Turkey - External debt, fiscal policy and sustainable growth (Vol. 2 of 2) : Methodological and statistical annex

Report No. 7162-TU Turkey External Debt, Fiscal Policy and Sustainable Growth (In Two Volumes) Volume II: Methodological and Statistical Annex September 9,1988 Country Department I Europe, Middle East and North Africa FOR OFFICIAL USE ONLY Document of the World Bank This report has a restricted distribution and may be used by recipients only in the performance of their official duties. Its contents may not otherwise be disclosed without World Bank authorization.  CURRENCY EQUIVALENT (annual average) 1980 TL 76.04 1981 TL 111.22 1982 TL 162.55 1983 TL 225.46 1984 TL 366.68 1985 TL 521.98 1986 TL 674.50 1987 TL 860.58 FISCAL YEAR March 1 to February 28 - through 1981 March 1 to December 31, 1982 January I to December 31 - from 1983 ABBREVIATIONS CPI - Consumer Price Index EBF - Extra-Budgetary Fund FX - Foreign Exchange Deposit MHF - Mass Housing Fund SEE - State Economic Enterprise SIS - State Institute of Statistics SPO - State Planning Organisation WPI Wholesale Price Index  FOR OMCIL USE ONLY This report is based on the findings of a World Bank mission that visited Turkey in October-November 1987. The mission consisted of Ritu Anand, mission leader (EMlCO), John Brondolo (CECEM), Ajay Chhibber (CECEM), David Robinsor (IMF), Roberto Rocha (EMTTF). Sweder van Wijnbergen (EMTTF) developed the models used and took part in the preparation of the report. This document has a restricted distribution and may be used by recipients only in the performance of their official duties. Its contents may not otherwise be disclosed without World Bank authorization.  TURKEY: EXTERNAL DEBT, FISCAL POLICY AND SUSTAINABLE GROWTH Table of Contents VOLUME II: METHODOLOGICAL AND STATISTICAL ANNEX"' Page No. Annex II: EXTERNAL ADJUSTMENT, EXCHANGE RATE POLICY 1 AND OUTPUT GROWTH A. Exchange Rate Fluctuations and External Debt: I A Decomposition Method B. Debt Dynamics, Output Growth and the 3 Current Account C. Solvency and Creditworthiness 6 C.1 Solvency 7 C.2 Creditworthiness 9 Table A 2.1 - Currency Composition of Total Net External Debt at Current Exchange Rates 11 and at 1980 Exchange Rates (in percent) A 2.2 - Currency Composition of Net Public 12 External Debt at Current Exchange Rates and at 1980 Exchange Rates (in percent) A 2.3 - Currency Composition of Net Private 13 External Debt at Current Exchange Rates and at 1980 Exchange Rates (in percent) Annex IlI: CONSISTENCY REQUIREMENTS FOR FISCAL POLICY 14 A. Introduction 14 B. Analytical Framework 16 L.1 Fiscal Deficits, Money Creation 16 and Debt B.2 Inflation Tax Revenue and the 20 Structure of the Financial Sector B.3 Creditworthiness and the Limits on Foreign Borrowing 22 B.4 Putting It All Together: the Design 26 of Consistent Fiscal Policies C. Empirical Requirements 28 C.1 The Structure of Money Demand 29 in Turkey Specification 29 Data 30 Estimation and Results 32 Annex numbers correspond with chapter numbers  (ii) Page No. Table 1 - OLS Estimates of the Demand for 33 Financial Assets 2 - Long-Run Coefficients 37 3 - TSLS Estimates of the Demand for 38 Financial Assets C.2 Export Demand Elasticities 39 D. Information on the Data 41 D.1 External Debt 41 D.2 Monetary Data 42 D.3 Domestic Debt 43 Table A 3.1 - Foreign Exchange Assets of the Central Bank 45 A 3.2 - Foreign Exchange Liabilities of the 46 Central Bank A 3.3 - Foreign Exchange Liabilities 47 of Government and SEEs A 3.4 - Average Interest on the Foreign 48 Liabilities of the Public Sector A 3.5 - Balance Sheet of the Central Bank, 1979-1987 49 A 3.6 - Non-Adjusted and Adjusted 50 Money Base, 1978-1987 A 3.7 - Net Domestic Debt of the 51 Public Sector, 19t9-1987 Annex IV: INTERNAL ADJUSTMENT: FISCAL POLICY, PRIVATE SAVINGS AND INVESTMENT, AND OUTPUT GROWTH 52 A. Model Structure 52 A.1 The Purpose of the Model '2 A.2 Real Interest Rates, Fiscal Policy, Output Growth: the Way the Model Works 54 B. Econometric Results 60 B.1 Private Consumption 60 B.2 Private Investment 62 B.3 Investment and Output Growth 64 B.4 The Model Equations 66 C. Data: 68 C.1 Effective Cost of Bank Loans 68 Table A 4.1 - Effective Cost of Borrowing, 1970-86 70 C.2 Other Data: Table A 4.2 - National Accounts, 1967-87 71 A 4.3 - Price Indices, 1967-87 73 A 4.4 - Data for Growth Equations 74 A 4.5 - Data for Investment Function 75 A 4.6 - Data for Consumption Function 76 REFERENCES 77  ANNEX II EXTERNAL ADJUSTMENT. EXCHANGE RATE POLICY AND OUTPUT GROWTH A. Exchange Rate Fluctuations and External Debt: a Decomposition Method. A 2.1 There are two channels through which exchange rate developments influence the debt-output ratio. One is through the real exchange rate, the relative price of aggregate foreign goods in terms of Turkish goods. This concept is closely related to the terms of trade. A real depreciation lowers the value of Turkish goods and thus raises the ratio of external debt to output. Equation (1) shows the mechanism: (1) b = (B$E/P)/y = (B /P ).(P E/P)/y -b e/y * * B is the dollar value of the debt and P the dollar-based foreign price level (see FN 1 on p.10 and on p.14, Chapter II for a more precise definition). b is the real value of foreign * * * debt in terms of foreign goods: b = Be/P. E is the nominal exchange rate against the dollar (local currency units per dollar), P the domestic price level and y real GNP. e represents the real exchange rate. In what follows we use capital letters for nominal values, lower case letters for teal variables, and a tilde to indicate variables expressed as a share of. GNP. Equation (1) brings out that a real depreciation feeds directly into the debt-output ratio: capital losses on foreign debt increase the burden of the debt.  2 A 2.2 There is a second channel through which exchange rate changes influence tli' burden of the debt. Cross-currency exchange rate fluctuations may affect the dollar value of the debt even if Turkey's exchange rate against the dollar remains unchanged. For example, an appreciation of the DM against the dollar raises the dollar value of the debt since a substantial part of Turkish external debt is denominated in DM. This is not captured in equ.(1) because of the implicit assumption of a single foreign good. A 2.3 To capture this effect, we define the value of the debt evaluated at 1980 * * * * values of cross-currency exchange rates, B $c* The difference bB8 = B$ - B$c is the cumulative loss (since 1980) incurred due to cross-currency exchange rate changes (CCE). In fact, CCE changes affect the foreign price level as well as the dollar value of the foreign debt. To capture this effect too, we define P c as the dollar-based foreign price index evaluated at 1]980 cross-currency exchange rates. The difference witb the actual price level * * * * is 6? 6P - P - PC. We can then define a measure for the real value of the debt, expressed in terms of foreign goods but evaluated at 1980 cross-currency exchange rates: (2) bc = B$c/Pc This is the measure used in Section C.3 of Chapter II (see the third column, labeled (A), in Table 2.2, and also Table 2.3 in Chapter II). To isolate the impact of cross-currency exchange rate changes (CCE), the expression in equ. (1) is decomposed further: (3) b = (bc + bb )ec/y = bce/y + bb eC/y =b + bbe/y c c  3 The first term in equ.(3) equals the debt-output ratio in constant TL, ccrrected for the effects on the dollar value of the debt of any change in cross-currency exchange rate since 1980. e is the real exchange rate evaluated at 1980 cross-currency exchange rates: eC = EPc/P. The second term in equ.(3)measures the cumulative impact (since 1980) of all such * cross-currency changes. bc is the measure used in the external debt decomposition presented in Chapter II, Section C.3. B. Debt Dynamics. Outout Growth and the Current Account A 2.4 The ratio of external debt to GNP can increase because of increased resource transfers from the rest of the world to the borrowing country, because of interest payments on past debt at a real rate in excess of the real growth rate of the economy, or because of capital losses incurred on outstanding debt as a result of depreciation of the real exchange rate. Chapter II uses a decomposition method designed to bring out the extent to which each of these factors has contributed to the actual changes in the debt-output ratio. A 2.5 The decomposition starts from the identity that the current account as a share of GNP equals the change in the value of external debt, excluding capital losses due to real exchange rate depreciation and to cross-currency exchange rate changes. Changes in the real exchange rate will be brought into the analysis below, but that is not easily possible with CCE effects. These effects are therefore eliminated from the analysi by a* * applying the decomposition to bc =bcec/y instead of to b . The result is:  4 (4) ebc = cad e is the real exchange rate. e=EP /P. where E is the nominal exchange rate against the dollar and P a domestic price index. cad is the current account deficit in real terms (i.e. local currency value of the current account deficit deftated by P). In addition, cad is a * defined to only include real interest payments r b e; the inflation component in foreign norrnnal interest rates, P b, is included in the capital account, in line with the approach to domestic debt used in Chapter III and in Annex III. A 2.6 However, a more fundamental concept of external balance involves the non-interest current account. The increase in foreign debt does not measure the net resource transfer received from foreigners: against it must be set interest payments made on the debt. But if one subtracts interest payments from both sides of equ.(4), it becomes clear that the net resource transfer equals the non-interest current account deficit (nicad/ rather than the current account deficit itself: (5a) e(b- r b )=cad - r b e = nicad Using tildes () to indicate variables scaled by GNP and expressing (5a) in terms of GNP yields: b - (5b) e( r b )=nicad y  5 where nicad=nicad/y and b =eb /y. Finally, straight differentiation yields for the change in the debt-output ratio b (6) b = b e/y -nb + eb. n is the growth rate of real output. e is the rate of change of the real exchange rate e. A 2.7 Using (6) to substitute out b from equ. (5b) gives the decomposition formula underlying much of the analysis of Chapter II: (7) b = nicad + (r -n)b + eb . (7) isolates the three components underlying any increase in the debt-output ratio: (a) the deficit on the non-interest current account or the net resource transfer received from abroad; (b) the "debt-dynamics" term measuring the extent to which interest payments offset the negative impact of real output growth on the debt-output ratio; and (c) capital losses on foreign debt due to depreciation of the real exchange rate. A 2.8 TNe formula also shows which factors are important for sustainability of any given non-interest current account deficit. Consider for simplicity the case where e = 0, and e = 1 by choice of units. (7) tien becomes: (8) b =nicad + (r -n)b Define the non-interest current account deficit that will just hold the debt-output ratio  6 constant as nicad. Clearly, (9) nicad =-(r - n)b Clearly, if r < n, then nicS < 0. A 2.9 If the interest rate exc',eds the growth rate of the economy, only a surplus on the non-interest current account is compatible with a constant debt-output ratio. Any non-interest current account dfiicit, and in fact any surplus less than (r -n)b will lead to explosive debt growth, with interest payments rising faster than GNP. This will eventually cause insolver.cy, as the next section shows. A 2.10 One final issue, before turning to the next section. Since we used a measure of real debt corrected for both the CCE effect on the nominal dollar value of debt and for the CCE effect on the price level, the real interest rate too needs to be corrected for these effects. This needs to be done in two steps. First, evaluate the average nominal interest rate at 1980 cross-currency exchange rates, to eliminate the impact of CCE effects on the dollar value of the debt. Second calculate the real rate using an inflation rate based on P to eliminate the CCE impact on the price level used to deflate the debt: r =(1+i )/(1+P )-1. C. Solvency and Creditworthiness A 2.11 It is important to distinguish between the concepts of solvency and  7 creditworthiness. Solvency involves the ability to pay; creditworthiness, on the other ha(jl. involves both the ability and the willinpess to pay. We discuss each concept in turn. C.1I Sol vellcv A 2.12 A country does not need to reduce the balance of its debt to remain solvent. Strictly speaking, solvency requires that the debt grows at a rate less than the rate of interest, a very weak condition indeed. This condition can be shown to be equivalent to the requirement that the discounted value of current and future consumption does not exceed the discounted value of current and fLtture output net of investment, minus initial debt. An equivalent condition is that the discounted value of current and future surpluses on the non-interest current account is at least as large as the initial debt. A 2.13 A more casual way of stating this is that a country should at least devote some of its resources to debt service; the increase in debt should not exceed interest payments. This condition rules out chain letter type Ponzi games. The equivalence of all these conditions is demonstrated below. A 2.14 Solvency requires the discounted value of total (public and private) consumption ct not to exceed the value of output minus investment (also tota!. public and * private), Yt- It, mqinus initial debt bo: (10) J exp(-r t)(yt- It)dt -b = J exp(-r*t)ctdt The non-interest current account equals nationally generated output, including remittances, minus expenditure, excluding interest payments on external debt:  8 ( 11 ) nicad t = -( Yt- It - ct ) Hence we can rewrite (10) as: (12a) Jexp(-r t) nicadt dt = -b; (12a) shows that solvency requires that the current and discounted value of the surplus on non-interest current account (minus the deficit) equals at least the initial value of the debt. A 2.15 Using the definition of the non-interest current account, equ.(5), allows one to reewrite (12a) as follows: (12b) J exp(-r t)(bt - r b)dt =-bo Integration of (12b) yields: (13a) limbtexp(-r *b0 b or (13b) lim btexp(-r t) = 0. t-4 (13b) implies the statement made in the introduction to this section, that the debt eventually will need to grow at less than the rate of interest for the country to remain solvent.  9 C.2 Creditworthiness A 2.16 The distinction between creditworthiness and solvency is unique to the issue of external debt. In the case of internal debt, say debt owed by a corporation, it is possible to seize a firm's assets through the legal system if the firm does not meet its debt-service obligations. If the value of the assets is high enough, the firm is effectively forced to pay. If the value of the assets falls short of the outstanding debt, the firm is bankrupt. Hence where domestic debt is concerned, creditworthiness and solvency are not really different concepts. A 2.17 However, foreigners will generally not be able to seize domestic assets on a significant scale, especially if those assets are the property of the debtor government. Hence the cost of default on external debt is generally less than the value of the debtor's assets. This implies that a country can fail to be creditworthy before it becomes insolvent. A 2.18 A practical problem is that it is not really possible to reliably assess the cost of default. However, if a country has not yet defaulted, the current burden of the debt clearly falls short of at least what the debtor country thinks is the cost of default. Otherwise it would in fact already have defaulted. Creditworthiness can thus be maintained by not allowing the burden of the debt to increase further (Cohen (1987)). A 2.19 The approach to creditworthiness taken here thus requires the debt service burden r b /R not to increase, with R a measure of the resource base of the country. In the approach proposed by Cohen (1985,1987) and used in this report, R is a weighted average of real exports and real GNP, both expressed in terms of foreign goods. nR is the S~~~~~~~~~~ growth rate of R. r is the foreign real rate of interest. For a constant real debt burden, the following must hold by definition:  10 (14) b -R=O. Also, the rate of increase in foreign debt equals: (15) b = nicad/(eb ) + r from equ. (5a). Combining (14) and (15) yields the restriction on the non-interest current account this approach to creditworthiness requires: * * (16) nicad -(r -n)b e (16) implies that creditworthiness requires a non-interest current account surplus equal to the debt times the excess of the real interest rate over the real growth rate. This implies that as long as the growth rate is positive, a deficit on the current account, including interest payments, is in fact compatible with the constraints creditworthiness imposes, as (17a,b) show. If * * (17a) nicad =-(r -n)eb < 0 then: * * (17b) cad = nicad + r eb * = neb >0O.  11 Table A 2.1: CURRENCY COMPOSITION OF TOTAL NET EXTERNAL DEBT AT CURRENT EXCHANGE RATES (in percenc) US$ DM SF ST YEN FF SDR 1979 75.1 11.3 2.5 2.1 1.2 2.4 5.4 1980 63.4 18.1 2.3 3.0 1.7 2.7 8.8 1981 57.9 20.0 2.8 2.4 3.4 3.2 10.3 1982 55.5 20.2 2.8 2.0 3.6 5.1 10.9 1983 57.3 21.1 1.8 1.5 4.0 3.4 10.8 1984 56.8 20.3 6.1 1.4 4.6 2.3 8.5 1985 55.8 23.8 5.6 1.5 5.0 2.4 5.9 1986 53.4 26.7 5.4 1.1 7.6 2.1 3.6 CURRENCY COMPOSITION OF TOTAL NIET EXTERNAL DEBT AT 1980 EXCHANGE r4ATES (in percent) . ..................................................... ............................................ USS DM SF ST YEN FF SDR ............................ ....................... ............................................. .... ........................ 1979 7o.1 10.3 2.3 2.3 1.5 2.2 5.3 1980 63.4 18.1 2.3 3.0 1.7 2.7 8.8 1981 55.5 21.4 2.7 2.8 3.4 3.8 10.5 1982 51.6 21.6 2.7 2.7 3.6 6.7 11.1 1983 52.0 24.2 1.9 2.0 3.8 5.2 10.9 1984 49.8 24.9 6.7 2.2 4.4 3.7 8.4 1985 52.1 26.1 5.7 2.2 4.3 3.6 6.0 1986 53.7 26.7 4.9 1.7 6.0 3.1 3.8 ..... ......................... ................................................... .................................. Source: Undersecretariat of Treasury and Foreign Trade, and Central Bank.  12 Table A 2.2: CURRENCY COMPOSITION OF NET PUBLIC EXTERNAL DEBT AT CURRENT EXCHANGE RATES (in percent) . ...................... ........................................................................ ............... US$ DM SF ST YEN FF SDR .................................................................... 1979 71.5 12.7 2.8 2.4 1.4 2.8 6.4 1980 63.1 18.2 2.3 3.1 1.7 2.7 8.9 1981 57.0 19.9 2.9 2.5 3.4 3.2 11.1 1982 55.1 20.2 2.8 2.1 3.6 5.1 11.2 1983 56.9 21.1 1.8 X.5 4.1 3.4 11.3 1984 55.9 20.2 6.3 1.5 4.8 2.3 9.0 1985 53.6 23.9 6.1 1.7 5.4 2.7 6.7 1986 51.5 26.3 6.0 1.2 8.5 2.4 4.2 .................................................................... CURRENCY COMPOSITION OF NET PUBLIC EXTERNAL DEBT AT 1980 EXCHANGE RATES (in percent) ................................................................................ US$ DM SF ST YEN FF SDR .................................................................... 1979 72.4 11.5 2.6 2.7 1.7 2.6 6.4 1980 63.1 18.2 2.3 3.1 1.7 2.7 8.9 1981 54.6 21.3 2.7 2.9 3.5 3.8 11.2 1982 51.2 21.6 2.7 2.7 3.7 6.7 11.4 1983 51.7 24.1 1.9 2.0 3.8 5.2 11.3 1984 49.2 24.5 6.9 2.3 4.5 3.8 8.9 1985 50.3 26.0 6.1 2.4 4.6 3.8 6.8 1986 5L.8 26.2 5.5 1.9 6.7 3.4 4.4 .................................................................... Source: Undersecretariat of Treasury and Foreign Trade, and Central Bank.  13 Table A 2.3: CURRENCY COMPOSITION OF NET PRIVATE EXTERNAL DEBT AT cucR&zvr &4'Cr-GCE @RAs (in percent) US$ DM SF ST YEN FF ........................................................... 1979 93.4 4.4 0.7 0.6 0.3 0.6 1980 84.5 11.5 0.9 1.2 0.7 1.1 1981 69.8 20.4 2.3 2.0 2.8 2.6 1982 68.8 20.5 2.2 1.6 2.8 4.0 1983 69.2 22.2 1.5 1.2 3.2 2.8 1984 70.5 22.8 2.8 0.7 2.2 1.1 1985 71.5 23.0 2.1 0.6 1.9 0.9 1986 65.1 29.6 1.8 0.3 2.5 0.7 .................. ..................................................................... CURRENCY COMPOSITION OF NET PRIVATE EXTERNAL DEBT AT 1980 EXCHANGE RATES (in percent) ... ............... ........ .......................... ............................. US$ DM SF ST YEN FF . ..... ...... ........................................... .................................... ........... 1979 93.9 3.9 0.6 0.6 0.4 0.6 1980 84.5 11.5 0.9 1.2 0.7 1.1 1981 66.7 22.4 2.3 2.4 2.9 3.2 1982 63.8 23.1 2.3 2.2 3.0 5.6 1983 61.5 27.4 1.6 1.7 3.3 4.5 1984 60.0 31.1 3.5 1.2 2.3 1.9 1985 66.6 26.9 2.3 0.9 1.8 1.5 1986 65.4 29.4 1.6 0.6 2.0 1.0 ......... ........................................................................................ .......... Source: Undersecretariat of Treasury and Foreign Trade, and Central Bank.  14 ANNEX III CONSISTENCY REQUIRENIENTS FOR FISCAL POLICY' A. Introduction A 3.1 This annex presents and applies an integrated framework to assess t he consistency between fiscal deficits and other macroeconomic targets, in particular output growth and the rate of inflation. The model centers around the government budget constraint and can be used to either derive the financeable deficit given inflation targets, or to derive a sustainable inflation rate for which no fiscal adjustment would be necessary. It can assess the impact of financial sector reforms affecting money demand, of changes in interest rates paid on foreign and domestic public sector debt, of output growth targets, and of exchange rate policy on the trade-off between fiscal adjustment and sustainable inflation rates. The analysis furthermore incorporates an approach, due to Daniel Cohen (1985, 1986), to derive a sustainable external debt policy. The model is applied to an analysis of inflation, external debt and financial sector reform. A 3.2 The framework is based on the public finance approach to inflation. In this approach (see Phelps (1973) for an early analysis along such lines), inflation is interpreted as the residual tax. Residual because it restores balance between the expenditure plans of the public sector, its debt management and revenue from traditional sources of taxation. Inflation acts as a tax because it forces the private sector to reduce expenditure just to maintain the real value of money balances it desires to hold for a given rate-of-return structure. It drives a wedge between income and expenditure not offset by any real 'This annex draws on three background papers listed in the references at the end of this Section: Anand and van Wijnbergen (1987), Rocha and van Wijnbergen (1988) and van Wijnbergen and Anand (1988).  15 accumulation of assets, just like any other tax does. Also, since its counterpart is (Central Bank advances to the public sector, it allows the public sector to cover a matchilng excess of expenditure over more traditional sources of revenue. A 3.3 This approach does not deny the fact that, in the short run, demand pressure or cost-push factors such as nominal exchange rate changes may be more important determinants of inflation. However, such factors contribute little to the understanding ol sustained inflation. Excess demand pressure, as Friedman points out in his celebrated presidential address (Friedman (1968)), should lead to accelerating rather than sustained inflation. Cost-push factors such as wage or exchange rate changes can explain price level shifts but not really sustained inflation. Of course continued nominal devaluation could explain a matching excess of domestic over foreign inflation, but such an explanation begs the question of whet is behind the apparent need to resort to such a continued policy of nominal devaluation? A similar objection can be raised against strict monetarist explanations of inflation. The claim that sustained inflation is not possible without a matching growth in nominal balances is almost tautologically true since otherwise the real money stock would go towards zero or infinity. It however begs the question of what drives such a process of sustained monetization. A 3.4 To explain sustained inflation rates, an analysis of the fiscal implications of inflation is necessary, and this is what the fiscal approach to inflation does. This approach has received renewed attention after an influential paper by Sargent and Wallace (1982), who used this view on inflation to explain paradoxical links between inflation and money growth sometimes observed in practice. The concept of sustainability plays an important role in their analysis. Sustainability of inflation targets requires consistency between those inflation targets and their implied consequences for revenues from the inflation tax on the one hand, and on the other hand the public sector's surplus of expenditure over other  16 sources of revenues (see also Buiter (1985)). This analysis has been extended to the open economy so as to allow discussion of exchange rate policy in Drazen and Helpman (1986) and van Wijnbergen (1986). B. Analytical Framework B.1 Fiscal Deficits, Money Creation and Debt. A 3.5 This Section explores the link between fiscal deficits, money creation and debt. First, the analysis focuses on the precise measurement of the public sector's net claim on real resources ( the "real" or "operational" deficit), and on the ways it can be financed. Then the implications for feasible financing of given macroeconomic targets for inflation and output growth, of changes in financial sector regulatory policies and of interest rates abroad and at home are explored. A 3.6 To derive the relation between fiscal deficits, money creation and debt, start from the following relation: (la) D+iB +i B E=B+B E+DCg = B + (B -NFA*)E + DCg + NFA E The left hand side of (1) lists the expenses (net of taxes) of the public sector: its non-interest deficit D plus nominal interest payments on domestic and foreign debt. i (i ) is the nominal interest rate on domestic (foreign) debt B (B ). E is the nominal exchange rate (TL per dollar). These expenses need to be covered by the issue of domestic or foreign  17 debt (B and B ), plus Central Bank advances to the public sector, DC The non-interest deficit D and the interest payments should include the non-interest deficit and interest obligations of all government entities: the central government, state enterprises, municipalities, local governments, and extra-budgetary funds. The proper treatment of the Central Bank is discussed below; its obligations too should be included. A 3.7 The second line of equation (1) regroups tne sources of financing somewhat, to bring out the link between money financing and public sector deficits. Increases in base money, M, equal the increase in net domestic credit to the government, DCg, and increases in net foreign assets of the Central Bank, NFA E, minus the increase in the Central Bank's net worth. Therefore base money can be seen as a source of revenue to the government. This analysis also suggests the proper definition of money for an analysis of deficit finance. Clearly, revenue derived from inflationary erosion of the private sector's deposits that is offset by inflationary erosion of loans outstanding to the private sector does not increase NET revenue. Hence the need to use base money or "inside money" only ( See Anand and van Wijnbergen ( 1987) on how to deal with such real world complications as Central Bank's credit to the private sector and so on). Reserve or base money equals currency in circulation and ( net) reserves held by commercial banks at the Central Bank. These components form the net liabilities of the Central Bank towards the private sector. However, counting all of the Central Bank's liabilities (base money) as a public sector liability means that of the Central Bank's assets, the claims on non-government agents need to be subtracted from the public sector's debt. In particular, the public sector's foreign debt needs to be measured NET of the Central Bank's foreign assets (i.e. public ** sector foreign debt equals B -NFA). A 3.8 If in addition the Central Bank's profit and loss account is included into the  18 definition of the public sector. the link between deficits and money rinancing becomes watertight. In a simplified set-up, the Central Bank's profits consist of interest earnings oii reserves, i NFA , which are added to net worth: i NFA E = NW. Subtracting that simple expression from the budget identity yields: * *I (Ib) D + i B + i (B-NFA) E= B + (B -NFA)E + DCg + NFA E-NW = B+ (B -NFA )E + M A 3.9 The importance of the latter correction cannot be stressed enough. Examples are rife where countries run a balanced budget, sometimes under the force of a constitutional amendment, but in fact continue deficit spending by effectively shifting Treasury expenditure to the Central Bank. In Turkey, a substantial part of the interest payments on the Central Government's foreign debt is handled by the Central Bank, without being recorded in the Central Government's budget. A 3.10 Further modifications to (1) are necessary to tighten the link between the public sector's net claim on resources to increases in the real value of domestic and foreign debt, to inflation and to money creation. The most important one involves incorporating: capital losses due to inflation and exchange rate changes into the accounting framework. A 3.11 Making this correction to equ. (1) results in the following link between deficits and inflation (see Anand and van Wijnbergen (1987) for a more detailed analysis of the steps involved):  19 * * g 2) d + rb + (r +e)(b -nfa )e = b + ((b -nfa )e) + M/P = b + ((b -nfa )e) + m + Pm Lower case letters denote real variables, so d = D/P, the real value of the non-interest deficit; P (P) is the domestic (foreign) price level. P (P ) is the corresponding domestic (foreign) inflation rates. b ( b) is the real value of domestic (foreign) debt in terms of domestic ( foreign) goods, and m is the real money stock. nfa is the real value of the banking system's net foreign assets in terms of foreign goods. r ( r) is the real rate of interest: r = i - P, r = i -P (this formula is only :n approximation when used on discrete time data; in the applied part of this report we use the exact formula (1+r)=(1+i)/(l+P)). e is the real exchange rate, e = EP /P. e is the rate of depreciation of the real exchange rate. M is the nominal, and m the real money stock: m = M/P. A " " indicates changes in the variable below it. Finally, n denotes the real growth rate of the economy. A 3.12 Equ (2) states that the fiscal deficit, inclusive of the Central Bank's profit and loss account, but counting REAL interest payments only, equals changes in the real value of domestic and foreign debt, plus revenue from the inflatior tax ( Pm) and from seignorage ( nm). The sum of the latter two equal revenue from monetization, M/P. Outside steady state there could be other sources of revenue from monetization than these two: once-off changes in the real money stock because of changes in inflation or interest rates, or similar once-off changes due to financial innovations shifting money demand. A 3.13 Equation (2) as it stands is nothing but accounting, but it is at the basis of  20 most of the analysis to come, Macroeconomic variables such as growth and inflation hlave implications for the amount of money the private sector is willing to absorb for givezl interest rates. Similarly, changes in financial structure and regulation will also have aii impact on the amount of revenue from monetization that can be expected. Finally, creditworthiness and sustainability imply constraints on the issue of interest bearing debt. All such restrictions can be incorporated in equ. (2). In the next two Sections, we consider first monetization and then debt issue. The next Section then puts it all together to derive the constraints on fiscal policy that consistency with financial structure and macroeconomic targets implies. B.2 Inflation Tax Revenue and the Structure of the Financial Sector. A 3.14 The previous section argued that the appropriate money concept to use is net Central Bank liabilities to the private sector, or reserve money. It is thus important to understand how demand for base money reponds to changes in finaacial sector regulation, to interest rates and to inflation. This creates a practical problem in that those liabilities are towards different agents in the economy. It is therefore unlikely that an aggregate demand function for reserve money would adequately capture the sensitivity of reserve money demand with respect to changes in the inflation rate, financial structure and interest rates. This report uses a more structural approach. Underlying this approach is a model describing private portfolio choice as a function of inflation, output growth and interest rates. This gives the amount of currency, demand deposits and time deposits the private sector is willing to hold given output, inflation and interest rates. This is coupled with a simple financial sector model incorporating reserve requirements and other bank regulatory policies to derive the demand for reserves by commercial banks. The demand for reserves is then added to the demand for currency already derived to get an estimate of the total demand for base money given inflation, interest rates and so on. All this is then used to  21 calculate revenue from monetization for different inflation rates, output growth rates. interest and inflation rates and different regulatory policies. A 3.15 This indirect, structural approach has a number of advantages. It allows explicit calculation of the effects of changes in financial sector regulation on the financeable deficit through the impact of such regulation on the aggregate demand for base money. For the same reason, this approach is more likely to be stable across changes in financial sector regulation. The case of a simple fractional reserve banking system (i.e. a system where banks are required to hold a fraction of their deposits as -eserves) is discussed below. More complicated regulatory systems are incorporated as an option in the LOTUS model that is available on diskette as a companion to this report. See also Rocha and van Wijnbergen (1988) for more extensive discussion. A 3.16 To analyse the determinants of the demand for the primary components of reserve money entering into such a financial sector model, consider first a simple portfolio approach to private sector asset demand: (3a) (CuI(PY)) = fC(P,iDD,iTD) (3b) (DD/(PY)) = fDD(F,iDD,iTD) (3c) (TD/(PY)) = fTD(P,iDD,iTD) Demand for currency Cut demand or sight deposits DD, and time deposits TD, all as a share of nominal GNP (PY in the formulas), depend on inflation P, and the interest rates paid on demand and time deposits, iDD and iTD. More sophisticated financial structures would introduce additional factors. For example, with foreign exchange deposits available  22 one would expect exchange rate depreciation and foreign interest rates to influence demand for domestic assets. However, the time period since their introduction has been too short to allow econometric analysis of the influence of the factors. A 3.17 Under a tractional reserve system, with reserve requirement ratios RRDD and RRTD against demand and time deposits respectively, demand for base money M is: (4) M/(PY) = Cu/(PY) + RRDDDD/(PY) + RRTDTD/(PY) fcU(P,iDD,iTD) + RRDDfDD(P,iDD,iTD) + RRo QfTD(PjiDD,iTD) (4) can be used to derive the impact of changes in inflation, interest rates and financial sector regulation on base money demand and their likely impact on the revenue the public sector can expect from monetization. Combining this information with equ.(2) then allows assessment of the iiscal consequences of inflation and of financial sector reforms affecting say reserve requirements RRDD and RRTD or the interest rates paid on deposits. B.3 Creditworthiness and the limits on foreign borrowing. A 3.18 The cost and availability of foreign financing are clearly important determinants of what constitutes a consistent fiscal policy. Required total revenue depends on the cost of the existing foreign debt, r (b -nfa )e; in addition, for any total revenue requirement, the amount to be covered from domestic sources depends on the amount of  23 A 3.19 Assessing a country's room for external borrowing involves two considerations: solvency and creditworthiness. Solvency concefs ability to pay and is intricately linked to the non-interest current account, real inQest and output growth rates, and, of course, the initial level of debt. Solvency is assumed to be in jeopardy if the discounted value of an estimate of minimum current and future consumption levels exceeds current wealth net of foreign debt. In practice, one would calculate the discounted value of current and future feasible trade surpluses and compare that value with the current debt; in Section C.1 of Annex II it is shown that these two approaches are in fact equivalent. Cohen (1985) assesses the solvency of the major debtor countri, and concludes that in almost none of them solvency is at all at risk. A 3.20 Creditworthiness may however be a constraint even if solvency is not. Creditworthiness depends on the lenders' perception of a country'& ability AND willingness to pay. Therefore creditworthiness often imposes tighter constrahts than solvency alone. Repsyment not only requires a sufficiently high value of wealth to be able to pay, but also the generation of a surplus of traded goods production over traded goods consumption (net exports). This is likely to be much more burdensome in a 0ountry with most of its resources in the non-traded sectors of the economy than in an outward oriented economy. But if it is more burdensome, a country might be more tempted not to repay, even if solvency requirements are met. Hence the importance of debt-export ratios in the assessment of creditworthiness. A 3.21 Assessing the precise limits imposed by creditworthiness constraints is difficult for several reasons. First of all, while debt-export ratio are important, they are a biased estimate of the ratio of a country's debt to its output oa;tradeable goods. Some domestically produced traded goods are likely to be sold at how On the other iand, the debt-output ratio would be an underestimate, since GNP incorporates non-traded goods.  24 We follow an approach suggested by Cohen (1986). Cohen suggests the use of anl intermediate measure of a country's resource base, R, constructed in such a way that no incentive to drive a wedge between actual and social costs of foreign exchange remains, at least for the purpose of assessing creditworthiness. A 3.22 A second problem is that it is extremely difficult to assess the costs of default. Cohen (1986) suggests a simple pproach to this problem, by observing that, if a country has not defaulted yet, the current debt service burden must fall short of whatever the cost of default is. An obvious definition of a "prudent" debt strategy then is a borrowing path that will not, at any time, raise the debt service burden above its current value. This may sound almost tautologically true, but it has important implications. For example the trade surpluses that many Latin American countries had to run after 1982 in fact do not qualify under this approach. Clearly, being forced to run trade surpluses of around 8% of GNP raises the -burden associated with the current level of external debt and so might trigger default. A 3.23 The measure of resources R should be set up such that any improvement in the debt-output ratio as a consequence of a real appreciation will be just offset by the negative impact on the debt-export ratio. Define R as a weighted average of the real value (in terms of foreign goods) of exports X and of domestic output, also measured in terms of foreign goods, Y * * (5) R = 7yX + (1-y)Y with X (Y*) the value of exports (home output) expressed in terms of foreign goods: X = X/e (Y = Y/e). The issue then is how to choose t. Choosing -y such that R does not depend on the real exchange rate requires that the following holds:  25 dR dX dY (6) ' -I This leads to the following expression for ar -CY (7) e e ( x - Y*) x* V* ce (Ce ) is the elasticity of X (Y) with respect to the real exchange rate e, and iP is the export share in GNP: X = X /Y. A 3.24 A feasible external debt strategy that maintains creditworthiness at least at current levels then consists of a time path for foreign borrowing that will not include a rise in B /R. For later convenience we define the growth rate of R, nR: (8) nR __=_X__X* . + ( 1-7 ) ny* ( OXt + (1-7)) (-x + (1-7)) Note that R has been constructed such that R is insensitive to real depreciation. nX* ( nys) is the growth rate of exports (GNP) expressed in terms of foreign goods. nX* can in turn be linked to output growth in the countries Turkey is exporting to, n (9) nX* = cy* n X*~~~~~~~ Here fy* is the elasticity of X, Turkey's exports in constant "dollars", with respect to output in the countries Turkey is exporting to.  26 A 3.25 Combining all this leads to the following expression for feasible externial debt acumulation as a percentage of GNP, eb /y: .** (10) eb =eb nR Equations (8-10) allow empirical application if at least the various elasticities have been estimated. Econometric estimates of these elasticities and of the parameters of the asset demand functions are presnted in Section C. B.4 Putting it all Together: the Design of Consistent Fiscal Policies A 3.24 A consistent fiscal policy is defined as a policy that can be sustained over the medium term without compromising any other macroeconomic target and without reliance oin unsustainable debt finance. Unsustainable debt finance implies debt issue at a rate in excess of the growth rate of the resources available for eventual debt service. This is in fact a more stringent requirement than solvency if the real interest rate exceeds the real growth rate of the economy ( see Annex III on the concepts of solvency and creditworthiness). * A 3.27 Taking R as the resource base concept for foreign debt, and domestic output y as the resource base for domestic debt, we obtain the following restrictions on debt issue: ( 11 ) b = nb, b nfa = nR(b - nfa*)  27 n ( nR) is the growth rate of y ( R ). A 3.28. Inserting these restrictions on debt issue into equ. (2), and expressinig variables as percentages of GNP then yields the following expression for the deficit reduction required for consistency, RDR: (12) RDR={d + rb + r (b -nfa )e} - {nb + n*R(b -nfa + nm + Pmn or, RDR equals the actual deficit inclusive of real interest payments on foreign debt, minus the financeable deficit. A deficit reduction equal to RDR will bring fiscal deficits in line with other macroeconomic targets in the sense defined before. Oince such an adjustment has been made, macroeconomic targets can be met over the medium term without any further need for fiscal policy adjustment and without unsustainable debt accumulation. Changes in macroeconomic targets (domestic growth rate, the rate of inflation, debt management policies), export prospects, real interest rates, and financial structure will influence RDR in ways that can be quantified using the model set up here. A 3.29 Note that if R is used as the benchmark to measure foreign debt accumulation against, real depreciation (e>O) does NOT affect fiscal balance: higher debt servicing costs are just offset by additional foreign borrowing under such a rule. But this will not be the case if either exports or output are chosen as benchmark. If the debt-output ratio is used as a benchmark, (12) becomes: (13) RDR = {d + rb + (r + e)(b - nfa )e} - {nb + ny(b -nfa ) + nm + Pm}  28 Clearly, under a debt-output rule, a real depreciation (e>O) increases the RDR, ie reduces the room for fiscal expansion. A 3.30 Under a debt-export rule, (12) becomes: (14) RDR = id + rb + (r + e)(b - nfa )e} - {nb + nX(b -nfa ) + nm + Pm} = {d + rb + r (b -nfa )e} - {nb + Cy n (b -nfa ) + nm - y * * * ' + ( 1 - ce )e(b -nfa ) e Where we used nX = Y*n* + fe e. Under this rule, real depreciation will relax the fiscal adjustment requirement if the export elasticity with respect to the real exchange rate exceeds one. C Empirical Reouirements A 3.31 Some empirical work is needed before the model set up in Section B can actually be implemented. First, the asset demand functions that form the input into the financial sector model need to be estimated. Section C.1 presents the results. Second, one needs the export elasticities with respect to price and income in recipient countries for the derivation of the resource measure used in the external debt part. These are presented in section C.2  29 C.1 The Structure of Money Demand in Turkey Specification A 3.32 The estimation of the demand for money in Turkey follows standard practice. The demand for each financial asset!" depends on income, the opportunity cost of holding the asset (interest rates on alternative assets), expected inflation and its own rate of interest. Sight deposits have offered positive, although moderate nominal interest rates. These are also taken into account in the estimation. The desired stock of money in period t is therefore written as: (15) log m*, * ba + bilog yt + b,iSDt + b3iTD, + b4P, + ut where m*, is the desired stock of money at time t, Yt is real output, iTD, is the interest rate on time deposits, iSDt is the interest rate on sight deposits, Pt is expected inflation and ut a random error term. Actual stocks are assumed to adjust to desired levels according to a standard stock adjustment function: (16) log mt - log mt_, = d (log m*& - log mt-,) Combining equations (15) and (16) produces: (17) 8log Mt ao + allog yt + a,iTDt + asiSDt + a4Pt + aslog mt,- + Ut That is, currency, sight deposits and time deposits. *5S43  30 where at = bid, i = L..4 and as = 1 - d. The as's are the short-run coefficients. The long-run coefficients are obtained by dividing the short-run coefficients by one minus the coefficient of the lagged term: bi - at/lI - as]; i = 1...4. The semi-logarithmic form implies variable interest rate and inflation elasticities along the demand curve. The demand for currency, signt deposits and time deposits in Turkey were estimated according to equation (17). The next section discusses the data. The final section discusses the estimation methods and the results. Data A 3.33 The data on monetary aggregates interest rates and prices are centered in the middle of the quarter (February, May, etc.) to avoid the year-end "window.-dressing" of balance sheets by commercial banks"'. The data on monetary aggregates--currency, sight deposits (of firms and individuals) and time deposits--are from the Quarterly Bulletin of the Central Bank. The domestic interest rates used in the estimations were the interest rate on sight deposits and six-month time deposits from the Quarterly Bulletin. The interest rates are expressed in percent per year, compounded and net of withholding taxes. The interest rate on six-month time deposits is the most representative interest rate on time deposits for this period. Six-month time deposits existed throughout the period and their return was usually the highestZ'. The Consumer Price Index (general) was used to lz This problem is particularly severe in the case of commercial sight deposits. 2/ By contrast, three-month time deposits were discontinued during 1983 and the effective interest rates on one-year time deposits were usually below the six-month deposits during most of the estimation period.  31 deflate the monetary variables and to measure the rate of inflation. Inflation was measured by the quarterly variations of the CPI. The CPI series are from the International Financial Statistics data base. A 3.34 Since there are no quarterly time series of real GNP, a proxy variable was used. To construct the proxy, first, annual real GNP was regressed on the annual output of electricity. Second, the regression coefficients were applied to the quarterly flows of electricity output. Finally, the quarterly flows in each year were proportionately scaled up or down, in order to equalize the sum of the quarterly estimates of real GNP and the observed annual flows of real GNP"'. Data on electricity output are from the Monthly Bulletin of the State Institute of Statistics. A 3.35 Data on exchange rates (TL/US $ and TL/DM) and foreign prices (US and German CPIs) are from the International Finar.cial Statistics database. Foreign interest rates (six-month euro-dollar and euro-mark deposits) are from Morgan Guaranty's World Financial Markets. The quarterly time series on population and the number of bank branches were both constructed by interpolation. Annual data on population are from the World Bank Atlas. Annual data on bank branches are from the Annual Reports of the Banks' Association of Turkey. The coefficient of correlation between real GNP and electricity output was 0.99 for the 1975-86 period. The improvement brought by the inclusion of additional variables such as cement output and oil imports was marginal.  32 Estimation and Results A 3.36 The OLS estimates of equation (17) with quarterly data are shown in Table 1. The equations were estimated in semi-logarithmic form in inflation and interest rates. The semi-logarithmic form performed better than the logarithmic form under a Box-Cox test "'. The proxy for expected inflation was a combination of current and past inflation rates with weights determined by a second degree polynominal without end-restrictionsa'. The upper half of Table 1 shows the estimates for the sample period 1977.1-1983.3. In the lower half of the Table, the sample period is extended to the fourth quarter of 1986. The sample period was not extended back to earlier periods for lack of homogenous time series of financial aggregates3. The equations were first estimated with a shorter sample to allow for tests of structural shifts due to the introduction of foreign exchange deposits in the fourth quarter of 1983. t' The deperdent variables were divided by their geometric means and the logarithmic and semi-logarithmic forms were estimated by OLS. The estimate of the semi-logarithmic form had a smaller residual variance. See, for instance, Maddala (1977). 2/ The equations were first estimated with a combination of the current and three lagged inflation rates. The introduction of four additional lags improved the performance of the equations. 3/ In 1977 the Central Bank introduced improvements in the collection and computation of monetary data. Since then, these data have been published in its Quarterly Bulletins.  33 Tabte 1: OLS 6STINAtES OF THE 'ESAJ FOR FINANCIAL ASSEtS Estimated Equation: Inmt - aO0 + aIlnyt + a2iTDe + a3iDDt + a4Pt + a5lnm t_i + 551CiD(I)t a/ ...............................I..................................I.............................................................. SA1LE O0 0 02 03 04 a5 r 12 K ow SER ................................................................................................................................ 197. 1 1983.3 1. Curr,ay -3.01 1.74 -0.77 -1.44 -0.97 0.53 -0.37 0.94 -1.33 0.0*3 (-1.28) (2.50) (-3.20) (-4.13) (-4.11) (4.39) 2. Indivdustlu -3.31 1.79 -1.63 -1.53 0.56 *0.5 0.98 -0.19 0.053 Siht Deposits (-1.57) (3.94 (-4.96) (-4.35) (5.09) *0.89 1.18 -1.56 0.40 -1.63 0.516 0.48 0.98 *0.13 0.052 (-0.32) (1.82) (-4.21) (1.30) (-4.69) (5.23) 3. FIre's Sight 4.01 0.9? -0.52 * -1.16 - 0.70 * 2.10 0.048 Ouposits (1.99) (1.92) (-3.08) (-4.19) 3.92 0.99 -0.52 0.17 -1.15 - 0.68 - 2.10 0.049 (1.62) (1.63) (-2.97) (0.06) (-2.83) 4. Total sight U/ 1.54 0.91 -1.07 -1.50 0.43 -0.4 0.97 -0.39 - 0.037 Deposits (0.93) (2.81) (-5.1) (-5.39) (3.62) 1.57 0.90 -1.07 0.01 -1.50 0.43 -0.4 0.97 -0.41 - 0.038 (0.75) (1.99) (-4.?7) (0.02) (-S.19) (3.50) 5. Time Oepoits 1.59 -0.00 0.93 - -0.74 0.79 - 0.99 0.33 - 0.061 0.54) (0.00) (2.86) (-1.07) (7.64) 1.55 0.04 0.92 -0.01 -0.73 0.79 0.99 0.42 - 0.055 (0.41) (0.00) (1.76) (0.02) (-0.71) (s.22) ............................................................... ...............................................................................................s .**@*Z 1977.1 - 1918.4 1. Curru- 3.563 0.571 -0.694 -1.068 -1.505 0.27a r 0.3 -.S3 * 0.069 (3.102) (2.42) (-3.6?) (-3.01) (-3.79) O1.7) 2. Irdiviasts 7.289 -0.479 -1.533 * -2.67 0.370 * 0.98 1.50 0.069 Sigt OpaIts (4.69) (-1.97) (-6.78) (-5.5) (3.90) 7.222 -0.530 -1.471 0.554 -2.437 0.397 - 0.98 1.70 - 0.066 (4.90) (-2.28) (-6.78) (2.05) (-5.450) (4.36) 3. Flrms Sight 6.919 -0.261 -0.300 - -1.305 0.257 0.7 -0.49 0.07T D_sits (4.41) (-1.41) (-2.92) (4.14) (1.82) 7.69 -0.050 -0.436 1.381 -1.536 * 0.7rr 1.86 0.056 (6.71) (0.17) (-3.94) (.92) (-5.72) 4. Total Sigt h/ 6.164 -0.382 -0.9W - -2.197 0.251 0.93 0.58 * 0.041 _eaits (6.70) (-2.64) (-8.17) (-7.93) (2.73) .899, -0.5t1 -0.964 0.599 -2.157 0.243 * 0.90 0.56 * 0.041 (6.71) (-2.66) (-7.97) (1.30) (-7.81) (2.67) S. Tim Oq. 1te 1.335 0.124 0.976 - -1.012 0.7s7 0.99 0.42 - 0.05S6 (1.96) (0.51) (5.04) (-2.09) 11.73) 1.298 0.277 1.181 0.383 -1.409 0.679 0.99 0.38 0.055 (1.94) (1.02) (4.61) (l.21) (-2.42) (7.46) ........ .... . ................ .................. .....................................................................................*_ ./ The co.ff cIis of seamt d1es are not show. bi The Intest ran m ight dpoits i ea w_eihed swersgs of the Interesr an sight deposits by firms w* individuls.  34 A 3.37 The estimated equations fit well the data for the shorter sample period". The coefficients have the expected sign and are usually significant. In addition, the coefficient of the lagged term indicates a relatively high speed of adjustment, except in the case of time depositsz/. The long-run income elasticities of the demand for currency and sight deposits are larger than unity, as is frequently the case in developing countries. That usually reflects the progressive switch from barter trade to monetized transactions. The estimated income elasticity of time deposits was zero for this period. Low or even negative income elasticities for time deposits are not unexpected. Given the wealth constraint, if the demand for some assets, such as currency and sight deposits, increase with the volume of transactions (proxied by income), then the demand for at least one asset must fall3'. A 3.38 The coefficient of the time deposit rate is negative and significant in the demand for currency and sight deposits, and it is positive and significant in the demand for time deposits. The coefficient on the sight deposit rate is negative and significant in the demand for currency. In the case of sight deposi6s, the coefficient has the expected sign, but it is not The methodology adopted in the analysis follows in good part previous work reported in Kopits (1987)--estimation with center of quarter data and construction of proxies for the output and expected inflation variables. We are also grateful to Kopits for the provision of data on interest rates and helpful discussions in the preliminary stages of the work. The results shown in the upper half of Table 1 are broadly in line with those reported by Kopits for the 1977.1 - 1982.4 period. The differences originate mostly from the choice of the price index (Kopits employs the WPI). Also, the analysis here provides separate estimates of the demand for currency and sight deposits and explores the significance of the interest rate on sight deposits. a2 In the case of the firms' demand for sight deposits, the coefficient of the lagged terms was not significantly different from zero. 3/ See, for instance, Friedman, B. (1978).  35 significant. Finally, the inflation coefficient has the expected sign and is significant, except in the case of time depocits. The fact that both the interest rate and inflation rate coefficients are usually significant reflects in part the very low correlation between these two variables during the sample period. Interest rates were controlled at low levels until mid-1980, while inflation rates were very high during the late 1970s and the first semester of 1980. Between 1980 and 1982 the Government adopted a more flexible interest rate policy and in 1982 liberalized the rates. In 1983 interest rates were set by a banks' cartel, which did not adjust them to increasing inflation in that year. The Government resumed control of interest rates in late 1983, while announcing the policy of adjustments of interest rates in line with inflation. However, the relationship has never been a close one. Although real interest rates have been generally positive since 1983-84, they have also oscillated significantly, reflecting the relative stickiness of interest rates vis-a-vis inflation rates"'. These factors justify the joint inclusion of interest rates and inflation rates in the estimation of the demand for financial assets in Turkey. A 3.39 The estimated coefficients of the demand for currency and sight deposits change significantly when the sample period is extended to the fourth quarter of 1986, as shown in the lower half of Table 1. The income elasticity The estimation of a Fisher equation for the period 1981.1-1986.4 produced the following results i, - 0.39 + 0.06 pt R , = -0.03, where it is the (77.7) (4.35) interest rate on six-month time deposits at quarter c and pt.i the CPI inflation rate in the two following quarters. The well-known error-in-variables problem in this rational expectations estimation was avoided through instrumental variables estimation (four lags of inflation were used as instruments).  36 in the demand for currency drops to one third of its former value, although it remains positive and significant, while in the case of individuals' and firms' sight deposits, the income coefficient becomes negative. The changes in the interest rate coefficients are not dramatic. The inflation coefficients increase significantly in the case of currency and the the individuals' demand for sight deposits. Finally, the coefficient of the lagged dependent variable decreases in the case of currency and the individuals' demand for sight deposits, indicating an increase in the speed of adjustment of actual to desired stocks. In the case of time deposits, the changes in the coefficients don't seem to be significant. A 3.40 The performance of F-tests confirms the existence of structural shifts in the demand for currency and sight deposits"". These shifts seem to be captured mostly in the income coefficient2'. The shift away from currency may partly be due to the increasing use of sight deposits in banks rather than cash for transactions. It may also partly reflect currency substitution. The major cause of the shift in the firms' demand for sight deposits is probably the increased use of repurchase agreements for cash management. Finally, the decreased demand for sight deposits by individuals probably reflects a substitution of foreign exchange deposits for TL deposits. The introduction of foreign exchange deposit accounts in end-1983 may have accelerated the ongoing process of currency substitution, although the increased stock of foreign exchange deposits also reflects tianefers of ~' The insufficiency of observations precluded the performance of a Chow-test. The stability of the estimated coefficients was tested against the increased sample size. See Johnston (1984). t' Shifts in the demand for money in the US during the 1970s also produced a decline in the income elasticity. See Garcia and Pak (1979).  37 Table 2: LONG-RUN COEFFICIENTS _-___________ ------------------ --------- --------- --------- --------- SAMPLE INCOME iTD iDD p _-___________ ------------------ --------- --------- --------- --------- Currency 3.70 -1.63 -3.06 -2.06 Demand Individuals 2.68 -3.54 0.91 -3.70 1978.1-1983.3 Demand Firms 0.97 -0.52 - -1.16 Demand 1.578 -1.877 _ -2.631 Time Deposits 0.00 4.428 _ -3.523 _-___________ ----- -- -- --------- --------- Currency 0.789 -0.958 -1.471 -2.079 Demand Individuals -0.878 -2.439 0.918 -4.041 1978.1-1986.3 Demand Firms -0.352 -0.404 - -1.756 Demand -0.510 -1.319 _ -2.933 Time Deposits 0.510 4.016 -4.164 -