|         Policy Rerch                 v4Ps   ) 2
WORKING PAPERS
Intemational Economic Analysis
and Prospect.
International Economics Department
The World Bank
February 1993
VPS 1102
Did the Debt Crisis
or Declining Oil Prices
Cause Mexico's
Investment Collapse?
Andrew M. Warner
Commodity price shocks (such as the decline in oil prices) have
been underestimated as a direct cause of declining investment in
the 1980s.
Pol;cyRceachWoikingPapedisunatedfindinpge cawoinpp aDdcnoouagetheexchangeofideasamogBankasuffand
othaes intiebred in tevelopBnkirssuLadoecpctobutedbytheRes adaAdvistor   Stafficanythennnesofotheauthors.f:crt
adlytheirviews,andshoWdbeuy and itedacoorndingly.Thefindings.intiptons,atdcwlsi rteathciors'own.Thcyshoald
not be anuted to the Wodd BaEn, its Board of Dixm, its mmagaentn, or cany of its member countries.



Policy Re  rc
~~~~~~~~~~~~~~- '
htondloc Prospok y
WPS 1102
This paper - a product of the Intemational Economic Analysis and Prospects Division, Intemational
Economics Department - is part of a larger effort in the department to examine the impact of extemal
shocks on low and middle-income countries. Copies of the paper are available free from the World Bank,
1818 H Street NW, Washington, DC 20433. Please contact Mila Divino, room S8-013, extension 33739
(February 1993, 31 pages).
Wamer proposes and estimates a microeconomic       The decline in Mexico's international terms
investment model to determine the relative     of trade was probably the most important ulti-
importance of three explanations for Mexico's  mate cause of the increased relative cost of
investment decline in the early 1980s:         machinery, but the reversal in net capital inflows
to Mexico probably also played a role in increas-
• The decline in oil prices.                ing this relative price. On this point, the evidence
is not as clear.
* The termination of capital inflows.
After controlling for these effects, Wamer
* The effects of debt overha,-g and uncer-   finds litle evidence that the effects of debt
tainty.                                        overhang and uncertainty had much to do with
the investment decline.
He uses investment data for private indus-
tries between 1981 and 1985, which have yet to     Warner points out that investment in Texas
be used in addressing the question under discus-    and Louisiana (which were also riding the oil
sion.                                          boom of 1973-81) also fell in 1981-86, and
adverse commodity price shocks also affected
Tne data indicate that the main microeco-   many other heavily indebted countries. At the
nomic mechanism driving lie decline in invest-  very least, commodity price shocks (such as
ment was a rise in the relative price of invest-  Mexico's decline in oil prices) as a direct cause
ment goods - especially the relative price of  of declining investment levels in the 1980s have
machinery (a traded good in Mexico). Moreover,   been insufficiently emphasized in the literature
the decline in trade (driven by falling world oil  on the effects of the international debt crisis.
prices) explains much of the increase in this
relative price.
ThePolicy ResearchWorking PapSeriesdisseminates thefindings ofwortkuwder wayintheBanlk Anobjectiveof the series
is to get these findings out quickly, even if presentations are less than fully polished. The fmdings, interpretations, and
conclusions in these papers do not necessarily represent official Bank policy.
Produced by the Policy ReseArch Dissenination Center



Did the Dsbt Crisis or the Oil Price Dledine Cause Mexico's
Investment Collapse?
Andrew M. Warner'
Room S-8018
The World Bank
1818 H Street, NW
Washington DC, 20433
'This paper is a revision of chapter 2 my dissertation and of International Finance Discussion Paper
#416, Federal Reserve Board. It has benefitted from comments and discussions with Susan Collins,
Ashok Dhareshwar, Steve Kamin, Yves Maroni, Jeffrey Sachs. Carlos Sales, Oscar Sanchez, Larry
Summers, Jeffrey Williamson, Ning Zhu and two especially helpful anonymous referees.



Table of Contents
1.    Introduction                                                         1
2.    Important Dates and Movements in Key Variables                       2
3.    The Micro Model of Investment                                        5
4.    The Mechanisms Behind the Terms of Trade and Debt Effects            9
5.    Data and Econometric Issues                                         13
6.    Results                                                             17
7.    Evidence with Annual Investment Data                                22
8.    Condusions and Discussion                                           23
Tables                                                                    26
References                                                                30



1 Introduction,
It is not necessarily surprising to learn that investment declined in a country when the world price
of its key export declined by about 50 percent. Yet the investment decline in Mexico in the early 1980s
is almost universally attributed to some aspect of the international debt crisis or to general uncertainty
rather than to the decline in the price of oil. The possibility that the oil price decline would have reduced
investment even in the absence of the debt crisis is rarely seriously considered. Instead, the usual
argument is that the oil price decline helped cause the debt crisis, and then some other phenomena
associated with the debt crisis, such as the debt-overhang, or heightened uncertainty about policy reforms,
in turn caused the investment decline.
Sorting out the role of the oil price decline versus the debt crisis is important because the view that
Latin America's economic problems in the 1980s were caused by debt problems is an important premise
in many policy discussions. Empirical studies attempting to estimate the effect of the debt crisis (broadly
defined) on investment have only recently begun to appear: Cohen (1990), Faini and de Melo (1990), Fry
(1989), Hofman and Reisen (1990), Sawides (1992) and Warner (1992), and so far there is no consensus.
These studies analyze entire groups of less developed countries and focus on investment aggregates from
national accounts data. This paper takes a more targeted approach: focussing on a country and a data
set where it is possible to define and estimate the effects more precisely.
The main focus of this paper is to distinguish empirically between three possible causes
for Mexico's investment decline in the early 1980s: the oil price decline, the termination of capital
inflows, and debt-overhang/uncertainty effects. The second goal of the paper is to help fill a gap in the
literature by providing structural estimates of a private sector investment demand function in a small open
economy. It is fortunate that good data are available for Mexico during a period when the available
instruments (world oil prices, world interest rates) exhibited substantial variation. Thus there may also
be sufficient statistical information to try to identify structural parameters.
1



The data are from & sector -level investment survey conducted in MexicG between the first quarter
of 1981 and the last quarter of 1985, which has not yet received much attention. This data set has several
virtues: it spans the crucial debt-crisis year of 1982, is of fairly high quality, permits explicit controls for
industry-specific investment determinants such as relative product prices and wages, and permits a focus
on industries which are in the private sector.
I begin with an assumption that concedes ground to the debt crisis side of this debate. That is,
I will assume that the sudden termination of international capital flows to Mexico, which happened in
1982, was exogenous instead of being caused by the continued decline in the price of oil, which began
in the middle of 1981. This assumption will lead to an underestimate of the magnitude of the oil price
effect, because it rules out any effect of the oil price decline working through the capital flow variable.
The paper will show that despite this assumption, the data suggests that the effect of the oil price decline
was large.
This paper is organized as follows. Section 2 sur eys the main facts for Mexico. Section 3
presents a micro-economic investment model. Section 4 spells out how the terms of trade decline and
the debt crisis relate to this investment model. Section 5 discusses the data and econometric issues.
Sections 6 and 7 present results using quarterly and annual data, respectively. Section 8 concludes.
2. Important Dates and Movements in Key variables
Mexico's debt problems in 1982 are usually attributed to the combined effect of high world real
interest rates, falling oil prices, and an inability of Mexican policy makers and their international creditors
to adjust to these new realities. The main facts are as follows. Between 1980 and 1982, world real
interest rates were very high by any measure. In July of 1981, world oil prices began to decline. In
February of 1982, Mexico devalued the peso by 46 percent and devalued again in August of 1982.
Throughout early 1982, macro economic reforms were repeatedly announced but only partially
implemented. Through July of 1982, international creditors were still lending heavily to Mexico. On
2



August 12, 1982, Mixic6 announced to the sii'prise of the ifiternational comxztinity that it could not meet
its short term obligations falling due in the coming week. By 1983, new capital inflows had virtually
stopped, Mexico was transferring resources abroad, and investment and growth were sharply lower than
the levels achieved in 1980 and 1981.
The investment decline after 1981 was severe by any measure. Real investment data from the
national accounts show that average annual real investment in the period 1983-1985 was 63 percent of
the 1981 level, and the decline was roughly similar for public and private investment. The same
calculation using the investment survey data puts this number at 53 percent.
Apart from debt crisis effects, the two key exogenous variables this paper considers to explain
this decline are the terms of trade and net capital inflows.' The capital inflow variable is the sum of net
capital inflows from the capital account (not including changes in official reserves) net interest payments,
and errors and omissions, all measured in billions of dollars and all deflated by the U.S producer price
index (1982= 1.0).2 In 1981, net capital inflows averaged $1.8 billion per quarter; in 1985, there was
a net outflow averaging $3.0 billion per quarter.
The terms of trade, defined as the ratio of export prices to import prices, also declined
substantially during this period. Driven by the decline in the world price of oil, the terms of trade fell
by almost 30 percent between 1981 and 1985 (using annual averages), and then fell a further 23 percent
in 1986. The decline in the terms of trade started in the middle of 1981, clearly preceding the debt
I The assumption that capital flows were exogenous is made more for the sake of the argument than
for realism. The capital flow reversal was probably caused in part by the continuing fall in the price of
oil as well as the sudden questioning of Mexico's solvency by its creditors. In contrast, the assumption
that Mexico's terms of trade decline was exogenous is easier to defead because world oil prices were the
key variable driving this index and Mexico only produced about 4 percent of world oil output.
2 By including the errors and omissions in this measure of capital flows, I follow numerous studies
in assuming that this primarily measures unrecorded private capital flight rather than other errors
(Anthony and Hughes-Hallett (1992) is a useful survey of possible capital flight measures). Casual
inspection of the data during 1981-1982, when capital flight was known to be extensive, support this
assumption.
3



problems of August, 1982.
4



3. The micro model of investment.
This section presents a microeconomic investment model to organize the analysis, and the
following section relates the exogenous variables to this model. I choose a baseline model which
highlights the impact of relative price movements on investment partly because the data are available but
also because adverse relative price movements provide the main alternative explanation for the investment
decline. Investment data is available at the sector level, and hence the model .s of a representative firm
at this level. There is a neoclassical constant returns to scale production function, F(K,L), and a convex
internal adjustment cost function, C(I), which measures adjustment costs in terms of the investment good.
The firm is assumed to be a price-taker in product, factor and financial markets, and to know the values
of future exogenous variables.'
The firm's problem is to choose investment and employment to maximize the present discounted
value of future cash flows. Whenever possible, time and sector subscripts are suppressed.
(1)                     Max f e   [pF(K,L) - wL - ptC(p ] ds
(52)                              ist.   k = i - 8K
where r is the real interest rate, p is the product price, w is the wage rate, and p' is the price of capital
goods. The solution to this problem, obtained by maximizing the current valued Hamiltonian, includes
the following four equations.
(3)                                  p FL(K,L) = w
I These assumptions are made to solve and motivate the investment model. The empirical section
allows for a less-restrictive set of assumptions and other determinants of investment not in this model.
5



(4)                                - pI C/I) = q-           X
(5)                            q - (r+6)q - -p FK(IL)
(6)                                lm         q   O
Equation (3) equates the marginal product of labor with the real product wage. Equation (4)
states that investment is chosen up to the point where the marginal cost equals q, which represents the
increment to the value of the firm from a unit change in investment. After integrating equation (5), q
can be expressed as the present discounted value of future marginal value products of capital, discounted
at the rate r+6. Note that higher wages depress investment demand by reducing L, from (3), which in
turn reduces the marginal product of capital and therefore q, in equations (5) and (4).
To obtain a closed form investment equation one needs simplifying assumptions on the functional
forms of C(.) and F(.). With these in hand, the solution procedure is to substitute optimal employment,
L, from equation (3) into equation (5) and then integrate eouation (5) forward from t to co using (6), the
no bubbles condition.
To provide an explicit investment function, I first assume a Cobb-Douglas production function,
F(K,L) = KI L'-. This assumption yields a simple and tractable expression for q:
(7)                                 q = A     6W
r+8
where A = e(o-1)('+ and 0 = 1/ax > 1.
6



Substituting (7) into (4), and assuming a quadratic investment cost function4, C(I) = 12/2, yields
an investment function which is increasing in p, and decreasing in w, (because 0> 1, and therefore 1-
0<O), r+b, and p.
(8)                                   1 = A  P M1
p (r+8)
There is data on investment, product prices, and wages which varies across sectors and time; data on the
price of investment goods and interest rates which varies across time only, and no data on depreciation.
Therefore, I will estimate a version of equation (8) whiich is modified to accommodate the available data.
That is, depreciation will be suppressed, and the constant, A, will be allowed to differ across sectors to
pick up differences in technology, sector size, or anv other sector specific but time invariant variable.
Letting "j" denote sectors and "t" quarters, I will estimate an equation similar to the following:
)2 1.
(9)                                  If    iPjt Wit
(Psl) '4r"
The key issue to discuss now is how exogenous variables such as the terms of trade and capital
inflows, and a mi-re general phenomena such as the debt crisis relate to equation (9). 1 will distinguish
three possibilities which are not mutually exclusive. The first possibility is that the terms of trade decline
caused the investment decline by affecting the prices on the right hand side of (9). The second possibility
4 From equation (4), a cost function with a positive second derivative will deliver the result that
investment increases in q and decreases in p': the quadratic functional form is not essential for this.
Including K or L in the investment cost function would make it difficult to obtain a closed form solution
for investment.
7



is that the debt crisis caused the investment decline by affecting the prices on the right hand side of (9).
The third possibility is that the debt crisis caused the investment decline by working through a variable
that is not present on the right hand side of (9).
8



4. The mechanisms behind the terms of trade and debt effects.
This section will specify the channels through which I allow the debt crisis and the terms of trade
decline to affect the investment equation (9) above. Briefly, the terms of trade and capital inflow effects
are indirect. They work through the prices on the right of (9), using a demand and supply framework
familiar from Dutch Disease models.5 Other debt crisis effects such as the debt overhang affect
investment more directly, and are tested by adding debt-crisis proxies to the right of (9). The rest of this
section explains the reasoning in more detail.
There is first of all a direct and obvious relationship between the terms of trade and prices in
export and import sectors. Because the terms of trade is measured as p'/pm, shocks to world export
prices will induce a positive correlation between domestic export prices and the terms of trade, and
shocks to import prices will induce a negative correlation.6
Apart from this direct relationship between the terms of trade variable and prices in traded
sectors, a terms of trade improvement represents an increase in wealth which may stimulate demand and
therefore affect product prices in non-traded sectors. This kind of effect is familiar from Dutch Disease
models (for example, Neary and Van Wijnbergen, 1986) where a rise in the terms of trade raises
spending or, home goods and increases the relative price of non-traded goods. Capital flows are assumed
to shift demand in a similar fashion. These assumptions lead to the following demand and supply
equations for any given product market, indexed by the "j" subscript.
I The model here is like an open economy model where the domestic interest rate is given by the
world interest rate and therefore shifts to the savings function from exogenous changes in capital inflows
do not have a direct effect on investment, but may have an indirect effect through prices. Although the
basic model here does not allow capital inflows to directly affect investment, I test for this possibility in
the empirical section.
6 The terms of trade index was computed by the Bank of Mexico using international (dollar) prices
of Mexico's main imports and exports. These prices are distinct from the sectoral value added prices,
pj, also used in this paper.
9



(10)                             pj = d( yj, P  , CF)
p 
(11)                                p1 = S( y , w1)
In equations (10) and (11), p is the product price, y is output, p'/p  is the terms of trade, CF is
the capital flow variable and w is the wage. The sectoral wage in turn is determined in the simplest
possible neoclassical labor market, described by equations (12) and (13) below. Labor demand is an
implicit function of the product price and the wage.
(12)                                wj = d( pj, LJ)
And labor supply is an implicit function of the consumer price index and the wage.
(13)                               WJ = s( CPI , LJ )
Equations (10) to (13) can be solved to obtain reduced form equations relating prices and wages
to the terms of trade, the capital flow variable, and the CPI.
(14)                            PJ = g( Px, CF, CPI)
ptm
(15)                            wJ = h( P   , CF, CPI )
10



The important point for this paper is just to establish that such reduced forms exist, in order to
motivate regressions of prices and wages on the terms of trade and capital flows. A possibly confusing
point is that the estimated regressions have pj/CPI and wj/CPI on the left rather than pj and wj as in
equations (14) and (15), implicitly assuming that the functions are homogenous with respect to the CPI.
Furthermore, because it happens not to matter for the main points, the reported regressions use weighted
average product prices and wages, Ewjp/CPI and rjwj/CPI, on tho left, where the co are investment
shares. If reduced forms exist for sectoral prices, they also exist for functions of sectoral prices such as
these price indexes. I do arbitrarily assume a log-linear functional form.
The expected sign of a(pj/CPI)/a(p/pm) depends on the nature of the sector. If pj is the price of
a non-traded product, a decline in p', holding pm constant, should depress demand and non-traded prices.
Given that the market basket of the CPI has a high non-traded component, both numerator and
denominator of pj/CPI should be affected similarly, and thus a regression of p/CPI on p'/pm should yield
a coefficient that is close to zero.
If pj is the price of imported machinery, and the price of oil exports falls, the denominator should
be depressed but not the numerator, yielding a negative coefficient. If pj is the product price of a non-oil
export sector, and the price of oil exports falls, the same thing should happen. Of course, if pj is oil,
both numerator and denominator should be depressed, yielding an ambiguous but probably small
coefficient. Since the capital flow variable is also modelled as a demand shifter, the same kind of
reasoning also applies for its effects on p/CPI.
This section now turns to a brief discussion of the debt theories. It is important to remember that
the main triggering mechanism for these theories is the perception that the country is not solvent rather
than simply the level of debt. The theories argue that this shift in sentiment lead to several other factors
which directly depressed investment. Therefore, this paper considers proxies for Mexico's perceived
11



creditworthiness by the international financial community to test these theories.7
Krugman (1988) and Sachs (1988) advanced the idea that the debt overhang can account for the
investment decline. Following the terminology in Krugman (1988), a debt-overhang is said to exist when
the expected present value of future resource transfers is less than the face value of the extemal debt.
The argument presumes that a debt-overhang creates a situation where creditors can siphon off some of
the additional output resulting from investment. In this situation a social planner would invest less due
to the implicit tax levied by foreign creditors. This is the formal mechanism linking high debt to low
investment in the models of Sachs (1988), Krugman (1988), Froot (1988), and Bulow and Rogoff (1989),
although the last two papers do not necessarily advocate the debt-overhang view.
A related argument is that the overhang caused the termination in lending by foreign commercial
banks. In Sachs (1988), the passage of a country from solvency to a debt-overhang places indebted
countries in a situation analogous to a domestic insolvent firm. Lending is restricted essentially because
semi-insolvency exacerbates creditor-debtor agency problems.
Krugman (1988) and Helpman (1988) also argue that a debt overhang causes investors to fear
higher investment taxes.  Ize (1989) and Rodrik (1989) stress that the debt crisis introduced
unprecedented policy uncertainty. Dixit (1989), and Rodrik (1989), following ideas in Bernanke (1983)
and Cukierman (1980) argue that even in the absence of risk aversion, it may be optimal to postpone
investments in the face of greater uncertainty, because the decision may be irreversible, and uncertainty
may resolve itself with the passage of time.
The evidence suggests that there was an important shift in perceptions of Mexico's solvency in
the late Summer of 1982. Solis and Zedillo (1985) report that Mexico's solvency was first questioned
during negotiations for a 2.5 billion dollar "jumbo" loan to the Mexican government in June 1982. The
I Fry (1989) and Cohen (1990) test specifications with the debt stock on the right; Savvides (1990)
also argues against such practice. In this paper, the debt stock is used in annual regressions, but not in
the quarterly regressions since the data are not readily available.
12



issue of the Economist magazine published just after the August incident reports that creditors were still
lending heavily to Mexico in the spring of 1982. Earlier issues of the Economist in 1982 fail to mention
Mexico's impending payments problems. Furthermore, the price of a Mexican government bond, and
the price of the "Mexico Fund" stock, both of which were traded on the New York Stock Exchange at
this time, did not exhibit any unusual decline until the week of the August announcement, suggesting that
Mexico's problems were not anticipated by the international financial community. Contemporary
accounts such as Kraft (1985) attest to the surprise caused by the August announcement.
5.Data and Econometric issues.
The investment data analyzed here is a sub-sample from a larger data set developed from a firm-
level investment survey conducted by the Bank of Mexico between 1981 and 1985. This larger data set
was published in the Bank of Mexico's monthly publication Indicadores Economicos in more aggregated
form (9 sectors rather than 68) during 1987 and 1988. The more disaggregated data used here was
obtained from the intemal files of the Bank of Mexico during a personal visit by the author. This data
set is not well known outside of Mexico and has not yet been examined at this level of disaggregation in
print. The key features of the data set are as follows.
*     Real investment spending: constant 1970 pesos.
*     4 kinds of investment goods: tr.msport equipment, machinery and operating
equipment, buildings and structures, and office furniture and equipment.
*     68 sectors, mostly mining and manufacturing.
*     20 quarters: 1981:1 to 1985:4.
*     Coverage includes 56,053 public and private firms; does not include
"maquiladora" or "in-bond" firms operating in tariff-free export enclaves; does not
include Agriculture, Forestry or Fishing.
The sub-sample I analyze excludes sectors in which public enterprises are dominant, and restricts
attention to machinery investment. This selection permits a focus on private sector behavior, and permits
13



a focus on a relatively homogenous investment good.5 All sectors with more than 25 percent of output
produced by public enterprises in 1980 were eliminated, reducing the number of sectors in the sample
from 68 to 48. Another 6 sectors were eliminated for lack of price data, leaving 42 sectors in the
sample. This final sample represents 61 percent of all private investment in the sample and 21 percent
of public and private investment.
Turning back briefly to the full data set, investment declined in virtually all (65 of 68) of the
sectors between 1981 and 1985, and typically was quite severe. Simple correlations do not reveal any
statistically significant association between the investment decline and ownership status (public/private);
nor any significant association between the investment decline and import or export shares. The only
significant s.mple association I found is that sectors with high machinery shares in investment in 1981
also tended to experience more severe investment declines after 1981.
The other key variables are listed and described briefly in table 1. The price indexes measure
the price for value added, and the wage indexes measure the mean cost of a man-hour of labor in the
sector.  The interest rate is the three-month government bond rate (CETES in Mexico) minus
contemporaneous CPI inflation. Many other measures of real interest rates were tried in the regressions
but not reported in table I to save space. MEXB is the ratio of the price of a Mexican government bond
to a risk-free bond with the same coupon and maturity.9 MEXF is the stock price of the "Mexico Fund"
mutual fund, traded on the New York Stock Exchange. MEXCW is Institutional Investor's index of
Mexico's perceived creditworthiness by major U.S. banks (a higher value means the country is more
' Machinery investment is by far the largest spending category in this data set (about 75 percent of
the total). Earlier versions of this paper included analysis of separate investment equations for machinery
investment and all other investment. The output was much more complex but the main conclusions were
similar.
9 The risk-free bond price was calculated by the author as the present discounted value (using the
average interest rate on 10-year U.S. government bonds for a given quarter) of the Mexican bond's
payment stream (the coupon was 8.125 percent and the bond matured in December 1997). The Mexican
bond price was obtained from Standard and Poors.
14



creditworthy)."0
The simple movements in the data are as follows. Real product prices do not change substantially
during the 1981 to 1985 period; however real wages fall by about 30 percent on average; and the real
price of machinery rises by more than 30 percent. Real domestic interest rates decline during the high
inflation quarters around 1982 but rise back towards zero by the end of the sample." The three debt
crisis proxies, MEXB, MEXF. and MEXCW, all are fairly flat before the Summer of 1982, and then
plunge precipitously around August 1982. After 1982, MEXB rises steadily up to the end of the sample
in 1985, while the other two variables remain depressed throughout the period.
The equations to be estimated are log linear versions of the investment equation, (9), and the
reduced form price and wage equations, (14) and (15). Recall that product prices, p, machinery prices,
p', and wages, w, are all scaled by the CPI.
(16)         lnQ)jt = toj + ajDt + CL21n(p),t + a3In(w)jt + ce4ln(p, + ar, + eij
(17)                     ln(pbar), = +0 + X1ln(')l +          22CFt + E2
p   
1' Institutional Investor asks about 90 banks to grade countries on a 0 to 100 scale, with 100
representing the lowest chance of default, and then weigbts the responses by size of exposure and
analytical sophistication of the bank's country analysis system. These ratings are published twice every
year, and linear interpolation was used to obtain quarterly numbers.
" Two other measures of expected inflation: one using fitted values from an AR(4) equation; the
other using the average of two leads of inflation, yielded similar data on real interest rates. It is difficult
to argue that negative real interest rates during this period reflect government controls on nominal interest
rates because a significant fraction of government bonds were sold on the open market to private
investors.
15



(18)                     ln(wbar), = eo + eln(AX), + 02CFt + e t
p
(19)                      In(p%) = P0 + P1ln(Px), + P2CFt + s4,
p 
Note that there is a separate equation, (19), for machinery prices and that (17) and (18) have the
averages of pi, and w;, across sectors on the left rather than the actual pj, and wj, (these are weighted
averages using investment shares for weights). This simplifies the presentation. Both (17) and (18) were
estimated sector by sector, but nothing essential in the argument is lost by reporting results from the
simpler framework above. Note also that D, in : ¢-) stands for any of the four debt-crisis proxies, and
that quarterly dummies were included in the estimation of (16) but not shown above.
Equations (16) through (19) make explicit that the effect of the terms of trade on investment is
indirect, operating through product prices, wages and the machinery price. The debt crisis is allowed
to affect investment in two ways. First, the capital flow variable affects investment indirectly by affecting
the relative prices. Second, debt-overhang and/or uncertainty are allowed to affect investment directly
in equation (16).
The model underlying equations (16) through (19) treats the terms of trade, capital flows and the
domestic interest rate as exogenous, and has a recursive structure. Although the recursive nature of the
model may be used to defend least squares estimation of the investment equation, there are also good
reasons to doubt whether this is the best approach. There could be standard supply and demand
simultaneity between investment and either the machinery price or the domestic interest rate. There may
also be a relationship between an unobserved variable in the investment equation and wages. For
16



example, the euphoria surrounding the oil boom in Mexico may have been associated with both an extra
stimulus to investment and wage bonuses or other implicit rent sharing arrangements with labor.
I will use three instruments in an attempt to deal with this issue: the terms of trade, the capital
flow variable, and the LIBOR interest rate. The first two instruments are clearly suggested by the model
above. It also seemed natural to include an international interest rate to instrument for a domestic interest
rate.12 With three instruments, the estimates of (16) treating three variables as endogenous, namely the
machinery price, wages and the interest rate, are exactly identified.
The reported regressions in table 2 have an error structure that was determined through some
preliminary examination of the residuals. First, a Bartlett test overwhelmingly rejected the hypothesis
that the error variances were equal for each sector. Second, separate Durbin-Watson statistics computed
for each of the 42 sectors were either close to 2 or were in the inconclusive range. Somewhat
surprisingly, only 3 of the 42 Durbin-Watson statistics provided evidence for positive serial correlation
(the regressions in table 2 report the average Durbin-Watson statistic across all sectors). Third, the
sample correlations over time of the residuals in all 42*41/2 pairs of sectors were examined and only
about 5 percent were even above 0.45 in absolute value. In sum, there was strong evidence for
heteroscedasticity but not very strong evidence for other departures from a diagonal error structure: hence
weighted least squares estimates will be reported below.
6. Results,
Table 2 presents instrumental variables estimates of the investment equation (16). The dummy
variable D, is a crude proxy to catch debt-crisis effects; it is coded to equal one after the second quarter
of 1982, and therefore should have a negative sign. MEXB measures the creditworthiness of the Mexican
government as perceived by the international financial community, and should have a positive sign.
12Several forms of the interest parity relationship imply that international interest rates can be treated
as instruments for domestic interest rates.
17



MEXE is an index of stock prices for Mexican companies and should also carry a positive sign. And
MEXCW is an index of creditworthiness for Mexico as a whole, and should again have a positive sign.
All of these are interpreted as imperfect, but still informative, indicators of debt-overhang and/or
uncertainty.
Of these four debt variables, only one (the dummy variable DJ) is significant and has the negative
sign predicted by the debt theories, while the other three are insignificant, and among these, the
coefficient on MEXB has the wrong sign. These results do not support either a complete rejection or
acceptance of the various debt hypotheses, but on balance cast doubt on their importance.
It is worth noting four additional points about the debt variables. First, the debt dummy is
arguably the most crude way to measure debt overhang effects, and yet it is the only one which shows
significant debt effects. Second, although the crudeness of the variable is an issue, in defence of the debt
theories, the estimated effect is large. The coefficient of -.83 indicates that ceteris paribus, this variable
can account for 70 percent (-.83/-1.18) of the investment decline in the sample."3 Third, fear of higher
investment taxes to solve government solvency problems is one of the main mechanisms offered to
explain the precise transmission from debt problems to investment. Yet, the most precise indicator of
perceived solvency of the Mexican government, MEXB, is also the worst performer of all the debt
variables. Fourth, MEXCW is arguably the best measure to capture international credit rationing effects
since it comes from direct surveys of country risk perceptions by major international banks, but it is not
significant either.
To check further the significance of the dummy variable, the investment sample was restricted
to the eight largest exporting sectors (sectors which sold more than 10 percent of their output outside
Mexico in 1980). One version of the debt-overhang hypothesis in Sachs (1988) states that investment fell
" Note however that three-stage least squares estimates of the same equation produced a somewhat
smaller coefficient of -0.63 (0.43).
18



because exporters feared expropriation of their output by foreign creditors. While it is unclear whether
this effect can have a large impact on total investment in a country which exported at most 20 percent
of output, it may still be expected to have a larger effect on sectors that export than those that do not.
However, when the model in column 2 of table 2 was re-estimated on these eight sectors, the debt-
overhang effect as measured by this dummy was actually smaller (-0.425, se=0.784)."
While the results in table 2 cast some doubt on the importance of debt crisis effects stressed in
the literature, they also point to a major effect that has not been emphasized. The most important result,
both in terms of size, statistical significance, and robustness across specifications, is the large estimated
machinery price elasticity, which ranges from -1.68 to 4.75.15 Since the relative price of machinery
rose by about 40 percent over the sample period, this is an important microeconomic channel behind the
investment decline, perhaps the single most important channel.
Another result in table 2 that seems robust across specifications is the positive elasticity on the
product price variable of about 1.0, although sometimes this is not significant. It should be mentioned
that the regression is controlling for changes in the prices of intermediate inputs through this variable
since the prices are value added prices."6
It is hard to draw reliable conclusions about the remaining two variables: wages and real interest
"' The other estimated coefficients, on p, w, p', and r, were 1.52 (1.66), -2.86 (5.14), -6.00 (3.74),
and -0.013 (0.010).
1" The low value of -1.68 comes from the regression where the debt crisis proxy, ln(MEXB), is
statistically insignificant and has the wrong sign. Of the regressions with correctly signed debt crisis
proxies, the range is -2.58 to 4.75. One reason that the regression with MEXB differs from the other
regressions is that MEXB is not highly collinear with the other debt crisis proxies. It measures market
sentiment about the Mexican government's creditworthiness rather than that of the entire country. This
variable rose during 1984-1985, whereas the other debt crisis proxies remained depressed.
16 On the issue of whether to treat the price variable as endogenous, I found that the residuals from
the price regression (equation 17 with p, on the left) had virtually zero correlation with investment or the
residuals from the investment equation, suggesting that endogeneity is not an issue. Regressions that
nevertheless treat the price variable as endogenous yield large negative price coefficients with extremely
high standard errors.
19



rates. The wage variable switches signs across specifications and is typically not significant. It is only
significant once MEXB is included in the regression, but MEXB itself has the wrong sign and is not
significant.
The estimated real interest rate coefficients range from about zero to -0.0045. The -0.0045
estimate means that a percentage point increase in real interest rates would reduce the log of investment
by 0.0045, and thus would reduce investment by 0.45 percent. This effect seems fairly small: if real
interest rates rose by 20 percentage points, the -0.0045 estimate predicts that investment would decline
by only 9.4 percent. While there are reasonable objections to the simple measure of real interest rates
used in this regression, alternative measures yielded smaller estimated effects.17
The estimates in table 2 also facilitate a more precise quantitative assessment of the impacts of
the right-hand-side variables on the investment decline. Two of the right hand side variables, real product
prices and real interest rates, did not change enough between 1981 and 1985 to have been major players
in explaining the investment decline, whatever their estimated coefficients. Of the other variables, real
machinery prices rose by about 30 percent and real wages fell by about 30 percent. The wage effect is
not reliably estimated and thus will not be stressed. In contrast, the machinery price effect seems reliable
and important. Taking the smallest estimated coefficient from table 2 of -1.68, the observed rise in
machinery prices can explain 46 percent (-1.68*0.324/-1.18) of the investmnent decline; taking the largest
estimated coefficient, it can explain 130 percent. The evidence therefore marks out machinery prices as
an important channel of transmission to investment.
Table 3 presents the estimates of equations (16) - (19). The estimates of the product price
equation (16) offer no evidence that the terms of trade decline or the capital flow reversal affected real
17 A proxy for expected inflation was calculated using one-step-ahead forecasts from an AR(4) time
series regression: the estimated coefficient was -0.0018 (0.0007). A proxy for the return on holding U.S.
bonds was also constructed using data on the 3-month futures premium for the Mexican Peso (an indicator
of expected devaluation), and U.S. 3-month treasury bill rates: the estimated coefficient was -0.0029
(0.0012).
20



product prices. On the other hand, there is evidence that the terms of trade reduction depressed real
wages. The absence of any effect on product prices probably indicates that there is little difference
between the traded goods content of the product price indexes in the numerator, and the CPI in the
denominator. In 1980, the sectors covered by this data exported only about 6 percent of their output on
average.
Table 3 provides strong evidence that the terms of trade decline served to increase the relative
price of machinery and somewhat weaker evidence that the capital flow reversal had a similar effect.
While Mexico sold only a small fraction of its non-oil output abroad during this period, at least 50
percent of machinery investment in Mexico was imported. Mexico was a price-taker in the world market
for traded machinery. Therefore, a demand contraction caused by the oil price decline could be expected
to reduce the CPI, with its high non-traded content, relative to the price of machinery, and yield a
negative relationship between p'/pm and p'/CPI. Note that in the machinery price equation, the
coefficients on both variables are negative, in line with this reasoning. The terms of trade coefficient is
easily significant, and the capital flow coefficient is marginally significant. In other specifications which
extended the sample size and tried other slightly different measures of capital flows, the terms of trade
was always significant, while the capital flow variable was sometimes insignificant. The evidence
strongly suggests that the terms of trade decline played an important role in explaining the rise in the
relative price of machinery.
The estimated coefficients in table 3 can also be used to split the increase in machinery prices into
the part attributable to the terms of trade and capital flows. Multiplying the estimated coefficients with
the actual changes in the right had side variables, it turns out that the change in the terms of trade
accounts for 67 percent 100*(40.836)(40.302)/(0.379) of the machinery price increase (the denominator
is the change in the fitted value of the dependent variable), while the capital flow reduction accounts for
the remaining third.
21



The model so far has allowed capital flows to affect investment only indirectly; that is, through
their effect on prices. Alternatively, a reduction in foreign capital inflows may shift the domestic saving
function to the left and cause a simultaneous rise in domestic interest rates and a reduction in investment.
In such a model capital flows would have a direct effect on investment even after controlling for other
relative price movements. However, adding the capital flow variable to the estimated investment function
yields a small and insignificant coefficient, 0.0025 (0.0178). Thus the data do not support the view that
the capital flow reduction directly depressed investment after controlling for the effect of relative price
movements.'8
7. Evidence with Annual Investment Data.
To assess the generality of the results both across time and for larger investment aggregates, this
section exan ines regressions using total investment in Mexico over a longer time span (1970-1988). The
estimated equations, reported in table 4, are reduced forms of the model outlined above: investment is
regressed on the terms of trade, capital inflows, an external interest rate, and other variables to capture
debt problems. The real investment data used as the dependent variable are scaled by population to
control for trend growth. It turns out that scaling by GDP delivers similar results.
The debt crisis variables in the regressions are the ratio of debt to GDP, and a dummy variable
taking the value one after 1981. The regressions use government consumption spending as an instrument
for the debt variable. The reason is to use the part of the variation in the debt variable that was driven
by government consumption spending rather than investment spending, because the latter may be
spuriously related with investment. The table shows that this part of the variation in the debt variable
is positively associated with investment after controlling for the other exogenous variables. This evidence
1' This can be seen as a test for what Borensztein (1990) calls the credit rationing effect. Note also
that this effect predicts that domestic real interest rates should have been higher after the debt crisis than
before. The Mexican real interest rates used in this paper, which cover the 1981-1985 period, are not
clearly higher after 1982:3 than before.
22



therefore does not support the simple notion that accumulated debt represents an investment deterrent.
The table also shows that the simple debt dummy either is not significant (column 1) or has the wrong
sign (column 5).1" Combining these results with the above results using the quarterly investment data,
it seems that the only debt indicator which provides evidence for debt-overhang effects is the debt dummy
in the quarterly regressions. On balance the evidence for this kind of debt effect is not strong.
The evidence in table 4 for the other kind of debt crisis effect, a demand compression due to the
reversal in capital flows, is somewhat more supportive but also mixed. This variable is always positively
signed as anticipated, but only significant when other incorrectly signed debt or interest rate variables are
included in the regression. In column 4, when the terms of trade and the capital flow variable are alone
in the regression, the terms of trade variable is significant while capital flow variable is not.
In contrast, the estimated terms of trade effect is significant in table 4, regardless of the other
variables controlled for in the regression. To gauge the size of the estimated effect, between 1981, when
investment in Mexico was at its highest, and 1988, when it was at the lowest level in the recent past, the
log of the investment variable changed by -0.725. During the sane period, the log of the terms of trade
index changed by -0.740. Using the estimated coefficient of 0.742, the terms of trade effect by itself can
account for about 75 percent (-0.742*0.740/-0.725) of the investment decline between 1981 and 1988.
Hence this evidence from annual data supports the earlier evidence on the importance of the terms of
trade effect.
8. Conclusions and discussion.
This paper has developed and estimated an investment model which distinguishes three ways in
which the debt crisis and Mexico's terms of trade decline could have affected investment in Mexico. The
model allows the capital flow reversal and the terms of trade reduction to affect aggregate demand and
19 Coding the debt dummy to equal 1.0 after 1982, rather than 1981, yields an insignificantcoefficient
on D in the regression in column 5 (0.067, se=0. 182), a significant coefficient on the terms of trade
(0.800, se=0.284) and an insignificant coefficient on the capital flow variable (0.007, se=0.01 1).
23



relative prices, and ultimately to affect investment through these prices. The model also allows debt
overhang effects or uncertainty effects to depress investment directly.
One of the main findings from this analysis is that only one of several specifications provide
evidence for debt overhang or unce,tainty effects. Overall, there is not robust evidence that this was the
main reason for Mexico's large investment decline in the "lost decade" of the 1980s.
The evidence instead supports two main points that have been underemphasized in the debt
literature. First, the main proximate cause of the investment decline was the rise in the relative price of
machinery between 1981 and 1985. This variable is the most reliable across specifications in accounting
for the investment decline. The argument to explain this rise is based on the fact that machinery is
essentially a traded good in Mexico. As spending declined in response to the terms of trade decline or
the capital flow reversal, this demand contraction reduced other product prices relative to this price,
depressing investment demand.
Second, the decline in Mexico's international terms of trade during this period, driven by falling
world oil prices, is probably the most important ultimate cause of this increase in relative machinery
prices, but the reversal in net capital inflows to Mexico may also have played an important role. The
econometric evidence using quarterly data during the 1981-1985 period indicates that both the terms of
trade reduction and the capital flow reversal played a role in increasing this relative price. Calculations
using the estimated coefficients and the actual changes in these two variab:es between 1981 and 1985
suggest that about two-thirds of the machinery price increase can be attributed to the terms of trade
decline, and the other third to the capital flow reversal. But in reduced form regressions using annual
data over a longer time span, it is unclear whether the capital flow variable is significant.
Finally, it is also worth mentioning that investment in two neighboring economies which were
also riding the oil boom of 1973-1981, namely Texas and Louisiana, also fell during the 1981-1986
24



period, and that adverse commodity price shocks also affected many other heavily indebted countries.0
Therefore, at the very least, the direct role of commodity price shocks such as Mexico's oil price decline
in causing low investment levels in the 1980s has teen insufficiently emphasized in the literature on the
effects of the international debt crisis.
20 These facts are from Warner (1992).
25



Table I
DATA
Variable          Mean            Description                Source
Ijt               47.45           Real Machinery Investment   Bank of Mexico
P= pj,/CPI,       62.94           Value Added Price Index    Indicadores Econonucos
wj, = w/CPI,      72.25           Real Wage Index            Indicadores Economicos
P= PI,/CPT,        3.71           Price Index for Machinery   Indicadores Economicos
CF,                0.34           Net Capital Inflows        Indicadores Economicos
r.               -10.79           Real Interest Rate         Indicadores Economicos
LIBOR,            11.84           Nominal LIBOR Interest Rate DRI
pxVPMt            78.87           Terms of Trade Index       Bank of Mexco
D,                 0.35          Dummy Variable Indicating   -
Debt Crisis (= I if t> 82:3)
MEXB,             81.35          Price of a Mexican          New York Stock Exchange
Government Bond
MEXF,              4.35          Price of Mexico Fund        New York Stock Exchange
MEXCW,            47.29          Index of Mexico's          Institudonal Investor
Creditworthiness
The j index ranges across sectors: for the investment sub-sample in the regressions, j = 1,..,42; for prices
and wages j = 1,..,9. The matching scheme, reported in the appendix, is not one-for-one.
The time index is quarterly, t= 1981:1,..,1985:4.
MEXCW is published bi-annually; quarterly data was obtained by linear interpolation.
Other measures of the real domestic interest rate, r, were used in the regressions but not reported above.
The index numbers were not scaled to a common base year because the relevant regressions were
estimated in logs.
26



Table 2
Instrumental Variables Estimates of Equation (16)
Dependent Variable: In of Real Investment, In()
Independent Variable                   Estimated Coefficients (Standard Errors)
Ln(p),                    1.01           1.90          0.60          0.78          1.14
(0.53)        (0.83)        (0.46)        (0.41)        (0.41)
Ln(w)j,                   0.63          -3.20          2.13          1.00         -0.01
(1.21)        (2.66)        (0.72)        (0.86)        (0.76)
Ln(p'),                   -3.20         4.75          -1.68         -2.58         -3.18
(1.30)        (1.85)        (0.78)        (0.79)        (1.62)
-t                         - .0037       -.0102         .0008        -.0030        -.0045
(.0027)       (.0048)       (.0026)       (.0019)       (.0014)
D,                         -            -0.83           -            -
(0.38)
Ln(MEXBX                   -             -            -0.43          -
(0.30)
Ln(MEXF),                  -             -             -             0.08
(0.11)
Ln(MEXCW),                -              -             -             -             0.38
(1.05)
R2                        0.50           0.37          0.51          0.51          0.49
SE                        1.02           1.07          1.00          1.01          1.02
DW                        1.63           1.58          1.67          1.64          1.62
N-K                     790           790            790           790           790
The instruments are the log of the terms of trade, the capital inflow variable, and the LIBOR interest rate.
The endogenous variables are the wage, the price of machinery and the domestic real interest rate. The
error structure allows for sector-specific error variances, estimated in a first-stage regression. This table
does not report the separate intercepts estimated for each sector nor the coefficients on quarterly
dummies.
27



Table 3
OLS Estimates of Equations (17) to (19)
(Robust Standard Errors in Parentheses)
Independent Variables                          Dependent Variables
Ln(pbar),     Ln(wbarX       Ln(p'),
pVpm                                    -0.051         1.057*       -0.836*
(0.163)       (0.339)      (0.265)
CF,                                      0.002         0.007        -0.026
(0.007)       (0.018)      (0.014)
R2                                       0.040         0.642         0.734
The reported standard errors are robust to quite general forms of serial correlation and heteroscedasticity.
They follow Wooldridge (1989), which proposes a computationally simple procedure. The sample for
all regressions is quarterly, 1981:1 to 1985:4. Statistical significance at the 5 percent level is indicated
by a * next to the estimated coefficient.
28



Table 4
Estimates of the Reduced Form using Annual Data
and Total Investment
Dependent Variable: In of Real Per-Capita Investment, ln(I)
Independent Variable                Estimated Coefficients (Standard Errors)
In(pVp'o                0.624         0.742        0.621        0.742        0.956
(0.300)      (0.202)      (0.207)      (0.229)      (0.211)
CFt                     0.024         0.031        0.025        0.004        0.042
(0.017)      (0.010)       (0.010)     (0.007)      (0.015)
rus,                    0.047         0.031        0.053
(0.038)      (0.022)      (0.022)
ln(DEBT/GDP),           0.217         0.220         -
(0.108)      (0.951)
D,                      -0.202        -             -            -           0.580
(0.439)                                              (0.218)
R2                      0.673         0.691        0.620        0.503        0.640
SE                       0.130        0.126        0.139        0.158        0.135
DW                      2.090         2.126        0.878        0.743         1.215
N-K                    11            12            15          16            15
The terms of trade and capital flow variables are annual version of the quarterly variables used earlier
in this paper. Of the other new variables, rust is the U.S. 10 year t-bond rate, DEBT/GDP is Mexico's
total debt in dollars, multiplied by the average peso/dollar exchange rate, and divided by nominal
Mexican GDP (source: World Debt Tables, World Bank, and Indicadores Economicos, Bank of Mexico).
Dt is a dummy equal to one after (and including) 1982. The regressions in the first two columns use
current and lagged government consumption expenditures as an instrument for the DEBT/GDP variable;
the other regressions are OLS. The sample is 1970-1988 except when the debt variable is used, in which
case it is 1972-1988.
29



References
Anthony, M. and Hughes-Hallett, A. (1992) "How Successfully Do We Measure Capital Flight?", Ihe
Journal of DevelopmeEt Studies, 28, 538-556.
Borensztein, E., (1990) "Debt Overhang, Credit Rationing, and Investment", Journal of Development
EconQmics, 32, 315-335.
Bernanke, Ben S., (1983) "Irreversibility, uncertainty, and cyclical investment", Quarterly Journal of
Economics, 98, 85-106.
Buffie, E. F., (1990) "Economic Policy and Foreign Debt in Mexico," in Developing Country Debt
and Economic Performance, Vol. 2: Country Studies, ed. by Jeffrey Sachs (University of Chicago
Press, 1990)
Bulow, J. and Rogoff, K. (1991) "Sovereign Debt Repurchases: No Cure for Overhang," Quarterly
Journal of Economics, 106, 1219-1234.
Cohen, D., (1990) "Slow G-owth and Large LDC Debt in the Eighties: an update." mimeo,
CEPREMAP, October.
Cukierman, Alex, (1980) "The effects of uncertainty on investment under risk neutrality with endogenous
information" Journal of Political Economy, 88.
Dixit, A. (1989) "Interseatoral capital reallocation under price uncertainty", Journal of International
Economics, 26, 309-325.
Dornbusch, R., (1988) "Mexico: Stabilization, Debt and Growth," Economic Policy, 3, 233-283.
Faini, R., and de Melo, J., (1990) "Adjustment, Investment and the Real Exchange Rate in Developing
Countries", Economic Policy, 11, 492-512.
Froot, K. (1989) "Buybacks, Exit Bonds, and the Optimality of Debt and Liquidity Relief',
International Economic Review, 30, 49-70.
Fry, M., (1989) "Foreign Debt Instability: an analysis of national Saving and domestic investment
responses to foreign debt accumulation in 28 developing countries.", Journal of International Money and
Finance, 8, 315-344.
Hayashi, F. (1982) "Tobin's Marginal q and Average q: A Neoclassical Interpretation", Econometrica,
50, 213-224.
Helpman, E. (1989) "Voluntary Debt Reduction: Incentives and Welfare", Staff Papers. International
Monetarv Fund, 36, 580-611.
Hofmnan, B., and Reisen, H., (1990) "Debt Overhang, Liquidity Constraints, and Adjustment Incentives",
Technical Paper No. 32, OECD Development Centre, October.
30



Ize, A., (1988) "Savings, Investment and Growth in Mexico: Five years after the crisis", mimeo,
IMF Fiscal Affairs Department, December.
Kraft, J. (1985) "The Mexican rescue", New York, Group of Thirty.
Krugman, P. (1985) "International Debt Strategies in an Uncertain World," in Smith and Cuddington,
eds., International Debt and the Developing Countries, World Bank.
, (1988) "Financing vs Forgiving a Debt Overhang", Journal of Development Economics,
29, 253-268.
Neary, 1. P., and van Wijnbergen, S., eds., (1986) "Natural Resources and the macroeconomy.",
Cambridge, MA, MIT Press.
Rodrik, D., (1991) "Policy Uncertainty and Private Investment in Developing Countries", Journal of
Development Economics, 36, 229-242.
Sachs, J., (1988) "The Debt Overhang of Developing Countries," in Ronald Findlay, ed., Debt, Growth
and Stabilization: Essays in Memory of Carlos Diaz Aleiandro, Oxford, Blackwell.
Savvides, A., (1992) "Investment Slowdown in Developing Countries During the 1980s: Debt Overhang
or Foreign Capital Inflows?", Kyklos, 3.
Solis, L., and Zedillo, E., (1985) "The Foreign Debt of Mexico" in G. W. Smith and J. T. Cuddington
International Debt and the Develovine Countries, World Bank.
Warner, A., (1991) "The Debt Crisis, World Investment Cycles, and American Exports", PhD
Dissertation, Harvard University, May.
, (1992) "Did the Debt Crisis Cause the Investment Crisis?", Ouarterly Journal of Economics,
T07, 1161-1186.
Wooldridge, J., (1990) "Regression-based inference in linear time series models with incomplete
dynamics", MIT working paper No. 550, 1-53, April.
31



Policy Re"arch Working Paper Series
Contact
Titb                            Author                  Date              for paper
WPS1082 What Do Govemments Buy? The      Shantayanan Devarajan   February 1993     C. Jones
Compositon of Public Spending and  Vinaya Swaroop                         37699
Economic Performance            Heng-fu Zou
WPS1083 Finance and Growth: Schumpeter    RobertG. King          February 1993     D. Evans
Might Be Riht                   Ross Levine                               38526
WPS1084 Stock Market Development and     Dong He                 February 1993     P. Infante
Financial Deepening in Developing   Robert Pardy                          37665
Countries: Some Correlation Pattems
WPS1085 Economic Approaches to Modeling   Cristino R. Arroyo iII  February 1993    0. Nadora
Fertility Determinants: A Selective Review                                31091
WPS1086 Teachers' Salaries In Latin America: George Psacharopoulos   February 1993  G. Psacharopoulos
A Comparative Analysis          Jorge Valenzuela                          39243
Mary Arends
WPS1087 Exchange-Rate-Based Stabilization: Alberto F. Ades       February 1993     R. Luz
Tales from Europe and Latin America Miguel A. Kiguel                      34303
Nissan Liviatan
WPS1088 A Primer on the MFA Maze         Rioardo Fainl           February 1993     D. Ballantyne
Jaime de Meio                             37947
Wendy Takacs
WIPS1 089 Equity Portfolio Investment in  Stiun Claessens        February 1993     R. Vo
Developing Countries: A Literature Survey                                 31047
WPS1I90 Gove:nment Expenditures as a     Thanos Catsambas        February 1993     A. Correa
Citizen's Evaluation of Public Output:                                    38549
Public Choice and the Benefit Principle
of Taxation
WPS1091 Capital Market imperfections Before  Fidel Jaramillo     February 1993     W. Pitayatonakam
and After Financial Liberalization:    Fabbo Schiantarelli                37658
A Euler-Equation Approach to Panel Andrew Weiss
Data for Ecuadorian Firms
WPS1092 The Effect of Financial Liberalization  Fidel Jararnillo  February 1993    W. Pitayatonakarn
on the Allicaton of Credit: Panel    Fablo Schiantarelli                  37658
Data Evidence for Ecuador       Andrew Weiss
WPS1093 Swiss Chilanpore: The Way Forward  Dimitri Vittas        February 1993     W. Pitayatonakam
for Pension Reform?                                                       37658
WPS1 094 The New Reglonalism: A Country    Jaime de Melo         February 1993     D. Ballantyne
Perspective                     Arvind Panagariya                         37947
Dani Rodrk



Policy Research Working Paper Series
Contact
Title                           Author                  Date              for paper
WPS1095 Are Faliproof Banking Systems    Samuel H. Talley        February 1993     B. Lu
Feasible? Desirable                                                       37664
WPS1096 How Trade Liberalization Affected    Mona Haddad         February 1993     D. Ballantyne
Productivity in Morocco                                                   37947
WPS1097 A Production Function-Based Policy  Takamasa Akiyama     February 1993     G. llogon
Simulation Model of Perennial   Jonathan R. Coleman                       33732
Commodity Markets
WPS1098 Fortress Europe and Other Myths    Jean Baneth           February 1993     M. St. Leger
Concerning Trade                                                          37148
WPS1099 Trade Unions and Collective      Harry C. Katz           February 1993     ESP
Bargaining                      Sarosh Kuruvilla                          33680
Lowell Tumer
WPS1 100 Indigenous Territories and Tropical   Shelton H. Davis  February 1993     S. Davis
Forest Management in Latin America  Alaka Wali                            33413
WPS1101 Transforming State Enterprises In   Brian Pinto          February 1993     M. Cheong
Poland: Microeconomic Evidence on  Marek Belka                            39618
Adjustment                      Stefan Krajewski
WPS1 102 Did the Debt Crisis or Declining Oil   Andrew M. Warner  February 1993    M. Divino
Prices Cause Mexico's Investment                                          33739
Collapse?
WPS1 103 Capital Mobility in Developing  Peter J. Montiel        February 1993     R. Vo
Countries: Some Measurement Issues                                        31047
and Empirical Estimates