Did the Debt Crisis Cause the Investment Crisis?

Board of Governors of the Federal Reserve System International Finance Discussion Papers Number 418

December 1991

DID THE DEBT CRISIS CAUSE THE INVESTMENT CRISIS?

Andrew M. Warner

NOTE: International Finance Discussion Papers are preliminary materials circulated to stimulate discussion and critical comment. References in publications to International Finance Discussion Papers (other than an acknowledgment that the writer has had access to unpublished material) should be cleared with the author or authors.

ABSTRACT

There is now a large literature which attributes the investment decline in heavily indebted developing countries to the effects of the international debt crisis which began in 1982. However, these theories have not been tested against the altemative that declining terms of trade and high world real interest rates in the early 1980s directly caused the investment declines. This paper is based on the idea that if the debt theories are true, then forecasts of investment in the 1980s which do not use debt variables should not perform very well. This paper points out that such forecasts perform surprisingly well, and in many cases

go against the predictions of the debt theories, casting doubt on the validity of the debt theories.

Did the Debt Crisis Cause the Investment Crisis? Andrew M. Warmer’ I. INTRODUCTION

The investment decline in heavily indebted less developed countries has stimulated a large literature examining how extemal debt problems can cause domestic investment to decline. There is general agreement that the international debt crisis, which began in 1982, was caused partly by declining export prices for the indebted countries, high world interest rates, and sluggish growth in industrialized countries.

It seems plausible, although rarely stressed in the debt literature, that these same world economic shocks that caused the debt problems in the first place probably also caused a simultaneous reduction in investment demand in these countries. In the absence of further evidence, it is not clear whether the observed investment decline was caused by debt-related effects, or by the world economic shocks directly, or by some interaction between the two.

This paper argues that a clear way to approach this issue is to examine out-of-sample forecasts of investment over the debt crisis period (1982-1989) using equations which incorporate the effects of the world variables mentioned above but do not incorporate debt crisis effects. The essential idea is that these forecasts should not track investment during the debt crisis period if the postulated debt-crisis effects are important but should track investment if they are not. More specifically, if the debt-crisis effects are important, then these investment forecasts which ignore debt-crisis effects should be higher than actual investment.

There are numerous debt-related theories in the literature. The important point for this paper is that,

'The author is a staff economist in the Division of International Finance. This paper represents the views of the author and should not be interpreted as reflecting the views of the Board of Governors of the Federal Reserve System or other members of its staff. I am grateful to Olivier Blanchard, Jeremy Bulow, Susan Collins, Neil Ericsson, Maria Hanratty, Steve Kamin, Jeffrey Sachs, Larry Summers, Jeffrey Williamson and other seminar participants at Harvard and the Federal Reserve Board for helpful discussions and comments. Errors remain my own.

1

despite their diversity, many debt papers agree that 1982 was a watershed year that saw the introduction of investment dis-incentives that did not exist before. Therefore, the prediction that investment from our forecasting equations should exceed observed investment is shared by numerous theories, and these theories can be evaluated together by examining these forecasts.

In the debt-overhang models of Krugman [1988] and Sachs [1988], investment falls because creditors can skim off additional output resulting from capital accumulation, a situation which they argue has prevailed since 1982. Helpman [1988], presents a model where the continucd failure to resolve the debt crisis, a failure which prevailed at least up to 1989 when the Brady initiative was announced, raises expectations of future capital taxes, and thus depresses investment. Sachs [1988] also argucs that the debtoverhang caused international credit rationing. Rodrik [1989] and others arguc that investment has declined because of unprecedented policy unccrtainty after 1982, which can reduce investment because of risk aversion or because investment is irreversible as in Cukicrman [1980], Bernanke [1983], or Dixit [1989].

A less formal, but equally influential, line of reasoning points to the fiscal problems caused by the drying up of external credit in 1982, and argucs further that this caused governments to adjust in ways which depressed investment: by printing moncy and thus raising the specter of hyper-inflation;, by issuing internal debt and thus crowding-out domestic investment; or by forcing domestic banks to hold public debt, thus forcing them to scale back domestic investment loans. These effects are all triggered by the cutoff in international credit to governments in late 1982, and therefore also suggest that new investment dis-incentives prevailed after 1982.

Others attribute the investment decline to intemmational credit constraints directly.” Although it is true

*This argument is sometimes based exclusively on the balance of payments identity that capital inflows must equal the difference between domestic investment and domestic saving. This identity implies that if saving is held constant, and if capital inflows fall exogenously, then investment must decline. But these

assumptions are questionable. Investment and capital inflows could have been jointly determined by falling commodity prices and higher world interest rates.

that indebted countries have been unable to borrow new moncy from private commercial banks since August 1982, it does not follow from this alone that international credit rationing caused the investment decline. First, investment demand may have fallen simultancously from detcrioraling world economic conditions, so that credit rationing may not have been a binding constraint. For example, in oil-exporting indebted countries like Ecuador, Mexico, Nigeria and Venczucla, the 50 percent fall in oil prices between 1981 and 1986 may have caused investment demand to decline in the absence of a debt crisis or intemational credit rationing. In addition, other sources of investment financing besides international bank lending were available, including direct forcign investment. or perhaps more significantly, repatriation of flight capital. One problem with the simple credit rationing view is that it fails to explain why we did not observe repatriation of flight capital after 1982. If profitable investment projects did exist in these countries when discounted at the world rate of interest, then its unclear why wealthy Venezuclans failed to repatriate funds from Miami bank accounts to establish equity stakes in domestic firms.’

In an empirical study written contemporancously with this paper, Cohen [1991] pools data for 81 less developed countries and regresses investment as a percent of GDP on population growth, inflation, the ratio of exports to GDP, income per capita, the share of the population in primary school, time and regional dummies and the debt to export ratio in 1982. He finds that both the 82-87 dummy and the debt to export variable have negative coefficients (-1.67 and -0.13) but are not significant (standard errors are 1.90 and 0.69).* Cohen interprets this evidence as rcfuting the simple notion that the accumulated stock of debt represented an investment deterrent, and instcad argucs that if anything reduced investment it was

the forced debt service payments in the 1980s. He estimates this last effect essentially by regressing

3J. Bulow and K. Rogoff have expressed similar scepticism about debt effects in several papers dating back to 1988. Their 1990 paper provides a summary of additional criticisms.

“The investment and debt variable are measured in percentage points, and investment is in 1980 constant prices. Tests are not presented for the joint significance of the debt variable and the 82-87 dummy variable. Cohen also presents regressions with growth on the left hand side, but the same independent variables, and actually finds positive debt cocfficients.

investment on capital flows and concludes that the effect is statistically significant but small, with an increase of one percent of GNP transferred abroad estimated to reduce investment by 0.3 percent of GDP.°

While the Cohen study reaches similar conclusions to this paper, Borensztein [1990], is one study we have found which does not, although he analyzes only one country, the Philippines. He regresses investment on the relative price of investment goods, an estimate of the marginal product of capital, an estimate of the expected real interest rate, and several alternative debt variables. A representative result from his study is that a 1.3 billion dollar debt reduction, equivalent to 3.7 percent of GDP in 1987, is estimated to increase the ratio of investment to GDP by one percentage point. Five of the six debt coefficients he reports are significant, and all arc negative.

One potential methodological problem with thesc studies is that the debt crisis may work through other variables on the right of the regressions. We try to avoid this by using variables that are exogenous to the debt crisis and by using coefficient estimates which are estimated on data prior to the debt crisis. We also try to avoid simultaneity bias by using variables which are exogenous to the country. This paper also differs in that it focusses on the one clear prediction that all debt theories have in common, namely that the investment environment changed fundamentally after 1982.

After reviewing the basic facts in Section 2, Section 3 presents the estimated investment equations. Section 4 presents the forecasts, Section 5 discusses some criticisms and Section 6 concludes.

Il. THE FACTS

In the late 1980s, proponents of the debt view typically would point to the fact that investment rates were lower for the heavily indebted group of countrics than for other countrics to support the debt theories. However, the passage of time and some data revisions have made even this simple comparison

>The crucial assumption here is that capital flows are exogenous with respect to investment. Cohen

presents instrumental variables estimates, but the instrument list is long so that over-fitting in the first stage regression may be a problem.

less stark, and in addition, it was rarely pointed out that the debt group also experienced more dramatic declines in their export prices than did other countries over the same period. The essential facts are displayed in table I, which shows that investment declined by about 23 percent for the debt group and by 11 percent for all other developing countries. The table also shows that the terms of trade declined by 20 percent for the debt group and only 2 percent for all developing countries. Therefore, unless we adopt the extreme view that a decline in relative export prices had no effect on investment, it is still an open question whether the observed investment decline in heavily indebted countries was due to debt-crisis effects or to the terms of trade decline. Data on the terms of trade by country (Data appendix) show that all of the heavily indebted group with the possible exception of Brazil and Colombia experienced major reductions in their terms of trade in the 1980s.

It is also worth noting that investment declined in Texas and Louisiana in the 1980s as the price of one of their main exports, oil, declined, even though these two economies did not share many of the characteristics, such as sovereign risk or a separate monetary policy, of the heavily indebted group of countries®. Table II shows that even in non-petroleum manufacturing real investment fell by 36 percent in Texas and 26 percent in Louisiana between 1982 and 1986. For comparison, real private investment spending (including petroleum) declined 34 percent in Ecuador between 1982 and 1986, and 20 percent in Venezuela between 1982 and 1985. In Mexico, the National Accounts report that real investment declined by 40 percent between 1981 and 1986. A separate investment survey reports that real investment declined by more, about 50 percent, between 1981 and 1985. On balance, the investment declines in these

countries are not greatly different from those recorded in Texas and Louisiana.

‘I owe the initial suggestion to examine Texas to Jeremy Bulow and Larry Summers.

Ill. ESTIMATED INVESTMENT EQUATIONS

The forecasts are based on country specific estimates of the following reduced form equation.

I I

In = qd, + a In( ‘Spp ° GDP

det + (B, +LB,)P, + (6, +L6,)r, + (A, +LA,)gw, * E, (1)

Where I is national investment, P is the terms of trade, defined as the ratio of dollar export prices to dollar import prices, or a proxy; ris the 10-year U.S. t-bill rate minus contemporaneous U.S. PPI inflation; and gw is the percentage change in an industrial production index for developed countries, published by the International Monetary Fund.

The variables on the right of this equation are best viewed as instruments for country specific variables which are the more proximate determinants of investment in cach country. The reason for including the terms of trade variable is spelled out in the longer version of this paper and in a related paper on Mexico (Wamer, [1991}), where it is tested and supported by more extensive investment data. The argument in brief is that these countries basically imported capital equipment and installed it in primarily non-traded sectors. Partial evidence for this statement is in table II.’ Given this, if a terms of trade deterioration reduces non-traded prices relative to the price of imported machinery, as it will in a wide variety of open economy models, and as the data suggest (Edwards, [1991]), then almost any standard investment model will predict that investment will suffer.

The world real interest rate can be interpreted cither as the relevant cost of capital variable itself, or as an instrument for domestic interest rates via an uncovered interest parity condition. And finally, the

world growth variable can be interpreted as an instrument for shifts in domestic aggregate demand.

’ Table III shows that most of the imports of Latin American heavily indebted countries were either

intermediates or capital goods, and further evidence (not in table III) reveals that a large share of

machinery investment was imported (for example, 77 percent in Colombia and 78 percent in Venezuela in 1985).

6

The least squares estimates of equation (1) for cach country are presented in table IV. In most cases there are 21 years of available data (the terms of trade variable is usually not available before 1960). To improve efficiency slightly, we dropped the constant tcrm from the specification whenever it was Clearly insignificant, as it was in most cases (Argentina and Nigeria were the exceptions). This specification issue tumed out to be inconsequential as far as the forecasts were concerned.

The coefficient estimates in table IV show that in 11 of the 13 cases, the estimated short run terms of trade elasticity, given by B,+B,, has the anticipated positive sign. In 6 of the 11 cascs, it is significant at the 10 percent level. Also in 11 of the 13 cases, the estimated short run real interest rate effect, given by 5,+8,, has the anticipated negative sign, and is significant in 5 cases. In contrast to these results, the estimated world growth effect, A,+A,, has the anticipated positive sign in only 6 cases and is significant in only 2. Overall, the evidence in this table provides much stronger support for the terms of trade and world real interest rate variables than for the world growth variable. The estimated world growth effect often has the wrong sign, and when it has the right sign, it is frequently insignificant. There seems to be little supporting evidence that the world growth variable affects national investment after controlling for world interest rate and terms of trade effects. One possible explanation for this result is that shifts in world demand affect national investment only through their effects on export and import prices.

To provide additional information on the size of these estimated effects, pancl estimates which pool the data and estimate one equation are presented at the bottom of table IV.5 These estimates indicate that on average, the estimated short run terms of trade clasticity, (6,+B,), is 0.129 and the estimated long run elasticity, (B,+B,)(1-@), is 0.662. Similarly, the world real interest rate coefficicnt implies that a one

percentage point rise in the world real interest rate is estimated to change the ratio of investment to GDP

®To provide some intuition for the relationship between the individual country estimates and the panel estimates, it can be shown in a simple example with a one variable regression and two countries that the panel coefficient estimate is a weighted average of the country estimates, with weights proportional to the precision of the individual country estimates.

by -1.5 percent in the short run and by -7.8 percent in the long run.

Although the standard errors of the panel estimates of the terms of trade and real interest rate effects are fairly small, they are estimated under the possibly dubious assumption that the coefficients are equal across countries. In contrast, some, but not all, of the country specific equations have large standard errors and poor fits. In general, this imprecision is a Cost we are willing to bear in order to achieve exogenous forecasts, somewhat analogous to the sacrifice of efficiency for consistency in classical instrumental variables estimation. It also should be mentioned that we do not ignore this imprecision when forecasting because we will examine stochastic forecasts.

Because it would not make sense to base forecasts on estimated equations with incorrectly signed coefficients, the equations in table IV were re-estimated after dropping some of the variables. Table V presents the final forecasting equations. Since the world growth cffect frequently had the wrong sign, many of the forecasting equations do not usc this variable. but most use the terms of trade and world interest rate. In two cases, Argentina and the Philippines, onc of the exclusion restrictions was rejected but nevertheless imposed to arrive at a final forecasting cquation, bul in the remaining eleven cases the data did not object to the exclusion restrictions.

IV. INVESTMENT FORECASTS

To be clear about the nature of the forecasts that we will prescnt, consider a simple cquation of the form y, = ay,., + Bx, +e, Let "T" represent the last year for estimation, (T=1981 in this paper), and let "a" and "b" refer to the point estimates of « and 8. The graphs at the end of the paper contain a dark dashed line labelled "Predicted" which is the deterministic forecast, computed as §7,; = aye + bXq,). Sra2 = aVy,, + DX7,., and so forth. Note that after substituting for §,,,, the two-step ahcad forecast is non-linear

in the estimated parameters, 9;,. = a’y; + abx),, + bx,,,, which complicates but does not prevent

calculation of the forecast standard errors.”

These forecasts are deterministic because they substitute the point estimates for and B and because they implicitly set ¢=0. We also complement these forecasts with stochastic forecasts computed from Monte Carlo simulations which take into account the fit of the cquation and the uncertainty inherent in using estimated parameters. Using the equation above for illustration, we took repeatcd draws of the parameter vector (a B) from a N[(a b),V] distribution, where V is the estimated variance covariance matrix, and also took repeated draws of the crror tcrm c, from a N[0,s’] distribution, where s” is the estimated variance of e,. There were 500 draws for the parameter vector and 500*T” draws of the error term, where T’ stands for the number of years in the forecast period. For cach country, this produced 500 forecast paths, generating an empirical distribution of forecasts from which we obtained confidence bands."

The graphs at the end of the paper plot together three things: Actual investment over GDP (solid line); deterministic forecasts (darker dashed line); and confidence bands which correspond to upper quantiles of the simulated distributions of the forecasts. For example, the 90 percent confidence band in the figure corresponds to the 0.9 quantile of the top half of the simulated distribution. We take quantiles of the top half of the distribution to permit a visual one-sided hypothesis test. The confidence bands allow the reader to guage the strength of the evidence against the debt theorics on a time varying, and country by country basis. The finding that actual investment not only lics above predicted investment but even lies above these confidence bands is rather strong cvidence against the debt theories because they predict that if

anything actual investment should be below forecasted investment.

*Asymptotic standard errors can be calculated for any nonlincar function of estimated parameters by using the multivariate Central Limit Theorem together with the Delta Method. In our application, we found it was faster to program the Montc Carlo simulations presented below.

'©A third source of uncertainty is the use of a finite number of draws in the Monte Carlo simulations.

However, when we repeated the simulations, we found that the quantiles of the simulated distribution did not change very much. Marqucz and Ericsson [1990] discuss the statistical issues in more detail.

9

The graphs show that for 11 of the 13 countrics, actual investment docs lic above predicted investment for at least one year in the period after 1981. The persistent over-prediction of investment that the debt theories suggest is apparent in only two cases, Argentina and Nigeria. In many countries, predicted investment is substantially below actual investment. Even though some of the estimated cquations are imprecise, it does seem striking to observe that predicted investment is lower than actual investment for so many of the group of countrics which intemational organizations have singled out as problem debtors.

These results seem especially surprising in light of the presumption that some sort of debt-crisis effect must exist. If we performed these out-of sample forecasts on 13 countrics sclected at random, and found that the forecasts were too high for 2, too low for another 3, and about right for the remaining 8, this kind of finding would not support the conclusion that the 13 countrics were especially abnormal. If we reflect on the performance of these countries as a group, there docs not sccm to be strong evidence for debt effects after controlling for the cffects of the terms of trade declines and higher world interest rates.

The graphs also show that in the final year of the forecast period, actual investment lics above the 60 percent confidence band in 10 cases; and cven lics above the 90 percent confidence band in 3 cases: Brazil, Ecuador and Peru. This is especially strong evidence against debt effects in these three countries; on the other hand, the results for Mexico and Venezucla should be treated with caution, since the confidence intervals are quite wide."

To analyze further how the forecasts performed over time. we computed the average forecast error across all countries for cach ycar, where the forecast crror ts defined as actual minus forecasted investment. The debt theories would predict negative forecast errors, yet the average forecast error was positive for every year, and indecd rose stcadily over the 1980s, indicating perhaps that policy measures

"However, the evidence on Mexico in this paper should be superseded by the evidence in Warmer [1991], which analyzes far more detailed investment data for Mexico over the debt period and cstimates that about two-thirds of Mexico’s investment decline was attributable to the terms of trade decline and the

remaining third to the termination of capital flows to Mexico. Other more subtle effectS arising from the debt-overhang or uncertainty are cstimatcd to be negligible.

10

to stimulate investment were having an effect by the end of the 1980s.

The issue of whether we can statistically reject debt-crisis effects was further explored by estimating a panel investment equation over the full sample period, 1961 to 1989, with the same independent variables except for a 1982-1989 intercept dummy to allow for debt effects. The lag specification was similar to the earlier panel estimates except that we dropped the lagged dependent so that this variable would not pick up debt effects during the 1982 to 1989 pcriod, and thus cloud the interpretation of the coefficient on the debt dummy. The point estimate on this dummy, coded to equal 1 for the period 1982- 1989, was 0.102, with a standard crror of 0.050. The debt theorics would predict a negative valuc for this coefficient, yet it has the opposite sign and is in fact on the margin of statistical significance. This evidence is consistent with the carlier evidence that forecasted investment is not higher than actual investment on avcrage across all of these countrics.

We also checked robustness along several dimensions. First, we used real investment per capita from the World Bank rather than nominal investment over GDP from the Intemational Monctary Fund. We scaled by population to eliminate any possible debt effects working through the denominator of the earlier investment over GDP variable. We did not use this World Bank data at first because the sample extends only to 1965, whcreas the Intemational Monctary Fund data extends to 1960."

This second set of forecasts, not shown, also reveal predicted investment to be below actual investment for most of the countrics and most of the years aficr 1981. As in the previous forecasts, Argentina and Nigeria are in the debt camp. But unlike the previous forecasts, Mexico, and perhaps Peru, seem to be in the debt camp. With the exception of these two cases however, the results gencrally reinforce the earlier conclusions.

We also tried replacing our terms of trade data from the Intemational Monetary Fund with World Bank

When this research was completed, Summers and Heston data extended only to 1985. Forecasts with this data up to 1985 did not reveal major discrepancies.

terms of trade data, with the sample once again beginning in 1965 rather than 1960. The forecasts with these data were even more disapproving of the debt crisis hypothesis than the carlicr forecasts, since only Nigeria exhibited debt effects.

V. SOME CRITICISMS

One objection might be that terms of trade effects and debt-crisis effects are not separable. For example, the terms of trade declines may impede investment only in the presence of high debt or in the context of restricted international borrowing. But in fact the forecasting exercise sidesteps this criticism because the terms of trade coefficients are estimated on data before the debt-crisis emerged. If this criticism were true, we would be using lower terms of trade cocfficients than actually prevailed after 1981, and this error would reveal itself as an over-prediction of actual investment, which in fact we do not find.

Another objection might be that the terms of trade declines in the 1980s are not exogenous to the debt crisis. If an indebted country began to scll more of its exports on the world market to generate foreign exchange to service its debt, then world supply would shift out and cxport prices would be forced down somewhat. If this happencd, then the observed decline in the terms of trade would not be free of debtcrisis effects; nor would our forecasts based on this terms of trade decline.

While this effect is possible in theory, the evidence suggests that supply shifts from indebted countries are not the dominant cause of commodity price declines in the 1980s. A simple supply and demand analysis of world commodity markets would suggest that the size of this effect would be related to the elasticity of world demand, and the size of the shift in world supply, which in turn would depend in part on the market shares of the countries involved. Table VI reports the market shares of several key export commodities for the group of heavily indebted countries, first for the year just prior to the debt crisis, 1981, and then for 1987. The table shows first that the market shares are quite low for some commodities, so that even large shifts in supply from these countrics are unlikely to have much effect on the world

market price. The table also shows that for some commoditics, there is little direct evidence that such

12

supply shifts really have occurred in the 1980s, because the 1987 market shares are often below and usually not substantially above the 1981 market shares.

Perhaps a more relevant issue is whether the indebted countrics have market power as a group rather than individually. The bottom of table VI reports market shares for the four commoditics where this is a potential issue. To err on the side of the position we are arguing against, these numbers add the market shares of all developing countries, in effect assuming that they are all affected by the debt crisis, rather than just those in the heavily indebted group. The numbers show that the market shares of developing countries for the four key commodities did not change substantially between 1981 and 1987. We now tum to direct evidence on world quantitics and world prices of these commoditics to further check the supply shock hypothesis.

The movements in prices and quantities go in the wrong direction for the supply shock hypothesis for three of the four commodities. During 1981-1987, world production of Coffec, Petroleum, and Tin declined while the market shares of developing countries remained fairly stable. Since world prices also fell, the evidence indicates that demand shifts rather than supply shifts dominated these markets in the 1980s. On the other hand, world copper production and the share of output produced by the heavily indebted group both increased over this period while Copper prices fist fell and then recovered. Therefore, it possible that the heavily indebicd group drove down Copper priccs somewhat in the carly 1980s, but

for most commodities and countries, the direct evidence is not favorable to the supply shock hypothesis.

VI. CONCLUSIONS The forecasts in this paper were first conceived as a way to measure the size rather than to question the existence of debt crisis effects on investment. Yet the evidence casts doubt on the existence of debt effects because we fail to detect the systematic over-prediction of investment that the debt theories imply.

Instead, the investment declines in many of the countries on the heavily indebted list can be forecasted

13

out-of-sample by simple terms of trade and world reat interest rate equations which do not include debtcrisis effects. In 11 of the 13 countrics examined, forecasted investment in the final year of the forecasting period was lower than actual investment. In 10 of the 13 countrics, actual investment even lies above the 60 percent confidence band (one-sided) for forecasted investment. Finally, a debt crisis dummy added to panel regressions which pool the data on all of the heavily indebted countries not only fails to have a negative cocfficicnt as the debt theorics predict, but actually is positive and significant. The presumption that debt-crisis effects are needed to explain the investment declines in heavily indebted countries is strongly reduced by the fact that simple forecasts without debt-crisis effects can explain much of the declines. At the very Icast, the direct influence of world economic shocks in the

1980s on investment in heavily indebted countrics has received insufficient attention.

14

REFERENCES

Bernanke, Ben S., "Irreversibility, uncertainty, and cyclical investment,” Quarterly Journal of Economics, XCVIII (1983), 85-106.

Borensztein, Eduardo, "Debt Overhang, Debt Reduction and Investment: the case of the Philippines,” Working Paper 90/77, International Monetary Fund, September, 1990.

Bulow, Jeremy, and Rogoff, Kenneth, "Cleaning up Third World Debt Without Getting Taken to the Cleaners," Journal of Economic Perspectives, Winter, 1990.

Cohen, Daniel, "Slow Growth and Large LDC Debt in the Eighties: an Update," CEPREMAP unpublished manuscript, Paris, October, 1990.

Cukierman, Alex, “The Effects of uncertainty on Investment under risk neutrality with endogenous information, Journal of Political Economy, LXXXVIII (1980), 462-475.

Helpman, Elhanan, "Voluntary Debt Reduction: Incentives and Welfare,” NBER Working Paper, No. 2692, August 1988.

Krugman, Paul, "Market-Based Debt Reduction Schemes,” NBER Working Paper, No. 2587, May 1988.

Marquez, Jaime, and Ericsson, Neil, R., "Evaluating the Predictive Performance of Trade-Account Models," Federal Reserve Board, International Finance Discussion Paper, No. 377, 1990.

Rodrik, D., "Policy Uncertainty and Private Investment in Developing Countries," NBER Working Paper, Nv 2999, June 1989.

Sachs, Jeffrey, "The Debt Overhang of Developing Countries,” in Ronald Findlay, ed., Debt, Growth and Stabilization: Essays in Memory of Carlos Diaz Alejandro (Oxford, UK: Blackwell, 1988).

Warner, A., "The Debt Crisis, World Investment Cycles, and American Exports", PhD dissertation, Harvard University, May 1991.

. "Did the Debt Crisis or the Oil Price Decline Cause Mexico's Investment Collapse?" International Finance Discussion Papers, No. 416, November 1991b.

15

TABLE |

CHANGES IN INVESTMENT AND THE TERMS OF TRADE, 1980 - 1986

1980 Mean for Percent 1982 to 1986 change

Investment as a percent of GDP

Mean for Heavily Indebted Countries 21.90 16.96 -22.5

Mean for ali Less Developed Countries 26.50 23.60 -10.9 Terms of Trade Index

Mean for Heavily indebted Countries 100.00 79.90 20.1

Mean for all Less Developed Countries 100.00 98.08 “1.9

Sources. For investment: IMF 1990 Yearbook. For Terms of Trade: IMF 1988 Supplement on Trade statistics

16

TABLE Il

REAL INVESTMENT IN NON-OIL MANUFACTURING IN TEXAS AND LOUISIANA

Year Texas Louisiana 1982 5761.3 1399.9 1983 3814.0 894.0 1984 4194.4 1022.6 1985 4148.0 1133.0 1986 . 3677.0 1030.0

in millions of constant 1982 dollars, deflated by the U.S. producer price index for machinery. Source: U.S. Bureau of the Census, Census of Manufactures, 1982, and Annual Survey of Manufactures, 1983 to 1986.

17

TABLE lil

IMPORTS OF CAPITAL GOODS AND INTERMEDIATES AS A SHARE OF ALL IMPORTS FOR THE 9 LATIN AMERICAN HEAVILY INDEBTED COUNTRIES (percent of all Imports)

Intermediates Intermediate inputs’ Capital goods’ plus capital goods 1970 1984 1970 1984 1970 1984

BOLIVIA 33 31 34 38 67 61 BRAZIL 49 74 37 17 86 91 CHILE 37 50 42 35 79 85 COLOMBIA 40 50 44 33 84 83 ECUADOR 40 47 38 36 78 83 MEXICO 39 32 46 43 85 75 PERU 37 39 33 32 70 71 URUGUAY 55 71 17 15 72 86 VENEZUELA 35 36 39 37 74 73 Mean: 77 80

‘includes industrial supplies and fuets. "Includes machinery, transport equipment for industrial uses, and spare parts.

Source: Latin American Statistical Abstract.

18

TABLE IV

ESTIMATES OF EQUATION (1) eee

Durbin

O, B,+B, 6,45, Ata, R’ o H N

ARGENTINA 0.000 -0.418 -0.024 0.047 0.581 0.066 [0.510ja 17 (0.127) (0.017) (0.020)

BOLIVIA 0.558 0.298 -0.095 0.040 0.684 0.134 [0.331] 18 (0.160) (0.114) (0.034) (0.030)

BRAZIL 1.044 -0.069 -0.004 0.049 0.608 0.076 [0.174]a 16 (0.282) (0.202) (0.022) (0.029)

CHILE 0.518 0.386 0.080 -0.144 0.101 0.353 [0.144]a 21 (0.270) (0.177) (0.090) (0.111)

COLOMBIA 0.826 0.115 0.017 -0.002 0.177 0.083 [0.833] 21 (0.112) (0.068) (0.020) (0.019)

ECUADOR 0.921 0.068 -0.046 0.014 0.887 0.083 [0.065] 21 (0.118) (0.087) (0.020) (0.016)

MEXICO 0.877 0.082 -0.006 0.003 0.799 0.057 [0.113]a 21 (0.230) (0.151) (0.012) (0.012)

MOROCCO 0.745 0.233 -0.026 -0.068 0.773 0.183 [0.155] 21 (0.138) (0.102) (0.043) (0.039)

NIGERIA 0.513 0.142 -0.062 0.029 0.767 0.131 [0.740] 21 (0.212) (0.105) (0.032) (0.047)

PERU 0.823 0.133 0.021 -0.031 0.574 0.121 [0.419] 21 (0.126) (0.079) (0.027) (0.031)

PHILIPPINES 0.893 0.117 -0.024 -0.044 0.929 0.052 [0.427] 21 (0.039) (0.031) (0.011) (0.016)

URUGUAY 0.825 0.122 -0.032 -0.018 0.609 0.107 [0.000] 21 (0.217) (0.139) (0.034) (0.034)

VENEZUELA 1.022 0.006 -0.025 -0.001 0.759 0.132 [0.936] 21 (0.116) (0.080) (0.030) (0.038)

PANEL (b) 0.805 0.123 -0.020 0.010 0.682 0.163 [0.525] 261 (0.038) (0.026) (0.010) (0.009)

0.807 0.129 -0.015 - 0.681 0.163 [0.425] 261

(0.037) (0.026) (0.009) eee

These are least squares estimates. Standard errors are in parenthesis; significance levels are in brackets. Equation (1), estimated on annual data (1961 through 1981) is:

(1) In(VGDP), = a9 + oIn(VGDP),, + (B, + B,L)IN(P), + (3, + 6,L)R, + (4, + MLW, + e,

where /GDP is investment over GDP, P is the terms of trade, R is the world real interest rate, and gw is the world growth variable. The last two variables are measured in percentage points.

(a) These are significance levels for the lagrange multiplier test for first order serial correlation, based on Breusch (1978) and Godfrey (1978), instead of the Durbin H test, which could not be computed in these cases.

(b) The panel equations include fixed effects dummy variables for each country.

19

TABLE V

FORECASTING EQUATIONS

—_— OOO CC. ee eee

Durbin a By +B, 5,48, Mtr, R’ So H eS ARGENTINA 0.000 0.000 -0.013 0.000 0.047 0.086 [0.182] (0.017) BOLIVIA 0.558 0.298 -0.095 0.040 0.684 0.134 [0.331] (0.160) (0.114) (0.034) (0.030) BRAZIL 0.946 0.000 0.000 0.042 0.610 0.076 {0.812} (0.020) (0.015) CHILE 0.482 0.304 0.000 0.000 0.006 0.371 [0.089}a (0.217) (0.127) COLOMBIA 0.873 0.086 0.000 0.000 0.268 0.078 [0.464] (0.079) (0.053) ECUADOR 0.921 0.068 -0.046 0.014 0.887 0.083 [0.065] (0.118) (0.087) (0.020) (0.016) MEXICO 0.877 0.082 -0.006 0.003 0.799 0.057 [0.113]a (0.230) (0.151) (0.012) (0.012) MOROCCO 0.824 0.127 -0.038 0.000 0.745 0.194 [0.154] (0.140) (0.091) (0.042) NIGERIA 0.513 0.142 -0.062 0.029 0.767 0.131 [0.742] (0.212) (0.105) (0.032) (0.047) PERU 0.762 0.150 0.000 0.000 0.599 0.118 {0.063} (0.104) (0.066) PHILIPPINES 0.918 0.063 -0.028 0.000 0.878 0.068 [0.659] (0.050) (0.032) (0.015) URUGUAY 0.799 0.123 -0.035 0.000 0.585 0.111 [0.661] (0.160) (0.088) (0.025) VENEZUELA 1.026 0.003 -0.026 0.000 0.784 0.125 [0.837]

(0.057) (0.054) (0.024)

Standard Errors are in parentheses; significance levels are in brackets. To derive these forecasting equations, variables were dropped from the equation reported in table IV when the associated (sum of) coefficients had the wrong sign. For Argentina and the Philippines, this entailed imposing a restriction that was rejected by the data on the basis of the estimates in table IV.

(a) These are marginal significance levels for the lagrange multiplier test for first order serial correlation, based on Breusch (1978) and Godfrey (1978). The Durbin-H test was not computable in these cases.

20

TABLE VI

MARKET SHARES FOR KEY EXPORT GOODS OF HEAVILY INDEBTED COUNTRIES

Country Commodity Percent of World Production 1981 1987 BOLIVIA Tin 10.1 5.5 BRAZIL Coffee 24.6 34.4 lron ore 3.5 4.0 Soybeans 19.6 17.5 CHILE Copper 13.0 16.7 COLOMBIA Coffee 13.4 13.8 ECUADOR Petroleum - MEXICO Coffee 3.2 3.8 Petroleum 4.1 4.5 MOROCCO Fruits 52.7 26.7 NIGERIA Petroleum 2.5 2.3 PERU Copper 3.9 4.7 Zine 1.7 2.2 Lead 1.6 1.3 PHILIPPINES Coconut 41.7 35.4 Sugar 2.5 1.2 URUGUAY Wool 3.0 3.0 VENEZUELA Petroleum 3.8 3.0 ALL INDEBTED LDCs' Coffee 100.0 100.0 Copper 41.8 44.0 Petroleum 13.2 12.3 Tin 20.2 23.3

‘Includes ail non-OPEC less developed countries.

Source: International Financial Statistics Yearbook (1989), and World Commodity Yearbook

21

DATA APPENDIX

GDP Nominal investment over nominal GDP. Units: Percent. Source: IMF 1989 International Financial Statistics Yearbook updated with national accounts data from the Oct. 90 IFS. P This represents either a terms of trade index (export prices over import prices) taken directly from IMF sources or a constructed terms of trade index. ARGENTINA Terms of trade BOLIVIA Terms of trade BRAZIL Terms of trade CHILE Ratio of the price of Copper to U.S. machinery prices. COLOMBIA Terms of trade ECUADOR Ratio of an index of export prices [Petroleum (.69), Coffee (.13), and bananas (.18)] to U.S. machinery prices. MEXICO Terms of trade MOROCCO Terms of trade NIGERIA Ratio of crude petroleum prices (Saudi Arabia) to U.S. machinery prices. PERU Terms of trade PHILIPPINES Terms of trade URUGUAY Ratio of an index of export prices [Wool(.46), Beef(.37), Hides(.17)] to U.S. machinery prices. VENEZUELA Ratio of Crude Petroleum prices (Venezuela) to U.S. machinery prices. Units: Index number, 1980=100. Source: IFS 1988 Supplement on Trade Statistics for terms of trade data and commodity price indices. The terms of trade for Argentina are from the World Bank. r An ex-post, long term, real interest rate: the 10-year U.S. treasury rate minus contemporaneous U.S. producer price inflation. Units: Percent. Source: Economic Report of the President, 1989, and IFS Yearbook, various issues. gw Percent change in the IMF’s worldwide industrial production index. Units: Percent. Source: IMF 1989 international Financial Statistics Yearbook. Variable 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 P: ARGENTINA 100.00 97.93 90.24 96.21 96.97 89.77 85.30 81.76 86.25 - BOLIVIA 100.00 101.50 100.98 107.20 107.30 106.40 73.90 61.61 59.50 - BRAZIL 100.00 84.70 82.30 82.20 88.30 88.40 113.33 97.14 118.45 114.80 CHILE 100.00 72.66 58.32 61.02 51.53 52.05 49.56 80.80 90.40 95.52 COLOMBIA 100.00 84.40 81.90 83.90 86.80 86.00 104.70 52.87 81.32 70.80 ECUADOR 100.00 98.96 93.10 83.92 78.34 76.14 50.25 53.03 48.89 54.38 MEXICO 100.00 97.41 84.80 77.43 76.09 72.02 51.41 57.23 46.48 55.06 MOROCCO 100.00 94.00 94.30 87.40 89.40 89.40 93.90 80.90 93.94 91.10 NIGERIA 100.00 103.08 101.06 85.13 81.80 78.20 38.10 48.63 36.99 42.43 PERU 100.00 86.50 78.00 85.30 78.50 73.30 55.70 57.29 65.41 71.74

PHILIPPINES 100.00 88.00 85.60 89.30 87.20 81.40 87.60 95.40 106.80 - URUGUAY 100.00 84.81 74.59 73.15 74.29 66.65 69.02 88.21 105.34 105.83 VENEZUELA 100.00 105.72 99.86 85.04 80.08 76.83 33.20 46.98 34.81 41.50

gw

r

2.20 1.60 0.20 2.20 4.90 3.60 3.00 3.60 4.10

2.46 4.21 6.60 7.20 8.74 7.62 4.98 5.09 4.85

22

Aotual and Predicted Investment over CBP

Actual and Predicted Investnent over GBP ARGENTINA

23

Rotual and Predicted Investnent over CDP CHILE

Actual and Predicted Investment over CDP BRAZIL

3 i ft Hl

6 i a E 3 i

fictual and Predicted Investment over GDP

-~' 68 peroent conf idence band ++ 98 percent conf idence band

24

fictual and Predicted Investment over GBP MOROCCO

fictual and Predicted Investnent over CBP MEXICO .

puaq eousp! juno yusdsed 6g

25

26

IFDP NUMBER

418

417

416

415

414

413

412

411

410

409

408

407

406

405

404

International Finance Discussion Papers

TITLES 1991

Did the Debt Crisis Cause the Investment Crisis?

External Adjustment in Selected Developing Countries in the 1990s

Did the Debt Crisis or the Oil Price Decline Cause Mexico's Investment Collapse?

Cointegration, Exogeneity, and Policy Analysis: An Overview

* The Usefulness of P Measures for Japan and Germany

Comments on the Evaluation of Policy Models

Parameter Constancy, Mean Square Forecast Errors, and Measuring Forecast Performance: An Exposition, Extensions, and Illustration

Explaining the Volume of Intraindustry Trade: Are Increasing Returns Necessary?

How Pervasive is the Product Cycle? The Empirical Dynamics of American and Japanese Trade Flows

Anticipations of Foreign Exchange Volatility and Bid-Ask Spreads

A Re-assessment of the Relationship Between Real Exchange Rates and Real Interest Rates: 1974 - 1990

Argentina's Experience with Parallel Exchange Markets: 1981-1990

PC-GIVE and David Hendry’s Econometric Methodology

EMS Interest Rate Differentials and Fiscal Policy: A Model with an Empirical Application to Italy

The Statistical Discrepancy in the U.S. International Transactions Accounts: Sources and Suggested Remedies

—_—_—_—_—_——————

Please address requests for co

AUTHOR (s)

Andrew M. Warner

William L. Helkie David H. Howard

Andrew M. Warner

Neil R. Ericsson

Linda S. Kole Michael P. Leahy

Clive W.J. Granger Melinda Deutsch

Neil R. Ericsson

Donald Davis

Joseph E. Gagnon Andrew K. Rose

Shang-Jin Wei

Hali J. Edison B. Dianne Pauls

Steven B. Kamin

Neil R. Ericsson Julia Campos Hong-Anh Tran

R. Sean Craig

Lois E. Stekler

pies to International Finance Discussion

Papers, Division of International Finance, Stop 24, Board of Governors of the

Federal Reserve System, Washington, D.C.

20551.

27

IFDP NUMBER

403

402

401

400

399

398

397

396

395

394

393

391

390

389

388

387

International Finance Discussion Papers

TITLES 1991

In Search of the Liquidity Effect

Exchange Rate Rules in Support of Disinflation Programs in Developing Countries

The Adequacy of U.S. Direct Investment Data Determining Foreign Exchange Risk and Bank Capital Requirements

Precautionary Money Balances with Aggregate Uncertainty

Using External Sustainability to Forecast the Dollar

Terms of Trade, The Trade Balance, and Stability: The Role of Savings Behavior

The Econometrics of Elasticities or the Elasticity of Econometrics: An Empirical Analysis of the Behavior of U.S. Imports

Expected and Predicted Realignments: The FF/DM Exchange Rate during the EMS

Market Segmentation and 1992: Toward a Theory of Trade in Financial Services

1990

Post Econometric Policy Evaluation A Critique

Mercantilism as Strategic Trade Policy: The Anglo-Dutch Rivalry for the East India Trade

Free Trade at Risk? Perspective

An Historical

Why Has Trade Grown Faster Than Income?

Pricing to Market in International Trade: Evidence from Panel Data on Automobiles and Total Merchandise

Is the EMS the Perfect Fix? An Empirical Exploration of Exchange Rate Target Zones

Estimating Pass-through: Stability

Structure and

28

AUTHOR(s Eric M. Leeper David B. Gordon Steven B. Kamin Lois E. Stekler Guy V.G. Stevens

Michael P. Leahy Wilbur John Coleman II Ellen E. Meade

Charles P. Thomas

Michael Gavin

Jaime Marquez

Andrew K. Rose Lars E. O. Svensson

John D. Montgomery

Beth Ingram Eric M. Leeper

Douglas A. Irwin

Douglas A. Irwin

Andrew K. Rose Joseph E. Gagnon Michael M. Knetter

Robert P. Flood Andrew K. Rose Donald J. Mathieson

William R. Melick