 30 percent per a,inum in 1988, there remain concerns and questions about its underlying causes and consequences. Ghana has devalued the official exchange rate on average by well over 40 per cent per annum between 1983 and 1988 (See Table 1). Ghana's high inflation rate has been attributed to the large devaluations of the official exchange rate. This paper disputes this conclusions about the inflationary consequences of devaluation based on careful testing and simulations vith a model estimated with Ghanian data. The model simultaneously determine. the rate of inflation and the exchange rate premium and links the monetary arnd real sides of the economy through several channels. The results show that there is no direct relationship between the official exchange rate and inflation; prices had already adjusted to the shadow exchange rate represented by the exchange rate prevailing in parallel markets. The results also show that the Ghanian official devaluation had a positive effect on the budget. The revenue improvements came primarily from two channels -- (a) from higher grant aid disbursed at a more depreciated exchange rate, and (b) from a reduction in subsidies to importers due to exchange over-valuation. The official devaluation therefcre did not result in higher budget deficits and as a result excess 2As recorded by the official CPI.  3 demand pressure did n-t s' tl onto the parallel market and the exchange premium narrowed considerably. The key to the success of the program was the adequate level of foreign financing, combined with a coherent set of fiscal policies. The paper shows that acceleration in inflation in recent years is primarily a monetary phenomenon, reflecting as well underlying weakness in the financial system which need to be tackled to sustain the reforms. 2. The Modei The approach used for the analysis attempts to integrate within -ne framework three transmission mechanisms for inflation in Ghana: via excessive money creation, via direct cost push factors such as the cost of foreign exchange and food prices, and through real faciors. WhUat is often missing in the analysis of inflation is the interactions b,tween these transmission mechanisms. For example, changes in individual prices and the exchange rate have budgetary implications. How the esulting change in the budget gets financed in turn affects debt management and inflation. Changes in monetary and credit policy have consequences for the availability of working capital and inve,tment capital. This affects the level of real output which also has implications for the management of inflation. The framework set up in this paper is designed to quantify these interactions in the presence of parallel markets.  4 The approach adopted here, in some ways, synthesises the monetarist and cost- push schools of inflation.3 The monetarists argue that cost-push factors such as food prices, wage or exchange rate changes can only cLuse a shift in the price level but cannot explain how it translates into sustained inflation. For this to happen, one needs monetary accomodation, without which the inflationary process stalls. While this is no doubt true, we argue that although money can be viewed as the proximate cause of inflation, cost-push "actors can trigger, directly or indirectly, changes in the rate of monetisation and in the velocity of circulation of money. In order to understand the inflationary process, it is necessary to disentangle the underlying factors which ultimately result in higher growth of domestic liquidity. The focus of previous analytical work in this area has been on the interaction between exchange rate changes and inflation. When the exchange rate is administered it acts as a stabiliser to the inflationary process but at the cost of trade imbalances or gtowth.4 This happers because as inflation accelerates due to high fiscal deficits or cost-push factors such as an increase in food prices, the real exc, aIge rate appreciates. Where the nominal exchange rate is allowed tc adjust to inflation automatically the inflationary process can become self-feeding.5 Similar problems can 3 For a comprehensive treatment of the monetarist and cost push analyses of infation in developing economies, see Kirkpatrick and Nixson, (1976), Williamson (1985), Taylor, (1983). See for example Khan and Lizondo (1986). See Drazen and Helpman (1986) and van Wijnbergen (1988) for an analysis of this issue,  5 arise in the case of dual exchange rates' when the government attempts Lo control the cxchange rate changes and there is a parallel market for foreign exchange. As the official rate remains fixed and the parallel market depreciates, export earnings channelled into the official exchange market decline and the duality in the exchange market becomes unsustainable. With a rising parallel market premium the implicit tax on exchange rate proceeds in the official market increases. When the government attempts to unify the two markets, it loses the implicit tax. Unless accompanying fiscal changes are brought about, the higher fiscal deficit (due to the loss in the implicit tax on foreign exchange transactions) is usually monetised leading to higher inflation which could lead to further depreciation of the parallel market exchange rate. A destabilising process from unification to higher deficits, higher inflation and an increase in the parallel market premium can be established. Pinto (1988) shows the conditions under which this destabilising process can develop. This has obviously not happened in Ghana as the premium between the two rates has gradually declined. Moreover, the government's budgetary position has, if Lnything, imp.oved due to the devaluation of the official exchange rate for relsons that are explained in later sections. 6 See Dornbusch (1986) and Pinto (1988)  6 2.1. The Theoretical Model A segmcnted goods7 market is hypothesized. There is an official market at which goods are available at a subsidised price (P3). In the parallel market and the uncontrolled legal market there are traded goods and non-traded goods. Traded goods are priced equal to foreign goods (Pr) converted at the parallel exchange rate, plus a mark- up for the unit costs of smuggling(s).$ PI = (I + s) (ep. Pr). We assume for expository convenience that s = 0 PX = e. + ;f ;S = O. A third category of final goods transacted in the parallel market are non-traded goods. The production of these requires labor and imported inputs. The foreigb exchange for the purchase of imported inputs comes from official sources (e0) as well as from the parallel market (ed). The following variable mark-up model is used to explain non-traded goods prices (Pz):9 7Refers to goods and services. The variable arguments of this unit costs function are the extent of policing and the volume of smuggling. A larger volume of smuggling decreases the unit costs of smuggling. The unit costs of smuggling are assumed to be invariant to the level of demand in the economy. 9See Kalecki (1971).  7 P2 = (I + u) 6eOp . Pr + (1-4) eoPfll (wp) 2 where wp denote unit labor costs, and B is the parallel market's share of foreign exchange used for imported inputs. The size of the mark-up is a function ^f excess demand (ED) in the economy. P2 = ao ED + a, O[ep + Pj + al (1.O)(^9 4 P,) + a2 wp The average rate of inflation is weighted average of the changes in prices of traded (P,) and nontraded (P2) goods and goods at controlled prices (P3). 1= + X2P2 + (1-As - X2 P3 Equilibrium in the goods market is depicted in Figure I by the PP line. In area a, there is excess supply as the mark-up over unit costs is too high. Prices must fall to clear the goods market. At b the opposite holds. There is excess demand since the mark-up is too low and therefore prices must rise to clear the market. An increase in aggregate demand brought about by a larger fiscal deficit shifts PP to P'P, thereby increasing the mark-up. We assume that there is a surp!us in the labor market such that an increase in aggregate demand decreases the size of the labor surplus, but has no effect on unit labor costs. In the asset markets, there is a domestic asset cedis, and a foreign asset, foreign exchange (F).  Mark-Up Model of Inflatior, Determination 12.0 9-5 7.0 - b Mark-up { Unit Wage Costs . |~~~~~~~  9 Overall demand is a function of total financial assets which are defined as: A = M + epF Where M represents the money supply, ep is the parallel exchange rate, F iQ the foreign asset. The demand for foreign exchange is composed of transactions demand and portfolio demand. ep. Fd = h(ep/eo, P/P3, A) + j [(i - I - ep), A] where hi < 0, h2 > 0, h3 >