Interest rates, growth, and external debt : the macroeconomic impact of Mexico's Brady deal

Policy Research WORKING PAPERS Debt and International Finance International Economics Department The World Bank June 1993 WPS 1147 Interest Rates, Growth, and External Debt The Macroeconomic Impact of Mexico's Brady Deal Stijn Claessens Daniel Oks and Sweder van Wijnbergen The debt-relief package worked in Mexico because it reduced uncertainty, not because it reduced transfers. The Policy ResearchWorking Papers disseminate the findings of work in progress and encourage the exchange of ideas among Bank staff and all others interested in development issues. Thesepapers, distributed by the Research Advisory Staff, carry the names ofthe authors, reflect onlytheirviews,and shouldbeused and cited accordingly. Thefindings,interpretations. and conclusions aretheauthors'own.They should not be attributed to the World Bank, its Board of Directors, its management, or any of its member countries.  Poiicy Research Debt and International Finance WPS 1147 This paper -a joint product of the Debt and Intemational Finance Division, International Economics Department, and the Country Operations Division 1, Country Department II, Latin America and the Caribbean - is part of a larger effort in the Bank to ziudy the impact of debt reduction on developing countries. Copies of the paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Rose Vo, room S8-042, extension 33722 (June 1993, 27 pages). Interest rates fell sharply after Mexico's Brady "debt overhang" hypothesis (the benefits to deal, and private investment and growth recov- growth of a reduced tax burden) as the dominant ered. factor. Claessens, Oks, and van Wijnbergen show Their main conclusion: Debt reduction can that the main benefit of debt relief was not to have a much greater impact than the magnitude lower expected payments but to reduce uncer- of relief, coupled with standard growth models, tainty. Reduced uncertainty was found to be the would suggest. The secondary effects on private dominant factor in explaining the positive investment of reduced uncertainty about govern- macroeconomic response (largely because of its ment policy is likely to be more important than favorable effect on exchange rate crises). the direct amount of debt reduction itself. Econometrically, they find that the variabil- But private investment is unlikely to increase ity of the future net transfer had a significant if uncertainty remains about future domestic impact but the average of the future net transfer macroeconomic stability and reform. The debt itself did not. package would not have succeeded if the govern- ment had not put through a successful domestic Their results confirm that debt reduction has refonn program before the debt relief package. a positive macroeconomic effect, but reject the The Policy Research Working Paper Series disseminates the findings of work under way in the Bank. An objective of the series is to get these findings out quickly, even if presentations are less than fully polished. The findings, interpretations, and conclusions in these papers do not necessarily represent official I .nk policy. Produced by the Policy Research Dissemination Center  INTEREST RATES, GROWTH AND EXTERNAL DEBT: THE MACROECONOMIC IMPACT OF MEXICO'S BRADY DEAL by Stijn Claessens World Bank, Daniel Oko World Bank, and Sweder van Wijnbergen University of Amsterdam and CEPR This paper was written while van Wijnbergen was at the World Bank.  INTEREST RATES, GROWTH AND EXTERNAL DEBT: THE MACROECONOMIC IMPACT OF MEXICO'S BRADY DEAL Non-Technical Summary On March 10 1989, the US Secretary of the Treasury Brady announced US support for external debt reduction (the "Brady plan"). The rationale for supporting what amounts to breach of contract was that debt reduction was considered easential for a restoration of growth and economic stability in the most indebted countries. The Mexican debt package, concluded in 1989 and implemented March 1990, was a spectacular success from that perspective. Interest rates on local currency debt fell by 20 percentage points within days of the agreement, private investment has boomed ever since and economic growth took off for the first time since the debt crisis started in 1982. The main channel suggested by theory for an impact of debt relief on future growth is a reduced tax burden (the "debt overhtng" hypothesis). But even on the most generous of assumptions, this seems implausible. In net present value terms Mexico received about $12 billion debt relief. This would on a permanent basis allow at most a 1.5 percentage point cut in the corporate tax rate.! Econometric evidence suggests that this falls far short of explaining the observed private investment boom of on average 14% real growth for the two years fcllowing the Brady deal's implementation. Much has been made of the impact of the deal on "investor confidence". In this context it is important to notice that the transfer Mexico has to make was not only reduced, but also smoothed out. Thus the specter of recurring crises associated with particular peaks in repayment obligations lost much of its threat as the new debt service obligation schedule was not only less 1/ Mexico receives about 3.5% of GDP at a 35% corporate tax rate. Due to a comprehensive definition of the tax base and the absence of almost all special creditse. average and marginal rates are close. Assuming a conservative 3 percent growth-adjusted real interest rate (real interest rate minus real growth rate) yields the claim stated in the text.  onerous in discounted value terms, but also had a much lower variance. In this paper, we assess the impact of the Brady deal through the channel of increased smoothness of debt service, and explicitly contrast it with the debt overhang hypothesis. We fJnd that reduced variance explains most of the effect of the deal, and that a proxy for the debt overhang effect has no explanatory power at all. These results confirm the potentially beneficial macroeconomic effects of debt relief, but reject the debt overhang hypothesis as an explanatory factor for them. Econometric evidence presented in this paper indicates that the impact of debt relief on uncertainty is the most important channel through which debt relief influences the macroecoromy. We furthermore sharpen the results by identifying specifically the favorable impact on uncertainty about future exchange rate crises as the dominant mechanism in explaining the growth and investment response to debt relief. The most important conclusion that follows from our results is that the likely impact of debt service relief can be much larger than the magnitude of the relief coupled with standard growth models would suggest. The secondary effects on private investment through reduced future policy uncertainty are likely to be more important than the direct amount of the relief itself. Another important result follows also, almost as a corollary; these secondary effects through reduced uncertainty will obviously not come into play unless other, potentially dominant sources of future policy uncertsinty have been removed first. Thus the successful domeatic reform program that the Mexican government put through in the years Preceding the debt package was a necessary precondition for the debt package to be successful. In that sense the debt package was so successful because the Mexicans had been so successful in reforming their economy first; without that the package would not have had such a spectacular effect.  1 1. INTRODUCTION On March 10 1989, the US Secratary of the Treasury Brady announced official support for external debt reduction (the "Brady plan"). The rationale for supporting what amounts to breach of contract was that debt reduction was considered essential for a restoration of growth and economic stability in the most indebted countries. The Mexican debt package, concluded in July 1989 and implemented March 1990, was a spectacular success from that perspective. Interest rates on local currency debt fell by 20 percentage points within days of the agreement, private investment has boomed ever since and economic growth took off for the first time since the debt crisis started in 1982. Such success was not widely anticipated (cf. Dornbusch and Modigliani (1989) for a skeptical assessment); there was a widely shared view that the amounts were simply not enough to make much of an impact. Mexico received about $4 billion cash flow relief per year, 2 of which would likely have come anyhow as amortization could have been expected to be rolled over. In a $200 billion economy, 2 or even 4 billion seems a small tail to wag a large dog.3 The optimistic prediction in van Wijnbergen (1989) was based on the assumption that real interest rates would drop by 20 percentage points; this assumption turned out to be correct, but as it was not itself endogenously linked to the debt deal, we are still left with a question mark about the exact mechanisms through which debt relief affects private sector investment and output growth. Oks (1991) also finds a significant impact of the Brady deal on Mexican 2/ De Long and Eichengreen discuss exactly the same issues in the context of the Marshall plan of 1948.  2 interest rates: a dummy for the debt deal shows up significantly in a risk premium equation. However, the use of a dummy once again leaves the exact mechanism unspecified. The main channel suggested by theory for the impact of debt relief on investment, ani from there on future growth, is the removal of debt overhang through a reduced tax burden (Helpman (1989), Krugman (1989), Sachs (1990)). But even on the most generous of assumptions, this seems implausible. In net present value terms Mexico received about $12 billion debt relief (van Wijnbergen (19911). This would on a permanent basis allow at most a 1.5 percentage point cut in the corporate tax rate.y Econometric evidence suggests that this falls far short of explaining the observed private investment boom of on average 14% real growth for the two yea- s following the implementation of the Brady deal.y Much has been made of the impact of the deal on "investor confidence" (Gurria and Fadl (1991), van Wijnbergen (1991)). In this context it is important to notice that the external transfer Mexico has to make was not only reduced by the deal, but also smoothed out. Thus the specter of recurring crises associated with particular peaks in repayment obligations lost much of its threat as the new debt service obligation schedule was not only less onerous in discounted value terms, but also had a much lower variance. The lengthening of debt maturity brought about by the 1989-90 debt restructuring 3/ The Government receives about 3.5% of GDP at a 35% corporate tax rate. Due to a comprehensive definition of the tax base and the absence of almost all special credits, average and marginal rates are close. Assuming a conservative 3 percent growth-adjusted real interest rate (real interest rate minus real growth rate) yields the claim stated in the text. 4/ van Wijnbergen (1989) presents econometric evidence on the sensitivity of private investment with respect , the after tax rate of interest.  3 was the key factor behind the sharp reduction in the variance of the net tr,.nafer to foreign creditors. The projected variance of the transfer was further reduced by the issue of fixed-interest rate debt instruments (par bonds). The importance of smoothness of public debt service obligations for financial stability has been stressed in the lir.erature on domestic debt and financial crises (Giavazzi and Pagano (1989)), but has not received attention in the literature on foreign debt. In this paper, we assess the impact of the Brady deal through the channel of increased smoothness of debt service, and explicitly contrast it with the debt overhang hypotnesis. We find that the reduction in the variance explains most of the effect of the deal. A proxy for the debt overhang effect has no explanatory power at all for the investment recovery, and the results indicate that the debt overhang effect played only a minor role in the reduction in interest rates. Another important result follows also, almoet as a corollary; the effects through reduced uncertainty will obviously not come into play unless other, potentially dominant sources of future policy uncertainty have been removed first. Thus tha successful domestic reform program that the Mexican government put through in the years Preceding the debt package was a necessary precondition for the debt package to be successful, while the reforms alone were not sufficient to allow for a recovery of investment. In that sense the debt package was so successful because the Mexicans had been so successful in reforming their economy first; without that the package would not have had such a spectazular effect. The remainder of this paper is organized as follows. In Section 2, we provide a brief overview of macroeconomic developments in Mexico before and  4 after the package, and a short description of the deal itself. Section 3 then presents a model formalizing the conjecture that the variance of debt service has an important impact on the macroeconomy. Section 4 describes the data used and the regression results. Section 5 concludes. 2. BACKGROUND: DEBT CRISIS, ADJUSTMENT AND THE DEBT REDUCTION AGREEMENT 2.1 Debt Crisis and Adjustment. Between 1950 and 1974, Mexico enjoyed high growth, low inflation and moderate external debt accumulation. This era of fiscal conservatism came to an abrupt end in the early seventies. The period of single digit inflation ended in 1973, the real excha:,e rate started to appreciate and the accumulation of external debt accelerated above the GNP growth rate. A financial and economic crisis in 1976 ended following majo. oil a-acoveries in 1977. The ensuing prosperity lasted until 1982, when soaring domestic inflation, falling international oil prices, rising world interest rates and massive capital flight ($21 billion in 1981-82 according to recent estimates; cf. Brideau, Eggerstedt and van Wijnbergen (1992)) led to a refusal by external creditors to roll over the principal payments falling due in that year of about $8 billion of Mexico's external public debt and subsequent reachedulings of Mexican external debt principal repayments. A sharp real depreciation of the Peso and unprecedented fiscal tightening wera necessary to effect a net transfer to foreign creditors of about 5% of GDP in the years following 1982. Economic growth was zero on average over the period 1982-1988. In late 1987 the Governmont announced the  5 "Economic Solidarity P&ct" (Pacto), which called for accelerated structural reform, further tightening of fiscal and monetary policy, a freeze of minimum wagss and basic public and private sector prices, and a freeze of the nominal exchange eate against the U.S. dollar. The Pacto brought inflation down irom a three digit level in 1987 to 20% in 1989-1992. Since more than 90% of Mexico's US$100.8 billion foreign debt outstanding at the end of 1988 was public or publicly-guaranteed, pressures on the Government for reaching an agreement with foreign creditors were building up. High and volatile external transfers generated uncertainty about whether the future transfer burden could be met. This generated increased uncertainty about future exchange rate developments which, in turn, was translated into high interest rates on domestic debt. "Ex post" real interest rates were almcst 50% in the months before the debt accord was reached. With domestic public debt at approximately 20% of GDP in 1989, interest rates this high severely threatened the fiscal underpinnings of the stabilization program and led to a reduced level of private investment. 2.2 The Debt and Debt Service Reduction (DDSR1 Agreement. Against this background, debt negotiations started in February 1989, after earlier agreements with Mexico's bilateral (Paris club) and multilateral creditors had been reached. Mexico was the first country to negotiate and obtain a debt and debt service reduction (DDSR) agreement under the Brady Initiative. An agreement in principle was reached on July 23, 1989, and the final package was implemented on March 28, 1990 under which $48,231 million of commercial bank debt was restructured. Of this 47% was committed to a par  6 bond at unchanged principal but at a reduced, fixed interest rate c4 6.25%; 43% was committed to a discount bond at unchanged interest rates but with principal reduced by 35%1 and 11% to the new money option, which left both principal and interest rate unaffected but called for 25 cents new money per dollar committed. The package as a whole delivered aiout $12 billion debt reduction in discounted value terms (van Wijnbergen (lV91)). 2.3 Macroeconomic impact of the Debt Agreement. When the deal was announced, nominal interest rates on local currency denoml.nated debt dropped 20 percentage points within day". This implied annual savings on domestic public debt service of about 4 percent of GDP, several times the direct impact of the deal on foreign debt service. Pt the same tim., private investment rebounded after years of slow progress if any; real private investment, which had stagnated during 1982-1988 (on average 1 percent growth), grew by 11.2 percent over 1989-1991. 1989 also saw a recovery of economic growth; over 1989-1991, real growth averaged almost 4 percent. The most spectacular reversal took place in the capital account of the balance of payments; on various estimates, around $10 billion flight capital returned in late 1989 and 1990 (Gurria and Fadl (1991), Brideau, Eggerstedt and van Wijnbergen (1992)). And total private capital inflows averaged $21 billion annually in 1991-92. Interest rates on dollar denominated debt fell too, but by much less than rates on Peso denominated debt ("CETES"). On short term dollar linked instruments ("PAGAFE"), rates fell from around 20% t- 16% on an annual basis, or about a fifth of the drop in Peso rates (see Figure 2 on page 11). The  7 bigger drop in Peso denominated debt strongly suggests the ,.mportance of expectations about impending 'lance of payments crises and their impact on exchange rate oustainability rather than considerations of public solvency. 3. THE MODEL The model presented below is a simplified implementation of the ideas outlined in Giavazzi and Pagano (1989).V' Take the following situation. The private E actor has access to foreign and domestic lending and investment opportunities, and is risk neutral. Lack of creditworthiness prevents the government (Central Bank and Treasury combined) to borrow abroad. The government faces variable external debt service payments of B. in period t, t= 1,...T, and rolls over every period the total principal on its internal debt of D at & nominal interest rate, it. The government can finance the debt service payments Bt and it*D in three ways: from its primary surplus (PS), through domestic borrowings, and by a (temporary) increase in the domestic monetary base (monetization). While Mexico has made large fiscal adjustments over the 1980s, in the short-run the primary surplus can not easily be adjusted. This leaves Mexico with the choice between domestic debt financirg and monetization. We use a to denote the fraction of the foreign and domestic debt service remaining (after taking account of the primary surplus) that the government decides to monetize. Clearly a is a reflection of the type of monetary policy the government is 5/The setup here is somewhat different from Giavazzi and Pagano (1989) where the domestic debt service falling due is variable. Here, the main source of variability is the foreign debt service falling due. The results are identical since what matters is that the total debt service falling due net of the primary surplus is variable.  8 pursuing. Ex-ante, the private sector is, however, not perfectly informed about the conduct of monetary policy and regards a as a random variable, distributed uniformly, T1(O,a,.). The variability of debt service payments and uncertainty about the conduct of monetary policy can lead to two (or more) equilibria: in one equilibrium the private sector has confidence that there will be no crisis (N); and in the other(s) the private sector is convinced that a crisis is possible (C). The crisis is assumed to take the form of devaluation of the currei-y. More specifically, in the crisis equilibrium, the private sector believes, because the government may monetizes the share a of debt service due in excesp eof its primary surplus, that there will be a devaluation of x percent (assumed to be exogenously determined) with probability a. As the domestic currency is fully backed by reserves,@ a devaluation is more likely to happen when the government monetizes in order to service its debts. The larger the amount monetized, the higher the chance that the currency will devalue as reserves decline to too low a level (here the lower bound is assumed to be zero) and a speculative attack on the currency occurs. Which of the equilibria (C or N) will prevail will depend on whether the private sector (in rational manner) assigns ex-ante a positive or a zero probability to a devaluation. Both can, under certain circumstances, be rational expectations outcome. Assume that money demand is of the constant elasticity type: M = A*e-dt, where it is the domestic interest rate. Consider now what happens to money demand in both equilibria: in N, it will be equal to the foreign rate r-, and M = A*e-', which will equal the money supply, which equals the reserve level 6/We equate the two as we abstract from the domestic financial system.  9 R; in C, it will be the f' eign interest rate plus the expected depreciation, and Mc = A*e-09"+u). Moving from equilibrium N to C implies thus a sudden drop in money demand of A*e-r**(l-e-mm) = R*(1-e-0u) and a loss in reserves. In the crisis equilibrium, the government will also face a higher interest rate, r + fx, on its domestic borrowings, D, further increasing the chances of an actual devaluation (if the government monetizes the share a of debt service due in excess of its primary surplus). The fall in money demand and the (temporary) increase in the monetary base will create an actual foreign exchange crisis--the government has to devalue--when R*(1-e-m) + a(B, + (r*+irx)*D - PS) > R, or, rearranging, when a > aA(,r), where aA(wr) = R*e0"X/(Bt + (r*+,r)*D - PS). The probability of a devaluation is now defined by the implicit function Xr = Prob(a>aA(,r)), where Prob(a>a-(ff)) is equal to 1 - aA(w)/ac . Figure 1 plots the probability function P(a>aA(ir)) as a function of , for different levels of B,. (We use the following parameters: D = 80, o - 0.05, R = 12, x 0.5, r = 0.1, PS = 30, a = 1]. Whether the equilibrium with Fiure 1 a positive probability of a crisis 10 0.9 - 45-degree can be an outcome will depend on 08- B1=20 whether the line P(a>aA(V)) 80.7 -2=30 006. crosses the 450-line. For low 0.5 - B 2 values of B this will not be the 4X 0.31 case, as for B, - 20 in this 0.2 0.1] example, while for higher values 0.0 0 0.2 0.4 0.6 0.8 Pt  10 it is, as for B2 = 30.2' As a result, the probability of a devaluation for two different debt service obligations wi.ll always be higher than (or equal to) the probability of a devaluation for the mean of the two debt service obligations. In the example, the probability of a devaluation is zero for the mean of the two debt service obligations, B3 = 25, as the line P(a>a^(ir)) does not cross the 450-line, while the mean probability of a devaluation for B, = 20 and B2 = 30 is strictly larger than zero (about 0.32). This implies that domestic interest rates, which will reflect devaluations expectations, will be higher when the variability of debt service payments falling due is large. Any reduction in the variability in expected debt service payments falling due would then lower interest rates. And, because with irreversibility investment will depend negatively on the level of uncertainty (e.g., see Pindyck, 1991), it also implies that investment will increase when the variability in debt service falling due is reduced. 4. EMPIRICAL APPLICATION. 4.1 Data We look at three Mexican assets to assess the impact of debt service variability on asset prices. The most obvious example is physical capital; because of its irreversibility, the impact of variability-induced rate of return uncertainty is clear. So the first part of the econometric analysis looks at private capital accumulation. An index of private fixed investment is available with monthly frequency. 7/For high levels of B (as in the case of B2 = 20)), there will actually be two possible outcomes for w, a low a and a high v. There is nothing in the model that can determine which equilibrium will prevail.  11 The second asset we Figure 2 consider is CETES, which is the Interest Rates on Peso- and Dollar- Mexican local currency Denonminated Domestic Public Debt 120- equivalent of a T-bill at 28 10 l 100 |Brady Deal Annwouncement| days maturity. Clearly CETES 0 / 80 m m |Implem~entatlon| is a much more liquid asset ' 60 than investment in physical 40 plant; there are two reasons why even for such short term 1988 1989 1990 monthly assets fears of being locked in {-C 4ts () -+-PqOW-(dc4) -Po*sebd*valiw while a crisis unfolds are in fact not unfounded. The first point is that even 28 days may be too long to get out when a balance of payments crisis with associated exchange rate response threatens. Second, investors may end up getting locked in if capital controls are imposed during or even in the run up to a crisis. Mexico did impose capital controls once before in a period of crisis (August 1982). Mexico offers a unique opportunity to decompose the interest differential between CETES and equal maturity debt instruments in the US. For this we look at a third asset. The Mexican Government issues a debt instrument similar to CETES, but denominated in dollars (PAGAFE). We can thus split the differential between CETES and the US 1-month T-bill rate in a pure exchange risk component and a pure country risk premium. The first is captured by the CETES-PAGAFE differential (corrected for the preannouncedw rate of devaluation), and the second by the differential between PAGAFE and 8/In the empirical application we use the actual rate of devaluation (see the Data Appendix).  12 the US T-bill rate. We analyze the response of both interest differentials to measures of the projected net transfer and projected variance of the net transfer to foreign creditors. Our measure of the projected transfer to foreign creditors is a moving average discounted net transfer (interest plus amortization less disbursements) over the 48 months following the observation. An 8% annual discount factor (or a 0.64% monthly discount factor) is assumed (this was about the LIBOR rate at the time). We first define the discounted net transfer of month t+i measured at month t (DSt) as: (5) DSt- = (r,,*Ft. + At4 - Dt-)/(1.0064)i where: rt*Ft+ is the interest on the stock of debt at month t+i At4, DtZ are projected amortizations and disbursements at month t+i; 1/1.0064 is the constant monthly discount factor The average projected net transfer over the following 48 months measured at month t (NT,) is then: (6) NTt = E (DStj)/49 i = 0...48 The measure of the variance of the net foreign transfer at time t is: (7) SNTt = E ((DSti - NTt)2/(1.0064)'] i = 0 ... 48  13 The NT and SNT measures were Figure 3 computed for the period March 1988 to Variance of the Projected Transfer to Foreign Creditors. December 1990. The evolution of the ,oz, variance of the projected net 90% \ 80% i ^ transfer to foreign creditors is 70%t shown in Figure 3. s07 40% 30%. 1988.3 1989.1 1989.7 1990.1 1990.12 yeas 4.2 Regressions zzmizz The model predicts that a reduction in variability lowers the domestic interest rates and raises investment. As the literature has focussed on the effect of the level of net transfers (the tax in the debt overhang literature), we also investigate whether the interest rate and investment are affected by the average level of net transfers. We first consider the impact on physical capital accumulation and then on interest rates. We then separate out currency risk from more general country risk by analyzing the response of both local currency and dollar denominated Mexican debt instruments. 4.2.1 Debt Relief and Priv;te Investment Equation (8) links private investment2' to the measures of average net transfer and variance of future net transfers using Ordinary Least Squares (OLS). Consider first the debt overhang hypothesis, with NT, the measure of 9/ The monthly index of private fixed investment published in "Indicadores Economicos" of the Bank of Mexico.  14 average projected net external transfers, as a proxy for future tax burden (associated with the foreign debt service): (8) log(Ip) = 0.856 - 0.006*log(NT) + 0.808*1cg(Ip l) (1.32) (-0.32) (6.34) Al = 0.59 H-Statistic = -1.16 F = 21.6 Sample: 1988.04-1990.12 The value for the H-statistic (the more conventional Durbin Watson test cannot be used because of the presence of lagged endogenous variables on the right hand side of the equation) indicates the absence of serial correlation. While the net transfer variable has the right sign, it is entirely insignificant, with a dismal t-statistic of only 0.32. Inserting the variance measure significantly improves regression performance: (9) log(Ip) = 2.276 + 0.020*log(NT) - 0.146*log(SNT) + 0.476*log(Ip.1) (3.30) (0.56) (-3.03) (2.99) = 0.69 H-Statistic = -1.409 F = 32.62 Sample: 1988.04-1990.12 The coefficient of the lagged endogenous variable goes down, and the variance of net transfers enters with the right sign and is highly significant. The debt overhang proxy remains insignificant and now even gets the "wrong" sign. The H-statistic again indicates absence of serial correlation. Not surprisingly, regression performance does not change much when the NT variable is left out:  15 (10) log(Ip) = 2.294 - 0.140*log(SNT) + 0.51*log(Ip-1) (3.37) (-3.02) (3.00) R2 = 0.69 H-Statistic = -1.256 F = 32.6 Sample: 1988.04-1990.12 The results suggest that NT, the proxy for debt overhang, is not significant, and that the variance of the transfer variable is. An intriguing question is whether the variance variable captures the entire impact of the Brady plan or whether there are other factors at play that we have failed to account for. We test this in an admittedly crude way by rerunning equation (10), but now %.>th an additional "Brady Deal" dummy included1ol: (11) log(Ip) = 2.271 - 0.149*log(SNT) - O.OO9*BRADY + 0.477*Iog(Ip(-1)) (3.26) (2.64) (0.30) (2.97) = 0.69 H-Statistic = -1.491 F = 21.1 Sample: 1988.04-1990.12 The Brady dummy enters with the wrong sign and insignificantly. Thus equation (11) strengthens our results: it suggests that the variance measure captures the entire impact of the Brady deal. These regressions show that the variance index (SNT) was a statistically significant explanatory factor of private investment (Ip) whereas the average projected net transfer (NT) was not. The debt overhang explanation of the macroeconomic impact of debt relief thus finds no support in the data; the 10/ BRADY = 1 from 1989.07 onwards and zero before that date.  16 empirical analysis suggestd that the impact of debt relief on uncertainty is a more important channel through which debt relief influences the macroeconomy. 4.2.2 Interest Rates and Debt Relief I: Exchange Rate Risk. CETES and PAGAFE Consider next a similar analysis, but now for CETES. This asset allows us to focus the analysis directly on exchange rate expectations by first looking at the rate differential between CETES and PAGAFE. As already indicated, the difference in currency denomination is the only difference between the two debt instruments. Figure 2 showed the interest rate on CETES and on PAGAFE, with the latter corrected for the preannounced rate of devaluation. Regressing the interest rate differential corrected for preannounced devaluation, INTDIFFlU', on the same set of measures as were used in the investment equation gives similar results!V: (12) INTDIFF1 = -0.162 + 0.076*NT + 0.429*SNT (-1.01) (0.81) (2.56) R= 0.77 DW = 1.67 Sample: 1988.3 - 1990.12 INTDIFI = (1+1)- 11/ INTDIFF1 = (lvi( ) (1 y) , with i for CETES, iv for PAGAFE, and I the preannounced rate of devaluation. 12/ All interest rate equations were estimateQ using the Maximum Likelihood procedure in SAS assuming first order serial correlation, as the DW statistic in OLS regressions always indicated presence of serial correlation. The DW statistics reported in this section refer to the equation estimated under the assumption of first order serial correlation.  17 This regression also indicates that the variance of net transfers, rather than the projected net transfer itself, is the main factor behind the dramatic interest rate response to Mexico's debt package. The variance of the net transfer is statistically significant, and the net transfer itself is not. Not surprisingly, almost no explanatory power is lost when the net transfer is left out: (13) INTDIFF1 = -0.139 + 0.489*SNT (-0.89) (3.20) R2 = 0.76 DW = 1.67 Sample: 1988.3 - 1990.12 Adding the Brady dummy to equation (13) does not change anything significantly: (14) INTDIFF1 = 0.001 + 0.356*SNT - 0.084*BRADY (0.04) (1.89) (-1.12) R2 = 0.77 DW = 1.70 Sample: 1988.3 - 1990.12 Again the variance term captures all the effects of the Brady plan; the Brady dummy, a proxy for any effects of the deal not taken into account, is insignificant.  18 4.2.3 Interest Rates and Debt Relief II: Country Risk. PAGAFE and the US Treasury bill rate Consider finally the differential between PAGAFE and the one month US T- bill, INTDIFF2. Because both are denominated in US dollars, the difference between them is a pure measure of country risk. Regressing INTDIFF22I on the variance and expected value of the net transfer yields the following: (15) INTDIFF2 = -0.005 + 0.115*NT + 0.07*SNT (0.02) (1.99) (0.76) A2 = 0.80 DW = 1.24 Sample: 1988.3 - 1990.12 Now the variance of the transfer is completely insignificant, and the transfer itself is only borderline significant. The DW statistic is in the middle of the inconclusive range (1.11-1.36) for the relevant number of degrees of freedom; however regressing the error term on its own lag yields a t-statistic on the latter's coefficient of only 1.35, so we did not apply any further error corrections. Rerunning the equation omitting the insignificant variance term yields similar results, except that the significance of the net transfer term increases: 13/ INTDIFF2 + (1 iPAG) (1 iv  19 (16) INTDIFF2 = 0.044 + 0.12*i'T (0.75) (2.34) = 0.80 DW = 1.29 Sample: 1988.3 - 1990.12 The DW statistic is now borderline in support of zero residual serial correlation (the inconclusive range is (1.172-1.291). Regressing the error term on its own lag not surprisingly yields a low t-statistic (0.91). Thus no further correction seems warranted. These results are strikingly different from the ones obtained for investment and the CETES-PAGAFE interest rate differential. The variance plays no role, and the transfer itself only marginally, in explaining the interest rate differential between PAGAFE and the one month US Treasury Bill rate. This fits in well with our hypothesis that it is mostly exchange rate uncertainty that matters; exchange rate uncertainty plays no role in explaining this interest rate differential since both assets are dollar denominated, and the transfer effect captures the partial reduction in country risk. 4.2.4 Summina UD The empirical results thus confirm the conjecture put forward in the introduction, that the impact of debt relief on uncertainty is the most important channel through which debt relief influences the macroeconomy; and they sharpen our hypothesis by pointing specifically at the favorable impact on uncertainty about future exchange rate crises as the dominant factor in explaining the investment and growth response to debt relief.  20 The transfer effect is significant in explaining the PAGAFE rate, or country risk; however, one should be aware that the PAGAFE rate dropped only 4 percentage points after the Brady deal. Reduction in pure country risk thus explains only 4 percentage points of the overall drop of 20 percentage points in the differential between CETES and the T-bill rate observed immediately after the Brady deal was announced. 5. CONCLUSIONS Interest rates fell sharply after Mexico announced the successful conclusion of the negotiations with commercial banks aoout debt reduction, and private investment and growth recovered sharply. Econometric evidence presented in this paper indicates that the impact of debt relief on uncertainty is the most important channel through which debt relief triggers such macroeconomic effects. The debt relief reduced uncertainty principally by reducing the variability of repayment obligations which, at the time, could not be smoothed out by Mexico as it was credit-rationed. The evidence shows that while the impact of the variance of the projected net transfer on investment is statistically significant, the impact of the projected net transfer itself is not. Our analysis of the interest rate response furthermore sharpened the results by identifying the favorable impact on uncertainty about future exchance rate crises as the dominant mechanism in explaining the macroeconomic response to debt relief. These results confirm the potentially beneficial macroeconomic effects of debt relief, but reject the debt overhang hypothesis as an important explanatory factor for them.  21 There is an intriguing analogy with a much earlier episode of large external assistance. In their analysis of the Marshall plan for Europe's reconstruction in 1948, De Long and 1ichengreen (1991) show that traditional channels cannot possibly explain the spectacular success of that program. The amounts involved (incidentally as a share of GDP similar to the debt relief Mexico received) are simply too small to have as marked an impact on growth as actually took place. They conjecture that mechanisms rather similar to the one explored in this paper need to be invoked to explain the success of the Marshall plan. The most important conclusion that follows from our results is that the likely impact of debt service relief can be much larger than the magnitude of the relief coupled with standard growth models would suggest. The secondary effects on private investr.ent through reduc d future uncertainty are likely to be more important than the direct amount of the relief itself.  22 References: Brideau-Hall, R., H. Eggerstedt and S. van Wijnbergen (1992), "Measuring Capital Flight in Mexico," mimeo, World Bank. De Long, Bradford J., and B. Eichengreen (1991), "The Marshall Plan: History's Most Successful Structural Adjustment Program," CEPR Discussion Paper no. 634. Dornbusch, R. and F. Modigliani (1989) "Easing the Mexico Interest Burden," Wall Street Journal, January 3, A12. Giavazzi, F. and M. Pagano (1989), "Confidence Crises and Public Debt Management," NBER Working Paper no. 2926. Gurria, A. and S. Fadl (1992), "Estimacion de la Fuga de Capitales de Mexico, 1979-1990," Serie de Monografias 4, Washington, Interamerican Development Bank. Helpman, E. (1989), "The Simple Analytics of Debt-Equity Swaps and Debt Forgiveness," American Economic Review, 1989. Krugman, P. (1989), "Financing versus Forgiving a Debt Overhang," Journal of Development Economics, No.29, 253-68. Oks, D. (1991), "Devaluation Expectations and Interest Rates in Mexico: the Role of Domestic Debt Management," World Bank, LAC Discussion Paper no 98. Pindyck, R. (1991), "Irreversibility, Uncertainty and Investment," Journal of Economic Literature, 29 no. 3, (September):1110-48 Sachs, J. (1990), "The Debt Overhang of Developing Countries," in J. de Macedo and R. Findlay, eds. Debt. Stabilization. and DeveloDment: Essays in Honor of Carlos Diaz Aleiandro, Helsinki: Wider Institute, and Oxford, Blackwell.  23 van Wijnbergen, S. (1989), "Growth, Debt and the Real Exchange Rate in Mexico," in A. Wick and D. Brothers (eds.), Towards a New Development Strateav for Mexico, Westview Press. van Wijnbergen, S. (1991), "The Mexican Debt Deal," Economic Policy, April, pp. 13-56.  24 Data Appendix: Table 1 lists the data employed to calculate the interest differential between peso-denominated domestic public debt and dollar-denominated domestic public debt adjusted by the preannounced devaluation (INTDIFF1). The interest differential is measured as follows: (1) INTDIFF1 = (1 + i/100)/[(1 + iL,,/100)*(1 + y/100)] - 1 i equals the nominal effective annual interest on peso-denominated domestic public debt (Cetes 28 days); ip0 is the nominal effective annial interest on dollar-denominated domestic public debt (PAGAFE 28 days); and y is the annualized actual devaluation of the peso against the US dollar (there is a minor discrepancy with the preannounced devaluation rate).  25 Table 1. Interest Differential i 'PAO y INTDIFF1 l 1988:Jan 332.5757 55.06078 36.86948 1.038232 Feb 327.9344 56.8795 0 1.72779 Mar 117.212, Q 36.43456 0 0.59206 Apr 80.52329 26.82091 0 0.42345 May 57.4644 22.28896 _ 0 0.287642 Jun 48.50383 22.27675 0 0.21449 Jul 49.51486 21.48562 0 0.230721 Aua 52.15244 23.92272Q 0 0.227801 Sep 52.1375 34.43564 0 0.131675 Oct 57.52609 42.3Q5816_ 0 0.106548 Nov 67.99901 40.51497 0 0.195595 Dec 68.29388 35.65322 16.03831 0.069145 89: Jan 64.85105 35.21051 15.24147 0.057968 Feb 62.01072 29.79448 18.73302 0.051274 Mar 61.47335 22.75371 17.11416 0.123199 Akr 66.11012 23.91036Q _ 15.72798 0.158377 May 70.32168 27.60596 15.54275 0.155198 Jun 76.83275 30.33728 16.45815 0.164995 Jul 41.364 21.073692. 15.67603 0.00V359 Aua 39.18834 17.35334 16.00408 0.022431 Sep 42.79426 16.27664 14.72536 0.070431 Oct 46.41416 17.27109 _ 14.56386 0.089794 Nov 50.01538 15.81136 16.49668 0.111914 Dec 49.39731 16.34658 13.69422 0.129408 90: Jan 52.4814 16.3116 13.04359 0.159706 Feb 56.35786 17.05985 17.13828 0.140284 Mar _ 59.605. 18.025 12.73658 0.19952 Aor 52.931 17.73 14.33424 0.136141 Mav 39.243 17.871 11.33425 0.061054 Jun 39.394 , 16265 10.48151 0.085189 Jul 35,.86 15.266 10.76589 0.064167 Aua 34.40904 14.15056 10.66843 0.063963 Seo 35.4787 13.6589 10.57424 0.077987 l, Oct 30.45391 14 j4143L_ 5.290494 0.082898 Nov 28.0638, 14.90902 5.858032 0.052807 Dec 29.93647 14.59818 7.440796 0,t5532  26 In Table 2 we report the estimates of the projected net transfer (NT) and the variance of the projected net transfer (SNT). The net transfer projections (i.e., projections of interest, amortizations and disbursements) are "pipeline" projections prepared by the World Bank's External Debt Division. Projections go as far as December 1994 for our latest observation, i.e., December 1990. However, since World Bank projections are only updated in December of each year, the net transfer data employed in the calculation for the months running up to December of the first year was actual (rather than projected) data obtained from the Bank of Mexico. We also assumed that the impact of the debt restructuring on net transfers was perfectly anticipated in July 1989 when the agreement in principle was announced. All this results in the following procedures to construct the net transfer variables used in the estimation: i) for the net transfers projected in 1988 we used the actual monthly net transfers for the months that actually fell in 1988 and World Bank projections (made in 1988.12) for the months after 1988.12; ii) for the 1989.1-1989.6 (prior to the DDSR plan announcement) projections we used actual monthly net transfers for the 1989 months and World Bank projections (made in 1989.12) for the months after 1989.12; and iii) for the 1989.7-1990.12 projections we used actual net transfers until 1990.12 and World Bank projections (made in 1990.12) for the months after 1990.12.  27 Table 2. Net Transfer (NT) and Variance of the Net Tranffer (SNT) (march r 9Y5 = 1) NT I SNT 1 1988:Mar I_1_ 7 1 Aor 1,043615 0.9136578 May 1.043069 0.9354615 Jun 1.071462 0.9649376 Jul 1.059288 0.9852648 Aua 1.056994 1.0094836 Sep 1.153636 0.9215588 Oct 1.095474 0.8586899 Nov 1.093671 0.8813044 Dec 1.111386 0.9056212 89: Jan 1.128727 0.9060599 Feb 1.14Q40.49144451 Mar 1.188965 0.9542012 Apr 1.153775 0.9631204 May 1.161921 0.9768235 Jun 2 0.9882428 Jul 0.712316 1.5335344 Aua 0.697654 1.5674955 Seo 0.698062 1.5983988 Oct 0.74706 1.5640814 Nov 0.751103 1.5944446 Dec 0.759426 1.6264214 90: Jan 0.743656.. 1.6489762 Feb 0.729128 1.6751762 Mar 0.577246 0.6995677 Apr 0.662674 0.4631334 May 0.690731 0.4484275 Jun 0.715014 0.4572061 Jul 0.739403 0.44121 Aua 0.751093 0.4433348 Set) 0.775085_ 0.4275248 Oct 0.811134 0.3995143 Nov 0.810544 0.4087421 Dec 0.82318 0.4174881  Policy Research Working Paper Series Contact Title Author Date for paper WPS1 123 Social Development is Economic Nancy Birdsall April 1993 S. Rothschild Development 37460 WPS1 124 A New Database on Human Capital Vikram Nehru April 1993 M. Coleridge- Stock: Sources, Methodology, and Eric Swanson Taylor Results Ashutosh Dubey 33704 WPS1 125 Industrial Development and the Adriaan Ten Kate April 1993 C. Jones Environment in Mexico 37699 WPS1 126 The Costs and Benefits of Slovenian Milan Cvikl April 1993 S. Moussa Independence Evan Kraft 39019 Milan Vodopivec WPS1 127 How International Economic Links Vikram Nehru April 1993 M. Coleridge- Affect East Asia Taylor 33704 WPS1 128 The International Ocean Transport Hans Jurgen Peters April 1993 J. Lucas- Industry in Crisis: Assessing the Walker Reasons and Outlook 31078 WPS1 129 How Policy Changes Affected Cocoa Jonathan R. Coleman April 1993 G. llogon Sectors in Sub-Sanaran African Takamasa Akiyama 33732 Countries Panos N. Varangis WPS1 130 Poverty and Policy Michael Lipton April 1993 P. Cook Martin Ravallion 33902 WPS1 131 Prices and Protocols in Public Jeffrey S. Hammer April 1993 J. S. Yang Health Care 81418 WPS1 132 An Analysis of Repressed Inflation Andrew Feltenstein April 1993 E. Zamora in Three Transitional Economies Jiming Ha 33706 WPS1 133 Macroeconomic Framework for an Ibrahim Elbadawi April 1993 A. Maranon Oil-Based Economy: The Case of Nader Majd 31450 Bahrain WPS1 134 Managing a Nonrenewable Resource: Ibrahim A. Elbadawi April 1993 A. Maranon Savings and Exchange-Rate Policies Nader Majd 31450 in Bahrain WPS1135 Inflation in Czechoslovakia, 1985-91 Zdenek Drabek May 1993 E. Zamora Kamil Janacek 33706 Zdenek Tuma WPS1 136 The Dynamic Behavior of Quota Kala Krishna May 1993 D. Gustafson License Prices: Theory and Evidence Ling Hui Tan 33714 from the Hong Kong Apparel Quotas  Policy Research Working Paper Series Contact Title Author Date for paper WPS1 137 Railway Reform in the Central and Philip W. Blackshaw May 1993 TWVITD Eastern European Economies Louis S. Thompson 31005 WPS1 138 The Economic Impact of Military Daniel Landau May 1993 C. Jones Expenditures 37699 WPS1 139 Should Sub-Saharan Africa Expand Jonathan R. Coleman May 1993 G. llogon Cotton Exports? M. Elton Thigpen 33732 WPS1 140 How Retail Food Markets Responded Bruce Gardner May 1993 C. Spooner to Price Liberalization in Russia Karen M. Brooks 32116 after January 1992 WPS1 141 Foreign Direct Investment in a Maxwell J. Fry May 1993 R. Vo Macroeconomic Framework: Finance, 31047 Efficiency, Incentives, and Distortions WPS1 142 Rent-Seeking Trade Policy: A Time- Martin Rama May 1993 D. Ballantyne Series Approach 37947 WPS1 143 Tariff Rates, Tariff Revenue, and Lant Pritchett May 1993 M. Fernandez Tariff Reform: Some New Facts Geeta Sethi 33766 WPS1 144 The Foreign Trade Dimension of the Bartlomiej Kaminski June 1993 P. Kokila Market Transition in Poland: The 33716 Surprising Export Performance and Its Sustainability WPS1 145 The Simple(r) Algebra of Pension Dimitri Vittas June 1993 P. Infante Plans 37664 WPS1 146 Is Poverty Increasing in the Shaohua Chen June1993 P. Cook Developing World? Gaurav Datt 33902 Martin Ravallion WPS1 147 Interest Rates, Growth, and External Stijn Claessens June 1993 R. Vo Debt: The Macroeconomic Impact of Daniel Oks 33722 Maxico's Brady Deal Sweder van Wijnbergen