Testing capital account liberalization without forward rates Another look at Chile 1979–1982

ELSEVIER

Journal of DevelopmentEconomics Vol. 52 (1997) 139-168

JOURNAL OF Development ECONOMICS

Testing capital account liberalization without forward rates Another look at Chile 1979-1982 William F. Maloney University of Illinois, Urbana-Champaign, Illinois, USA Received 6 July 1995

Abstract The paper offers a way to test the impact of capital account liberalization on domestic interest rates when proxies for expectations of depreciation are unavailable. The model is tested on the well-understood Japanese experience in 1979-1980 and then applied to the Chilean liberalization from 1979-1982. The evidence points to Chilean monetary authorities retaining significant control over interest rates across the period and offers support to an evolving interpretation of the period that stresses the excessive optimism about the reforms, rather than their lack of credibility, as the source of their ultimate collapse. JEL classification: F32; F4; O19 Keywords." Capital mobility;Chile; Interest parity; Japan; Liberalization

1. Introduction It is now common to measure the degree of capital mobility among industrialized countries by testing for covered interest parity (CIP): in the absence of barriers to arbitrage, similar assets should yield similar returns across countries when adjusted for the risk of currency realignment embodied in the forward discount. In the same spirit, convergence toward CIP has been used as a test of capital account liberalization by among others Frankel (1984) and Eken (1984) for 0304-3878/97/$17.00 Copyright © 1997 Elsevier Science B.V. All rights reserved. Pll S0304-3878(96)00436- l

140

W.F. Maloney/ Journal of Development Economics 52 (1997) 139-168

Japan, and Chinn and Frankel (1992) for the Pacific basin. Yet, for a majority of countries that are now contemplating liberalization, forward markets are nonexistent, rendering the straightforward arbitrage test inapplicable. This paper departs from the idea that it is possible to extract additional information from the behavior of the coefficients on the domestic determinants of interest rates; as a capital account opens, it would be expected that these coefficients should decrease in importance. This is implicit in the growing literature that has emerged based on an intuitively appealing technique introduced by Edwards (1985a, 1986) and Edwards and Khan (1985) to estimate the degree of capital market integration in Colombia, Chile, and Singapore. In essence, their model takes the interest rates that would prevail under the two extremes represented by a closed economy interest rate determination model (R') and the interest parity condition (R * ) and then weights them using a 'coefficient of openness', q~: R =

+ (1 -

(1)

A semi-reduced form equation can be derived for the domestic interest rate in terms of income, the real money supply, and the foreign return which, when estimated, permits q~ to be recovered. Building on this basic framework, Haque and Montiel (1990) inverted the final specification estimating money demand rather than interest rate equations and derived coefficients of openness for 15 countries from 1969-1987. Dooley and Mathieson (1992) employed a similar methodology for seven less developed countries (LDCs) from the mid-1960s to 1990 and tested for structural break in the relation across periods of capital account liberalization. Most recently, Reisen and Y~ches (1993) employed the original Edwards and Khan specification to measure the openness of the Korean and Taiwanese capital accounts. One objection to this approach is that it is does not make clear how the implicit averaging of closed and open economy interest rates arises from the behavior of individuals in the economy. As an example of why this is important, the present work addresses the question within a portfolio balance framework, adapting to an inflationary context earlier theoretical work by Kouri and Porter (1974) for a fixed exchange rate regime. Modeling agents' demands for domestic and foreign assets subject to a wealth constraint and then letting foreign assets become more substitutable for domestic assets as liberalization proceeds allow the derivation of a semi-reduced form for interest rates that differs critically from that of the Edwards and Khan tradition. In particular, the framework raises some questions about the interpretation of the Haque-Montiel real money balances specifications and, more importantly, about q~ as an absolute coefficient of openness. The paper then shows how the relative movements of the domestic coefficients of the portfolio balance specification can be used to draw inferences about the effectiveness of liberalization measures for two well-known capital account liberalizations. First, the reduced form specification is tested by tracking the determinants of Japanese interest rates across 1979-1980, a liberalization that has been

W.F. Maloney / Journal of Development Economics 52 (1997) 139-168

141

extensively analyzed using the interest parity approach, i It is shown that the same conclusions about the opening of the capital account can be deduced even when no measure of covered foreign returns is included in the regression, thus showing the validity of applying the methodology to situations where no forward rates exist. One such case, the celebrated and controversial capital account opening in Chile from 1979 to 1982, has presented something of a puzzle for ten years and one which this paper attempts to resolve. 2 Following the elimination of most capital controls, and amidst vast inflows, domestic interest rates rose in real terms to levels of 40% for over two years and contributed integrally to the eventual collapse of the economy. 3 Some observers (Dornbusch, 1985; Edwards, 1985b, among others) assume interest parity and hold responsible expectations of depreciation, brought on by imprudent exchange rate policy. Others (Harberger, 1985, and Arellano, 1983, among them) have stressed extraordinary domestic demands for credit as an alternative explanation, thereby implicitly embracing the view that the financial sector behaved as if relatively closed. By estimating the derived reduced form across the period of liberalization, the movements in the coefficients allow for distinction between these two views. The paper is thus similar in focus and complementary to the very ambitious complete system approach developed by McNelis and Schmidt-Hebbel (1993) for Chile and New Zealand. After explicitly modeling the key endogenous variables, the authors derive a reduced form specification for the nominal interest rate that differs substantially from that presented here and estimate it using Kalman filtering techniques to show how liberalization alters the relevant parameters. They conclude that " F o r Chile, covered arbitrage interest levels ... have a dominant influence on domestic nominal rates during a period characterized by massive changes in the access to foreign credit markets" (p. 268), and find that the impact of domestic credit decreases and that of covered foreign rates rises during the process of liberalization. This paper concludes the opposite. The results point to Chilean interest rates behaving as if the capital account were closed for much of the period after liberalization, despite massive inflows. A very satisfactory specification that explains most of the movement in nominal rates prior to liberalization with no influence of international interest rates shows no evidence of structural break at

i The ideal test case would have a well-developed forward market, credible data, a clear liberalization, and fixed exchange rates, to be strictly described by the Kouri and Porter model. However, the countries that meet the first three criteria, such as Japan, are among the industrialized economies that are perhaps best characterized as dirty floats since the early 1970s. 2 The Chilean case is well described by the Kouri and Porter model: from July 1976 to June 1979, the government employed a crawling peg system where small devaluations were done daily in accord with a preannounced formula. From June 1979 to June 1982, the exchange rate was fixed at 39 pesos per dollar. 3 Galvez and Tybout (1985); Mizala-Salces (1985).

142

W.F. Maloney/ Journal of Development Economics 52 (1997) 139-168

any point associated with the reduction of controls on capital inflows. Further, a very similar specification is then fit to the post-liberalization period again, with no recourse to international variables. By sequentially dropping variables, the influence of individual domestic factors is isolated. The evidence points to the fact that monetary authorities retained substantial control over interest rates across the period and that a non-accommodating and then very contractional domestic credit policy accounts for much of the rise in nominal and real rates. The key to the paradox of domestic factors dominating despite vast capital inflows is suggested to lie in the discontinuous nature of the capital inflows caused by regulations on short term inflows, and a persistent excess demand for credit. The findings offer support to an evolving interpretation of the period that stresses the excessive optimism about the reforms, rather than their generally assumed lack of credibility, as the source of the collapse of the economy in 1982.

2. Modeling a liberalizing capital account As with the several studies in the Edwards and Khan tradition cited earlier, the model developed here is partial equilibrium in nature and therefore can offer insight into how the results of those studies may be reinterpreted. Given the relative rapidity of adjustments in the financial sector, the lack of a general equilibrium framework is unlikely to be a major drawback in the tests for break, especially if potential endogeneity effects are controlled for at the estimation stage. It is assumed that in a simple economy domestic agents can hold four assets in their portfolios which are gross substitutes: money (M), domestic bonds (B) paying a return R, a foreign bond ( F ) paying R*, a nominal interest rate plus expectations of depreciation and a risk premium, and an asset, perhaps land (L), effectively indexed to inflation (Tr). 4 The wealth constraint

W=M+B+F+L

(2)

implies that all arguments enter into all asset demand functions so that

A d = A ( R , R * ,Tre,Y,W) VA = { M , B , F , L } .

(3)

4 In the long run, the return to bonds, representing a claim on real productive assets, should reflect the real return to capital plus expectations of inflation. In the short run, however, the real ex-ante return on bonds, in the present case bank deposits, can far exceed or fall short of any reasonable return to capital due to short-term monetary disequilibria. In the absence of land, or any other asset effectively indexed to inflation, in a closed economy, both money and bonds will be equally affected by inflation and bond returns will not necessarily reflect expectations of inflation. In an open economy, interest rates will track the international return which may or may not reflect expectations of inflation. McNelis and Schmidt-Hebbel (1993) explicitly model the risk premium as a function of the stock of foreign assets. Here it is taken, as expectations of depreciation, to be unobservable and not easily modeled and is excluded from the estimations.

W.F. Maloney / Journal of Development Economics 52 (1997) 139-168

143

The wealth constraint also implies that ~-,Ai _ ~ _ A A = 0 V i = { Y , R , R * ,Tr e} A --A h i

(4)

and that AA EAw = E a w = l. A

(5)

A

Monetary equilibrium requires that AM(Y,R,R*,Tre,w)

= A D C + CA + KA,

(6)

where D C is domestic credit and CA and KA, the current and capital accounts respectively, reflect the reserve component of the money supply. It is assumed, for expositional purposes, that the money multiplier is unity. The capital account can be seen as the change in foreign demand for domestic assets, represented by an analogous demand function where superscript * indicates the foreign country, minus the change in domestic demand for foreign assets: KA = A B * ( Y * , R , R * ,,rr¢* ,W * ) - A F ( Y , R , R * ,Tr% W ).

(7)

Combining Eqs. (6) and (7), expanding the asset demands around their arguments, and then invoking Eq. (4) to solve for the change in the domestic interest rate with the second-order foreign arguments dropped yields a semi-reduced form equation: A R = ~ 3 l A D e +/32CA + / 3 3 A Y + / 3 4 A'/Te q-/35AW +/36AR*,

(8)

where 1

/31,/32 Ly + By /33

- - '

0

=

D

O'

L= + B~, /34 = - - '

0 BR, + B R,

/36

0

Lw + Bw /35-- - - '

0

(9) '

BR, ,B~, ,L r , B r,B~ < O, B R,B/¢ ,L~ ,L w ,B w > O.

The magnitude of substitutability between domestic and foreign bonds represented by the four domestic and foreign partial derivatives B determine the value of each coefficient in Eq. (8). In an entirely closed economy there is no substitution between domestic and foreign assets so both BR, and B~ = 0, making 0 finite and /36 = 0. This leaves something reminiscent of a Fisher equation where

144

W.F. Maloney/ Journal of DevelopmentEconomics 52 (1997) 139-168

expectations of inflation affect the interest rate while the nominal return on foreign assets and expectations of depreciation do not. As domestic and foreign bonds become more and more substitutable with liberalization, B R and B~ ~ oo and BR, and B R. ~ -oo, forcing 0 to infinity, all coefficients with finite numerators, /31 to 5, to zero, and /36 to unity. The intuition is straightforward. Any domestic factor that might increase the interest rate - a contraction of domestic credit, an increase in demand for transaction balances, or an increase in expectations of inflation - leads to an immediate capital inflow that forces the interest rate back to the world rate. The return to foreign bonds inclusive of expectations of depreciation completely determines movements in domestic interest rates as in the standard uncovered interest parity condition. Eq. (8) differs from that estimated by Khan and Edwards in essentials only by excluding the endogenously determined capital account contribution to the money supply, but the difference may be critical. First, the absence of the total money stock variable in Eq. (8) suggests that it would be difficult to generate the Haque-Montiel specification with money stock as the dependent variable. More importantly, by replacing CA + A D C with the total A M 1, Khan and Edwards in effect ask: If we control for changes in the money supply, including those caused by capital inflows, what impact do movements in the foreign return have on the domestic interest rate? This is likely to short-circuit the very effect being measured since changes in the money stock due to capital inflows are a critical channel of influence from the international to the domestic economy: a fall in the world rate leads to a capital inflow which expands the money stock and forces down domestic rates. It is probably not 'openness' that ~ is measuring, but perhaps the ease of reallocation of portfolios when asset stocks are fixed, i.e. when the capital account is closed. Nor, theoretically, should there be any structural break in the relation expected during liberalization unless this ease of reallocation were affected. By contrast, testing for structural break in Eq. (8) can give us information on whether liberalization measures have in fact led to greater capital mobility. Further, if the foreign return can be credibly claimed to be orthogonal to the other regressors, we can test for increasing capital mobility with no recourse to forward rates and interest rate parity.

3. Estimation

The approach is first tested on the well-known Japanese case where a covered foreign return is available and then applied to the Chilean case for which forward rates are absent. 5 Analysis of both the Japanese and Chilean data begins with the

5 PC-GIVE, David Hendry, 1991.

W.F. Maloney / Journal of Development Economics 52 (1997) 139-168

145

standard tests for unit roots and co-integration (see Appendix B). If appropriate, an error correction model is fit to the pre-liberalization period that is stable and parsimonious and possesses spherical, normally distributed disturbances. Only lagged values of all variables are included to avoid simultaneity bias. 6 In neither case is a credible proxy for wealth available, although to the extent that the economy is fully open, the cumulative current account may capture its effect. 3.1. Japan Comprehensive studies of the Japanese liberalization have been published by Frankel (1984) and Eken (1984), and this paper will not attempt to offer more detail than they have provided. Throughout the 1970s Japan imposed restrictions on foreign holdings of many types of financial instruments in an attempt to prevent the yen from appreciating against the other world currencies. 7 In Fig. 1, the differential in the 1978-1979 period relative to the post-liberalization period suggests that these controls were binding. With the rapid depreciation of the currency in 1979, the Japanese loosened the restrictions prohibiting nonresidents from purchasing Japanese bonds with less than five years' maturity that had effectively ruled out short term capital inflows. 8 The impact is evident again in Fig. 1: the three-month gensaki and Euroyen rates, both denominated in yen and therefore absent any currency risk, converge rapidly and track one another closely thereafter. 3.1.1. Estimation Augmented Dickey-Fuller tests cannot reject the presence of a unit root in all series nor the existence of a co-integrating relationship between them. An error correction model is therefore fit to the pre-liberalization period from 1973-1979

6 In the Japanese case, instrumental variables techniques yielded some unexpected signs and, as is generally the case, much larger standard errors (see Kennedy, 1992) so that few coefficients were significant and the recursive tests for parameter change were inconclusive. See also Baker et al. (1993) for a discussion of the drawbacks of instrumental variables techniques when instruments are only weakly correlated with the endogenous variable. The costs of using lagged values are possible misspecification, especially in the implicit assumption of backward expectations. Forward-looking proxies, such as the realized future inflation rate were experimented with but had little explanatory power. The specification with lags of inflation can be viewed as assuming adaptive expectations. 7 These restrictions were eased only slowly and with some reversals across the 1970s. In 1973, limits on nonresident stock and bond purchases were abolished and those on foreign direct investment in various industries were eased; in 1974, ceilings on loans from abroad were raised and the issue of bonds abroad by residents was made easier. s Foreigners were permitted to hold yen denominated bonds, gensaki (three-month repurchase agreements), yen denominated certificates of deposit, and 'impact loans' (foreign currency loans to Japanese non-bank residents) Frankel (1984).

W.F. Maloney / Journal of Development Economics 52 (1997) 139-168

146 • ,1¢*

~-Yen

(London)

.14

.1:1

~"i

.1

.08

.06

ii ~ , .I.% :~ _ '

~,~ i'-1

~%~,

.

~

t

...........

t--~<:

, L:,_/i

.04

• 02

Q

~

,,:

,,

,",

:5

- . 02

- . 94

i

I

1977

i

; I

1978

i

:'l

1915

i

I

1780

i

I

1~81

,

I

1982

i

I

1983

i

I

1984

i

I

1905

i

I

1986

i

I

1987

Fig. 1. T h r e e - m o n t h interest rates: J a p a n ( a n n u a l i z e d ) .

that represents a stable parsimonious specification which explains 53% of the variation in the gensaki rate. Since the point of the exercise is to verify that we can detect break without a reliable proxy for expectations of depreciation, only the results without the Euroyen rate are presented although those from the complete specification are even more convincing• As the first column of Table 1 shows, the critical explanatory variables enter significantly and are of the expected sign. The residual from the co-integration regression is not significant and is dropped. By every measure of parameter stability, the specification breaks down after June 1979 (Table 2). The second column of Table 1 shows virtually all domestic factors becoming insignificant and, despite the R 2 of 0.12 taken as a whole, the F-test suggests that previous explanatory variables now have no significant explanatory power. The 'forecast' Chow tests (Fig. 2), similar to the CUMSUMSQ test, compute scaled Chow tests for each of progressively longer forecast periods and confirm the sharp and permanent break by crossing the 5% significance line (at a scaled value of one) at the expected point. The vastly increased variance in scaled one-step residuals (Fig. 3) at the break is verified by the forecast X 2 test which rejects numerical constancy at any significance level. The recursive parameter estimates (Figs. 4-8) show significant break in virtually all coefficients. They show some ambiguous movements immediately after, that are absent when

W.F. Maloney/Journal of Development Economics 52 (1997) 139-168

147

Table 1 Three-month interest rates (ordinary least squares, first diff.) Variable

Japan 1976:6-1979:6

Chile Lag

1979:7-1986:12

Inflation Industrial activity Current account Domestic credit Constant Seasonal Seasonal Seasonal R2 o"

F-test DW RSS

AR

1-j

ARCH White heteroscedasticity J-B normality X 2(2)

Lag

1979:7-1982:6

2

- 0.6.09e-3 (-0.914) 1.28e-3

2

(1.64) - 1.10e-3

3

(0.0270) -3.51e-4

(3.91 ) - 1.10e-2

4

( - 4.45) - 3.42e-3

( - 1.81) - 8.69e-7

3

(-0.897) - 4.06e-7

(-2.53) - 9.91e-2

5

(-2.14) - 3.58e-2

(0.779) 0.00085 (1.88) - 5.84e-3 ( - 2.20) -5.71e-3 ( - 1.67) - 4.02e-3 ( - 1.60) 0.122 0.00612 F[7,71] 1.40 [0.22]

( - 2.74) - 0.241 ( - 1.51) 1.20 (2.78) -l.38 (-2.18) - 1.23 ( - 3.78) 0.731 0.429 F[8,22] 7.47 [0.00]

0.0111 (3.02) 0.112

( - 2.77) 1.65e-3 (1.72) - 2.79e-3 ( - 1.37) - 8.86e-3 (-4.98) - 2.58e-3 ( - 4.32) 0.525 0.00279 F[7,29] 8.76 [0.00] 1.74

B-P Q

1976:12-1979:6 0.670 (0.802) 0.302 (2.68) 5.46

Libor

0.000226 N = 37 1.51 (3) j=3 F[3,26] 0.367 [0.77] F[3,23] 0.471 [0.71] El11,17]

1.34 [0.291 0.435

3

1 4 10

1.89

2.18

- 0.247 (-0.689) 2 -2.22

( - 2.58) - 0.126 ( - 1.57) 1 2

-0.831 ( - 3.17)

11 0.642 0.229 F[5,18] 6.45 [0.00] 1.58

0.000266 N = 79 46.39 (5) j=5 F[5,66] 1.29 [0.28] F[5,61] 1.56 [0.19] El11,59]

4.07 N = 31 4.41 (3) j=3 F[3,19] 0.48 [0.79] F[3,16] 0.700 [0.57] F[13,81

0.939 N = 24 0.347 (2) j=2 F[2,16] 0.12 [0.89] F[2,14] 0.31 [0.73] F[9,8]

1.85 [0.07] 160

0.339 [0.96] 1.71

0.31 [0.95] 0.717

Note: t-statistics beneath coefficients in 0, p-values next to value in [].

Lag

148

W.F. Maloney/ Journal of Development Economics 52 (1997) 139-168

Table 2 Tests for parameter constancy Japan 1976:5-1979:6/1979:7-1982:6 Chile 1976:12-1979:6/1979:7-1980:12 1976:12-1979:6/1979:7-1982:1 1979:7-1981:6/1981:7-1982:12

Forecast X2 a

Forecast Chow b

362.6 (36) [0.00]

F(36,29) = 8.83 [0.00]

24.5 (18) [ > 0.10] 107.0 (31) [0.00] 134.6 (18) [0.00]

F(18,22) = 0.819 [0.66] F(31,22) = 1.58 [0.14] F(18,18) = 4.83 [0.00]

Note: p-values in [].a Compares within sample and forecast residual variances and is a test of for numerical constancy of the parameters,b The forecast Chow test whether the RSS is significantly different when forecast period is included. the Euroyen rate is included as they pick up the 1980 rise in world interest rates. Nonetheless, they confirm that the breakdown is due to capital account liberalization and not some other disturbance: the parameters on domestic factors, with the possible exception of the current account, all move toward zero. The Euroyen coefficient, included for illustrative purposes from the complete regression, shows dramatic and significant upward movement. In sum, studying the stability of Eq. (8) is shown to be an effective method of determining the efficacy of reductions in legal barriers to capital flows.

3.2. Chile

The liberalization of the Chilean capital account occurred in the context of a nearly complete transformation of the country to a model neoclassical small economy. Tariffs were reduced to a uniform level of 10% by 1979, government enterprises were largely privatized by 1975, and the budget deficit was reduced to zero by 1976. The financial system was privatized, and allocative quotas and interest rate ceilings were abolished in 1974-1975 (Fig. 9). The capital account liberalization occurred in stages with the principal measures facilitating commercial bank access to the world capital markets. 9 In September 1977, banks were permitted to import capital through Article 14 of the exchange law subject to three broad categories of restrictions. First, they were allowed to borrow only up to 20% of capital and reserves. This ceiling was gradually raised to 25%, higher for longer maturity loans, lO and then abolished completely in June 1979 in what is seen as the critical movement toward freeing capital inflows.

9 See Ffrench-Davis and Arellano (1981, in Spanish) and Mathieson (1983) and Corbo (1985) for a more extensive discussion. lOIn April 1978, the limit was raised to 45% for capital flows over three years. It was then raised to 60% in December 1978 and 70% in April 1979.

W.F. Maloney / Journal of Development Economics 52 (1997) 139-168

149

3.2

Z.8

i

'\,

i

2.4

"\.

2

1.5

1.2

,8 t

.4

f"-S

f II

~ ~"Jn

1979

,

n

2981I

,

i

2991

,

n

/983

,

I

1983

,

n

1984

,

h

291S

,

a

].!I 8~.

d l

1987

Fig. 2. Graphical break tests, Japan: forecast F-test. Second, restrictions on the maturity structure of incoming loans and the requirement of non-interest paying reserves held with the central bank worked in tandem to encourage longer term inflows. A minimum maturity of two years for

.IttZ .........

" ........

r

,I .IQ4

0

-.I~l

-, 012

-.116

-.12

fill

1981

1991

1913

I914

IIIG

Fig. 3. Graphical break tests, Japan: one-step residuals.

1987

150

W.F. Maloney/ Journal of Development Economics 52 (1997) 139-168

t

..If

F ,_.._...___._.____.....

!¢I .,

i~~

',.. ! I f V "~'~i('" . :{I

"4- i ~

I]

-

Fig. 4. Recursive parameter estimates, Japan: Euroyen flu|! estimation). ,mq

•-~

]\J, i

"..d •I l l

k

iL..

~ I

!

,A./\

,~

ID

"

~ ~:

-,

i

L'

. . . . .

-.Ill

Fig. 5. Recursive parameter estimates, Japan." inflation.

I

)

....

...................

..

~, Ii ~.,~ .~ /

J

"/"

.~.'

;~ ",. :,~,~iJ ,. ............... ~l

-.@lu

..=::

~

v.,,J!f'

-.,~

'"

i{

~-

...,.-,

.

.

l~e~l

.

.

ii

.

.

.

i$ii

.

.

.

iii

,

ii

iIi~

,

ii

,

,

111~i

,

*

i~ii11

Fig. 6. Rec~sive parameter estimates, Japan: current account.

W.F. Maloney/ Journal of Development Economics 52 (1997) 139-168

151

.I

.al

t',.J ~'

. . . . . . .

.

ii

'"

~ ..... .

.

.

.

.

.

"'"

.

.

.

.

.

.

.

............

.

.

.

.

.

.

" - ~

.

.

.

.

i~ /

V I~t

~m

1~11

t~i

lU~

1~4

19|s

lm

tier

Fig. 7. Recursive parameterestimates, Japan: industrial production.

Irii! i'i ;:

gp., J. . . . . . . . . . . . . . . . . . . . . .

.....

1199

l$1lll

11411

ltlli

191|

|$14

,..

|tN

15116

|~17

Fig. 8. Recursive parameterestimates, Japan: domestic credit.

inflows was established in 1976, thereby excluding short term capital flows by definition. Graduated reserve requirements of 25% for loans under three years, 15% for three-four years, and 10% for over four years were implemented in June 1979 with the highest category reduced to 15% in July 1980. As a result, the average maturity of loans from 1979-1982 was around five years, ll While this would seem to make tests of interest parity only valid in the very long run, this is not the case. Banks were permitted to on-lend their foreign borrowings to domestic residents in pesos, but by law, the loans were denominated in dollars. Thus the

II Edwards(1985b, p. 55).

W.F. Maloney / Journal of Development Economics 52 (1997) 139-168

152 1B

9 8

? 6

.'!

I"3

5

i i,:: :.:','}

4 3

Inflation

[I/ ',

I

=

, ,". ,, :...',.... ,,.l ,.l

: t

_

,~J i'~ ;,,.: . :.

i :]i/

~\

,~. .\ .,,.

:

, H U ~ f\

.". ,,,,.,~..

\

!

: '

•

. ~

! ~ :~ . .. ,'-~.,\/ .

2 1 .. . . . . . . . . . . . . . , _ .

8

-1

,...--.-""

' '""

...................................................................................................................................

......

,

I

1977

i

I

1978

i

l

1979

,

I

1988

................ ! .........................

i

I

19B1

i

I

,

1982

I

1983

i

I

1984

Fig. 9. Three-month interest rates: Chile (monthly).

bank clients, most with only 90-day loans, held the exchange risk and were therefore the relevant arbitragers. A final restriction set the rate at which dollar loans could be converted into pesos in an attempt to control liquidity. Prior to June 1979, the larger of 5% of capital and reserves or $2 million could be converted into pesos each month. This restriction was eliminated in June 1979 with the global lending limits, but reimposed in September at 5% or $1 million until April 1980 when the limit was again raised to US$2 million. The repeated fine tuning would seem to imply that the restriction was binding. That critical controls were in fact lifted by June 1979 or at the latest by April 1980 is suggested by the sheer volume of inflows which doubled in 1980 and virtually doubled again in 1981, reaching 10% of GDP. But, after an initial decline of real interest rates to 13% in 1980, the first months of 1981 saw a rapid rise to sustained levels of 40%. Given the vast flows of capital and the increasingly severe overvaluation of the peso as inflation persisted despite a pegged exchange rate, an appeal to uncovered interest parity and high expectations of depreciation as an explanation seems logical. However, from the model elaborated earlier, it is clear that the key factor determining whether domestic or foreign variables are determining interest rates is

W.F. Maloney / Journal of Development Economics 52 (1997) 139-168

153

l.m

,79

.9

,82

.72

.63

.54

.45

i'~....

• 36

;.2"

• 27

.,t8

J .@9

tN~' J i I

I

I

19?9

198e

Lg$L

il

I

I

1983

1992

Fig. 10. Graphical break tests, Chile: forecast F-test.

the degree of substitutability between foreign and domestic assets• Intermittent inflows, however vast, will simply leave the closed economy regime occasionally shocked by one-time increases in the money supply. Between inflows, however, a

1.2

,i

i:

I!I • -.2

-.4

t_l!

t

i!_!i !J

.....

.........~ rV! A ! ij ,i ....

-.6

:i -.8

,,,~

-I

-1.2 1979

i 191~

* 191) 1

i 19B2

Fig. 11. Graphicalbreak tests, Chile: one-stepresiduals.

i 1983

154

W.F. Maloney / Journal of Development Economics 52 (1997) 139-168

rise in the interest rate due to any d o m e s t i c factor w o u l d not be i m m e d i a t e l y offset by capital inflows and interest parity w o u l d not hold. T h e liquidity quotas m i g h t h a v e created just this effect as w o u l d the way m e d i u m and long t e r m capital arrived in Chile. Periodically, several large banks w o u l d c o o p e r a t e in a syndicated paper issue in the w o r l d financial centers. 12 In the periods b e t w e e n the arrival o f disbursements and their distribution through the various subsidiaries o f the national banks involved, there m i g h t be no response to e v e n substantial m o v e m e n t s in interest differentials such as m i g h t be e x p e c t e d in situations w h e r e short-term inflows w e r e permitted. 13 Despite the fact that the relevant market for establishing interest parity was the 3 0 - 9 0 - d a y market w h e r e the syndicated paper issues w e r e re-lent, the lumpiness o f the original inflows m a y h a v e left m a n y arbitragers rationed in their holdings o f dollar loans, leaving the peso d e n o m i n a t e d side o f the market d e t e r m i n e d by d o m e s t i c factors.

3.2.1. Estimation Since p r e l i m i n a r y testing was unable to reject the p r e s e n c e o f unit roots in m o s t variables, though it did reject co-integration, the m o d e l was estimated in first differences for the p e r i o d 1976:12 to 1979:6. Libor, though insignificant, is i n c l u d e d for w h a t e v e r i n f o r m a t i o n it can contribute on the foreign retum. 14 The

12 Ovando (1988). About 50% of capital inflows came through the Banco de Chile, where Ovando was manager. 13 Because demand for cheaper foreign credit continually exceeded supply, Chilean banks would provide a mix of peso and dollar loans with the weights dependent on the quality of the customer, sometimes adding a mark-up of close to 10-20% over Libor that accrued as pure monopoly rents. The remaining much-reduced spread between dollar and peso interest rates might represent expectations of devaluation on the part of the borrowers, but of a far smaller magnitude than is commonly believed (Ovando, 1988). The actual frequencies of syndicated issues is somewhat unclear. Ovando suggested every three to six months although others have argued that they were far more frequent. 14 It might be argued that a reasonable proxy for expectations of depreciation could be constructed to augment Libor. However, experience in the industrialized countries suggests that the pursuit of a trustworthy proxy is probably quixotic in general. Attempts to explain movements in the forward rates using surveys of exchange market speculators' expectations, presumably a good proxy for their true sentiments, perform notoriously poorly. Constructing proxies for those expectations by exploiting one or two variables that theory suggests should be important is one step further removed. Previous attempts using Bayesian updating techniques have yielded conflicting results for the Chilean case. Le Fort and Ross (1985), using diminishing reserves as the information variable, predict expectations of depreciation as mid-1982 approaches. Although not strictly comparable, Baxter (1985), keying on domestic credit which was contracting, argues that there was increasing confidence in the stabilization program across the same period. Knowledgeable observers disagree in their perceptions of the presence of expectations of depreciation. Bi-Minister of Economy and Finance Rolf Lfiders argued that there were, Minister of the Economy Sergio de la Cuadra argued that there were not (interviews, 1987). In the absence of a way to choose among such disparate series, and given the possibility of inducing bias if the choice is wrong, the present approach gains attractiveness. A similar argument can be made for not explicitly modeling country risk - unless we are sure of the manner in which agents form these expectations.

W.F. Maloney/Journal of Development Economics 52 (1997) 139-168

155

third column of Table 1 presents a good specification for 30-90-day deposit rates in the pre-liberalization period that explains 73% of the variance with the expected explanatory variables entering at least at the 10% level and of the predicted sign. In contrast to the striking results from the Japanese case, the forecast X 2 and graphical forecast F-tests (Fig. 10) for parameter stability and the one-step residuals (Fig. 11) strongly reject any change in how interest rates were determined after the capital account was opened in June of 1979, the raising of liquidation quotas in April of 1980, or the small readjustment of reserve requirements in July 1980. The recursive parameter estimates (Figs. 12-16) also reject break around these events although they do suggest some loss of significance of

. . . . .

. . . . . .

',,,J/ r)

¢

I 1 i !

ifll

INI

IHI

I~1

Ill

Fig. 12. Recursiveparameterestimates, Chile: Libor.

.(

\.-\ \

\ .11

j%

°-~%-'~

,11

L.-

I -Jl

,

a

|9~

i l~ee

I t~t

i

ngea

i

t~)

Fig. 13. Recursiveparameterestimates, Chile: inflation.

156

W.F. Maloney / Journal of Development Economics 52 (1997) 139-168 .n4

-.H, t./

-.oal

~° .............

/

c"

-.ill

/

,tkr,

-.m

~,, ,, ./

..nil, / -.gl~ /

-.•t I I - .lil* O41

|9~

15~e

tWt

19e2

t~03

Fig. 14. Recursive parameter estimates, Chile: current account.

,-.

e,e

0

\

\

. */'" ~

11.4 J

i . .i . . .

. .i . . . . i . .

;

;

Fig. 15. Recursive parameter estimates, Chile: industrial production. .Jl

.m

i

•

-.Jr

/ {

It"....r,.\

-.|e

-.l,l

r..,

..f~J r" ±

r~ r"

-.$

-.,Is -.,w

""°

,,g

,,--'

,,,1

,,d

,,,;

Fig. 16. Recursive parameter estimates, Chile: domestic credit.

W.F. Maloney/ Journal of Development Economics 52 (1997) 139-168

157

inflation, domestic credit and the current account later in the period, and a possible rise in significance (although not a significant rise) of Libor. Further, the one-step forecast residuals suggest the presence of outliers or break beginning at the end of 1980. To identify what these tendencies reflected, an attempt was made to fit a model to the later period. Column 4 in Table 1 presents a specification for the sub-period 1979:7 to 1981:6 that, although similar to the original, is consonant with the observed parameter movements and fits somewhat better. Inflation ceases to enter significantly, reflecting that nominal rates depart substantially from the inflation fate's slow decline to zero across the period, and domestic credit and the current account now act with less of a lag. This shift in specification could represent the market's adjustment to a low inflation regime or reaction to another dimension of the reform program. But the high explanatory power (0.64, or 0.63 without Libor) and significance of the parameter estimates strongly suggest that it does not reflect a reduction in the impact of domestic factors that a substantial liberalization of short term inflows would imply. Again, this might result from the lumpiness of the capital inflows: occasional large injections of liquidity might shift the overall level of interest rates, leading, theoretically, to a one-time prediction error in the formulation in differences, but not to a fundamental change in the coefficients of the interest rate equation. These findings are consistent with studies by Corbo (1985, 1986) and Edwards and Cox-Edwards (1991), who found little relation between the interest rate spread and capital inflows, and Sjaastad (1983) who found only a weak relationship between August 1979 and December 1980. 15 They are also consistent with work in the Kouri and Porter (1974) tradition of estimating the determinants of capital flows: Corbo and Matte (1984) found an offset coefficient of domestic credit on capital flows of - 0 . 3 2 from 1980:2 to 1982:1, suggesting that the central bank had some degree of control over monetary policy during the period; McNelis (1991), using a Kalman filter, found a smaller and almost negligible offset coefficient of - 0 . 0 7 ; and Laban and Larrain (1994) found an offset of - 0 . 0 9 across 1975:1 to 1992:4 that counterintuitively moves toward zero from 1980:2 to 1982:2. 16 Finally, the results are not in conflict with previous efforts to model interest rate behavior during the liberalization directly. Maloney (1989) applied locally weighted regression (Cleveland and Devlin, 1988) to a model similar in spirit to that presented above and also found no evidence of structural break using standard CUMSUM tests, although the technique did not allow the construction of confidence intervals that permit the explicit testing of break for individual

ts These cites reported in Edwards and Cox-Edwards (1991). 16 These cites reported in Laban and Larrain (1994).

158

W.F. Maloney/ Journal of Development Economics 52 (1997) 139-168

parameters. The same absence of tests for significance makes it difficult to evaluate the mixed results emerging from the work of McNelis and Schmidt-Hebbel (1993) employing a Kalman filter. The fall in the impact of domestic credit is dramatic, although of undetermined significance and perhaps late, occurring in the fourth quarter of 1980, several months after the last substantial liberalization measure and after a year of substantial inflows. Of more concern, the largest movement in the foreign interest rate parameter occurs long before the liberalization of the capital account, suggesting that it may not be statistically significant but rather represents the usual noise accompanying recursive updating techniques. 17 Weighing the mass of evidence, it appears that capital flows were relatively insensitive to interest rate differentials, that the interest rate was behaving as if the capital account was largely closed, and that the central bank had significant control, although perhaps decreasing with time, over the money base and interest rates. 3.2.2. Implications

This conclusion is important because it contributes to the construction of a view of the boom and collapse of the Chilean economy somewhat at odds with the more prevalent one that sees the current account deficit and indebtedness as the result of a lack of credibility in the new regime: the backward indexing of wages created the sharp appreciation of the peso, which in turn led to expectations of depreciation that led investors to demand very high rates of interest to hold Chilean vs. foreign assets. Consumers, seeing the damage to the real sector of the combination of appreciation and high real rates, speculated against the exchange rate by massive borrowing to purchase cheap imported goods. Although the relatively small number of observations advises caution, the 1979:7-1981:6 specification suggests that we do not need to resort to unobserved expectations of depreciation to explain the nominal, and implicitly, real interest rate movements we see. Fig. 17 presents the actual and fitted values within sample and then the out of sample forecasts of the new specification. The specification captures well the major movements in interest rates in sample. Figs. 1 8 - 2 0 present the same information although with one of the significant explanatory variables industrial production, domestic credit, or the current account - excluded from the specification to isolate its individual impact. Rises in industrial production appear

17The coefficient reaches its estimate in the constant parameter of 0.349 in 1977:4, suggesting that foreign rates were very influential from an early date and showed no substantial increase in influence with the reform measures. The two-quarter lag also seems long for financial variables given that it suggests that changes in the world rate, or the proxy for expectations of depreciation, affect domestic rates six months later. It is also possible that the constructed proxy for expectations of depreciation is capturing other factors whose influence remains constant across the period.

W.F. Maloney/ Journal of Development Economics 52 (1997) 139-168 Tittcd'-"

159

....

,4 .2

0

l i i! ,..:. f!i:i

..............., T . i a s ........~ " , ¢ - ~ - ~

-.2 -,4

" 71

7

i'il

7

i

-.8

iZi

i

lib II

i 1911

i

1981

1982

F i g . 17. P o s t - 1 9 7 9 : 7 C h i l e : actual a n d f i t t e d f o r e c a s t s - c o m p l e t e s p e c i f i c a t i o n .

associated with the brief jump in nominal rates of almost 0.6 of a percentage point (annualized ex-post real from 5 to 22%) from 1979:10 to 1980:1, and then the double peaked rise of 1.2 percentage points (annualized ex-post real, 0 to 44%) Fitted. = . . . . . .

.0

i

.6

,<

iil i/!

.4

.2

,

I!

!,, ,i/il;'i~

i,,'" '., ,

............ ,i,,.~ :~::i

~

i

il

{~ it ,t-

~t llll

.

, ,'t~;-;rr i 'i':'i ~ ............ .t , ,~, J , .,~

,z -~J

':i i

-,(

i:

i.k,'7"vt, E:iT.:,, :'., " . , ,,,~

,~,

{U

7

i

1901

i

1102

ltll

F i g . 18. P o s t - 1 9 7 9 : 7 C h i l e : actual a n d fitted f o r e c a s t s - n o industrial p r o d u c t i o n .

160

W.F. Maloney / Journal of Development Economics 52 (1997) 139-168 ritt~;

......

/i

.4

,Z

!

:"

j!:i

'

~, ~ii!

':'

i': il

0

-.2

,I'!,i L.~ I "+' l ,, "! '! /i

-,4

,:

!i! r'i

i++

-.G

-.8

i

i

i

i

198g

1891

1988

19113

Fig. 19. P o s t - 1 9 7 9 : 7 C h i l e : actual a n d f i t t e d forecasts - n o d o m e s t i c credit.

from 1980:10 to 1981:2. The last rise of 0.5 of a percentage point (annualized ex-post real, 18 to 37%) in 1981:5-7 is associated with a sharp contraction in domestic credit that began in the first quarter of that year.

1.2

Fitted~ ......

1

.8

i

.6

.4

.2

.....

,1

...................

i!i ].,j

0

-.2

-.4

i,r, , .

+.6

-.8

-1

I

I

I

198g

L911

191a

t

lgll

F i g . 20. P o s t - 1 9 7 9 : 7 C h i l e : actual a n d f i t t e d f o r e c a s t s - n o c u r r e n t account.

W.F. Maloney / Journal of Development Economics 52 (1997) 139-168

16l

All of these very substantial j u m p s in nominal and real interest rates appear to have their roots in the non-accommodating and later contractional behavior of the monetary base at a time when output was expanding rapidly. The monetary accounts show that this behavior is due almost entirely to movements in domestic credit and not in net foreign assets. These increase or remain constant to the end of 1981 while capital inflows remain stronger than previous years and would peak four months later. The important innovations are those such as the contraction of real domestic credit by over 20% in the first quarter of 1981. Minister of the Economy De la Cuadra explained that the government decided to use a substantial fiscal surplus to repay debt to the central bank rather than refund taxes or increase spending. It may be that the government expected compensating reserve movements as the M A B P would predict, but De la Cuadra also allowed that concern with the alarming current account deficit may also have influenced the policy. ~8 If the latter is the case, then the lack of credibility scenario could still be correct even if the high interest rates were not market driven by interest parity. The government, faced with speculation through the current account and increasing expectations of depreciation, may have driven up interest rates to maintain a steady supply of financing. The problem with this scenario is the timing of the first dramatic interest rate increases in 1980 and early 1981 described above. There is now substantial evidence that the roots of the current account deficit and the increase in absorption behind it during this period were not related to a lack of confidence in the reform program, but rather to the reigning ambience of 'triumphalism', where Chileans expected the economy to continue growing at rates of 10% a year and the Minister of the Economy was predicting a doubling of G D P in ten years 19 (Barandiaran, 1987; Schmidt-Hebbel, 1988; Edwards and Cox-Edwards, 1991). In the only formal test to date of the triumphalism hypothesis, Schmidt-Hebbel found in its support that 48% of the increase in traded goods consumption and 109% of the increase in non-traded good consumption were related to increases in myopically defined measures of wealth. 20 Conley and Maloney (1994) show that borrowing for consumption on both foreign and domestic durable goods increased at rates of 1 0 - 2 0 % in real terms per month

18 Interviews, 1987, 1990. i9 Minister of the Economy, Sergio de Castro, 1980 state of the economy speech. Also of interest was Labor Minister Jose Pifiera's 1980 speech to 3000 union leaders where he claimed that Chile "in ten years would be a developed nation..., where 70% of the population would have color T.V.s." 20 Measures of expected values of explanatory variables based on backward-looking behavior, and therefore exhibiting a myopia consistent with the euphoria hypothesis, statistically dominated those measures that were based on forward-looking behavior in his consumption function estimates. In addition, he found that the elasticity of intertemporal consumption substitution was not significantly different from unity, implying that real interest rates and future exchange rates did not affect-current consumption levels. Also see Schmidt-Hebbel and Serven (1994).

162

W.F. Maloney / Journal of Development Economics 52 (1997) 139-168

beginning very early in 1980, long before the trade regime appeared in peril, while the prices of real estate, the quintessential non-tradable, more than sextupled their level two years earlier in some areas of Santiago (see also Arellano, 1983). The behavior of both consumption and real estate suggests that the boom was not purely speculative, that people believed Chile would be far more prosperous than in the end it would turn out to be, and that the dramatic fall in savings and increase in borrowing resulted from an overconfidence in the reforms, rather than the reverse. Faced with this extraordinary increase in absorption, the Central Bank may well have embarked on a tight money policy to cool demand which in turn offered an attractive return to foreign capital. If, in fact, foreign credit was rationed as argued above, then evidence provided by Edwards (1986), Morand6 (1988) and Elbadawi and Soto (1994) that capital inflows may well have induced the appreciation of the peso gains new force since the inflows would have released constraints on domestic absorption. 21 Pulling these disparate findings together, we can construct a scenario that generates the entire boom-overvaluation, high interest rates, indebtedness, and resulting real sector collapse, from a consistent program undermined by unrealistic expectations of prosperity. If this alternate scenario is substantially true, the implication is that we cannot write off the Chilean collapse as something that can be avoided by governments who adopt credible policies and it puts a premium on understanding transition-related phenomena, such as consumption booms, that may undermine even the most solid of reform programs. There is already some evidence that a similar expectations-driven boom occurred in the Mexican miracle where we saw a dramatic fall in savings and tremendous demand for consumer borrowing in 1992 and 1993, long before the 1994 Christmas crisis. In late 1981, the traditional story may in some form come into play, as both interest rate specifications show signs of break. Visual inspection of the forecasts of the second regression also suggest a break in 1981:7 that is corroborated by the X 2 test and Chow tests. As Edwards and Cox-Edwards (1991) have argued, the end of the period of euphoria was marked by the collapse of the Crav Group in April 1981, followed by the failures of two major banks in November. Conley and Maloney (1994) show both housing prices and demand for consumer credit begin to fall in mid-1981 as well. This collapse in confidence may have led to short-term outflows which would have a more elastic response to expected return differentials. The evidence is mixed, however: capital inflows are still strong, and the

21 See also Copelman(1994) who argues that the increase in extended credit reduced the proportion of consumers that were liquidity constrained in the economy and helped to fuel the observed consumption booms.

W.F. Maloney / Journal of Development Economics 52 (1997) 139-168

163

forecasts of Figs. 1 7 - 2 0 suggest that domestic factors still appear to play a role. 22 Alternatively, other factors may have become important in delayed response to the reforms. As an example, in ailing sectors of the real economy, Arellano (1983) and Harberger (1985) have postulated a large demand for credit arising from illiquid and reform-damaged firms rolling over debts at high interest rates. The rampant self-lending practices of the banks had become evident and the superintendency of banks was implementing a portfolio risk-rating scheme to deal with the acknowledged large number of non-performing loans. In the absence of elastic capital flows, these and other related factors not explicitly modeled in the specification could have plausibly caused its breakdown in mid-1981.

4. Conclusion The model developed here suggests that a liberalization of short-term capital flows should lead to a reduced impact of domestic factors that determined interest rates previously and a greater impact of the implicit foreign return. The strong evidence emerging from tests for such parameter movements and structural break during the Japanese liberalization o f the mid-1970s for which forward rates are available, suggests that the approach is useful for measuring the impact of liberalization measures in countries where they are not. Chilean interest rates, however, are shown to behave as if determined domestically, despite vast inflows, until very late in the reform period. It is suggested that the very high real rates that emerged across this period were not the result of expectations of depreciation but of domestic factors, most particularly those associated with a contractional monetary policy and the euphoria arising from an excess of confidence in the economic reforms, rather than the reverse.

Acknowledgements M y thanks to Afonso Bevilaqua, Menzie Chinn, Barry Eichengreen, Albert Fishlow, Dale Henderson, Richard Meese, Ana Dolores Novaes, Perry Perone,

22 Capital inflows are strong throughout 1981 and substantially above the average level for 1980. Net foreign assets are at their peak, despite the large current account deficit, suggesting that speculators are not re-exporting inflows as short term outflows in significant quantities. Offsetting capital inflows appear to cause the first major prediction error occurs in 1981:7 when the specification predicts that the sharp contraction in domestic credit at the beginning of 1981 would have led to a continued increase in interest rates in 1981:7. The observed fall in real rates occurs concomitantly with one of the largest episodes of capital inflows across the period that prevented high powered money from contracting more than it did. However, domestic factors do not seem to have lost all of their impact after this one prediction error: the large spike in interest rates in 1981:10 coincides with the rise in industrial production; the dramatic fall in nominal rates in early 1982 seems driven largely by the large expansion of domestic credit at the end of 1981 that led to a real expansion of the monetary base of 26%.

164

W.F. Maloney / Journal of Development Economics 52 (1997) 139-168

Helen Popper, and Klaus Schmidt-Hebbel for helpful comments on earlier versions of this paper and to Hernando Vargas and Carlos Cavalcanti for inspired research assistance. I also gratefully acknowledge financial support from the Tinker Foundation and the University of Illinois Research Board.

Appendix A. Data See Tables 3 and 4.

Table 3 Japan 30-day gensaki (repurchase agreement) rates: monthly, Morgan Financial World 30-day Euroyen deposit rates: Morgan Financial World Output/income (ind.act.): monthly index of industrial activity, International Monetary Fund International Financial Statistics (IMF/IFS) Domestic credit (d.c.): IMF/IFS deflated by the PPI Cumulative current account balance (curr.acct.): IMF/IFS deflated by the PPI Expectations of inflation (intl.): several lags of realized monthly inflation, IMF/1FS.

Table 4 Chile 30-90 day deposit rates: monthly, Boletin Mensual, Banco Central de Chile (BCC) Output/income: monthly index of industrial activity, BCC Domestic credit: the Central Bank changed the composition of its domestic credit series numerous times and twice significantly from 1976-1982. The analysis uses the homogeneous series generated by Vial and Matin (1986) deflated by the consumer price index (CPI) Cumulative current account balance: the monthly trade balance was constructed from BCC export and import statistics standardized on the IMF BOP yearly aggregates. Transfers and non-interest services are the IMF yearly aggregates allocated equally across months. The monthly series on interest payments on Article 14 inflows was inflated so that yearly totals equaled the IMF total net interest payments. The sum yields a passable monthly measure of the current account surplus which is then deflated by the CPI, BCC, IMF Expectations of inflation: several lags of realized monthly inflation, BCC.

165

W.F. M a l o n e y / Journal o f Development Economics 52 (1997) 1 3 9 - 1 6 8

Appendix B. Unit root and co-integration tests See Tables 5 and 6.

Table 5 Dickey's t Variable

Japan

3 - m o n t h interest rate Foreign Inflation Industrial activity D o m e s t i c credit C u m u l a t i v e current a c c o u n t

Chile

Level

I st diff.

Level

1st diff.

-

- 10.97 - 13.07 - 4.97 - 6.50 -4.70 - 0.65

- 4.60 - 2.15 - 2.49 - 2.10 -0.42 - 0.26

- 8.11 - 10.60 - 7.90 - 7.30 - 11.40 - 2.93

1.93 2.20 2.02 1.36 1.60 0.60

Note: the null m o d e l is Yt = a + Y t - t + Et (et white noise). 5 % critical values (Dickey a n d Fuller, 1979) between - 2 . 8 9 a n d - 2 . 8 8 . The Japanese cumulative current account still appears non-stationary in first differences but because all other series appear 1(1), we do not difference further.

Table 6 Co-integrating regressions J a p a n ( R 2 = 0.90, D W = 0.81, s.e. = 0.0055) gen= 0.081 + 0.67 - 0.012 Euroyen ind.act, (2.35) (24.1) ( - 1.5) Chile ( R 2 = 0.596, D W = 0.71, s.e. = 1.265) dep= 2 4 . 7 2 - 2.15 - 3.93 + Libor ind.act, (4.02)

( - 2.5)

( - 3.02)

0.005 intl. (-0.79)

-0.00004 curr.acct,

0.167 intl.

+

(1.5)

-

0.0004 d.c.

( - 1.98) 0.001 curr.acct. (1.78)

(-0.52) -

0.015d.c. ( - 0.88)

Note: the t-statistic o f the model e t a e t 1 q'- et for the residuals of the co-integrating regression with six variables is - 5 . 8 8 for J a p a n and - 3 . 9 9 for Chile. T h o u g h Engle and Yoo (1987) tabulate critical values only up to five variables ( - 5 . 0 2 at 1% for 200 observations), they appear to increase at m o s t by 0.43 per additional variable. U s i n g this e x c e e d i n g l y crude measure, it is unlikely that co-integration can be rejected at the 5 % or even 1% level for J a p a n although it is rejected for Chile. See E n g l e a n d G r a n g e r (1987). =

W.F. Maloney / Journal of Development Economics 52 (1997) 139-168

166

Appendix C S e e T a b l e 7.

Table 7 Chile: selected macro indicators, 1973-1982 Year

Percentage of GDP

1973 1974 1975 1976 1977 1978 1979 1980 1981 1982

-3.7 0.4 - 8.4 1.6 9.4 9.6 9.9 9.7 5.4 - 15.1

Real int. rate

127.20 65.15 57.97 43.65 16.98 13.24 39.72 38.01

CPI 487.5 497.8 379.2 234.5 113.8 49.8 36.6 35.1 19.7 9.9

Sources: Corbo (1985), Banco Central de Chile.

Net K inflows

C.A. surplus

RER PT/PN

57.9 238.0 240.2 678.7 921.8 1809.0 2947.5 859.5

-294 -211 -491 148 - 551 - 1088 - 1189 - 1971 -4814 -2382

1.652 1.753 1.123 0.991 0.955 1.025 0.845 0.642

W.F. Maloney / Journal of Development Economics 52 (1997) 139-168

167

References Arellano, J.P., 1983, De la liberalizacion a la intervencion: E1 mercado de capitales in Chile, 1974-83, Coleccion Estudios CIEPLAN 11. Baker, R., J. Bound and D. Jaeger, 1993, The cure can be worse than the disease: A cautionary tale regarding instrumental variables, NBER Technical Paper 137, June. Barandiaran, E., 1987, La gran recesion de 1982, in: Del auge a la crisis de 1982 (Instituto lnteramericano de Mercados de Capital, Santiago, Chile). Baxter, M., 1985, The role of expectations in stabilization policy, Journal of Monetary Economics 15. Chinn, M.D. and J.A. Frankel, 1992, Financial and currency links in Asia and the Pacific, 1982-1992, Paper presented at Western Economics Association. Cleveland, W.S. and S.J. Devlin, 1988, Locally weighted regression: An approach to regression analysis by local fitting, Journal of the American Statistical Association, September. Conley, J.C. and W.F. Maloney, 1994, Optimal sequencing of credible reforms with uncertain outcomes, Journal of Development Economics. Copelman, M., 1994, The role of credit in post-stabilization consumption booms, Mimeo. (MIT). Corbo, V., 1985, Reforms and macroeconomic adjustments in Chile during 1974-84, World Development 12. Corbo, V., 1986, The use of the exchange rate for stabilization purposes: The case of Chile, Discussion Paper DRD132 (World Bank, Washington, D.C.). Corbo, V. and R. Matte, 1984, Capital flows and the role of monetary policy: The case of Chile, Working Paper 92 (Catholic University of Chile, Departrnent of Economics, Santiago). Dickey, D.A. and W.A. Fuller, 1979, Distribution of the estimators for autoregressive time series with a unit root, Journal of the American Statistical Association 74. Dooley, M.P. and D.J. Mathieson, 1992, Exchange rate policy, financial structure, and the transition from credit to monetary instruments, Paper presented at San Francisco Federal Reserve Board Conference on Exchange Rate Policy in Pacific Basin Countries, Mimeo. Dornbusch, R., 1985, Inflation, exchange rates and stabilization, Mimeo. (MIT) October. Edwards, S., 1985a, Money, the rate of devaluation, and interest rates in a semi-open economy, Journal of Money, Credit and Banking, February. Edwards, S., 1985b, Stabilization with liberalization: An evaluation of ten years of Chile's experiment with free market policies, 1973-1983, Economic Development and Cultural Change. Edwards, S., 1986, Monetarism in Chile, 1973-1983: Some economic puzzles, Economic Development and Cultural Change, April. Edwards, S. and A. Cox-Edwards, 1991, Monetarism and liberalism: The Chilean experiment (Ballinger, Cambridge, MA). Edwards, S. and M.S. Khan, 1985, Interest rate determination in developing countries: A conceptual framework, IMF Staff Papers, September. Eken, S., 1984, Integration of domestic and international financial markets: The Japanese experience, IMF Staff Papers. Elbadawi and Soto, 1994, Capital flows and long term real exchange rates in Chile, Policy and Research Working Paper 1306, World Bank. Engle, R.F. and C.W.J. Granger, 1987, Co-integration and error correction: Representation, estimation, and testing, Econometrica, March. Engle, R.F. and B.S. Yoo, 1987, Forecasting and testing in co-integrated systems, Journal of Econometrics 35. Ffrench-Davis, R. and J.P. Arellano, 1981, Apertura financiera externa: La experiencia Chilena en 1973-80, Coleccion Estudios Cieplan 5, Julio. Frankel, J., 1984, The yen/dollar agreement: Liberalizing Japanese capital markets (Institute for International Economics) December.

168

W.F. Maloney/ Journal of Development Economics 52 (1997) 139-168

Galvez, J.: and J. Tybout, 1985, Microeconomic adjustments in Chile during 1977-81: The importance of being a grupo, World Development 13. Haque, N.U.: and P. Montiel, 1990, Capital mobility in developing countries: Some empirical tests, IMF Working Paper 117, December. Harberger, A.C., 1985, Observations on the Chilean economy, 1973-1983, Economic Development and Cultural Change, April. Kennedy, P., 1992, A guide to econometrics (MIT Press, Cambridge). Kouri, P.J.K. and M.G. Porter, 1974, International capital flows and portfolio equilibrium, Journal of Political Economy, May/June. Laban, R: and F. Larrain, 1994, The Chilean experience with capital mobility, in: Bosworth, Dornbusch and Laban, eds., The Chilean economy: Policy lessons and challenges. Le Fort, G. and C. Ross, 1985, La devaluacion esperada: Una aproximacion Bayesiana, Serie de Investigacation 72 (Department de Economia, Universidad de Chile, Santiago). Maloney, W.F., 1989, Determinants of interest rates in a semi-open liberalizing economy, Chile 1976-82, Mimeo. (University of California at Berkeley). Mathieson, D.J., 1983, Estimating models of financial market behavior during periods of extensive structural reform: The experience of Chile, IMF Staff Papers, June. McNelis, P.D., 1991, Flujos de capital y political monetaria: Se compensan en Chile y Venezuela, in: F. Morandr, ed., Movimentos de capitales y crisis economica (llades/Georgetown University and IIMC, Santiago). McNelis, P.D. and K. Schmidt-Hebbel, 1993, Financial liberalization and adjustment: The cases of Chile and New Zealand, Journal of International Money and Finance. Mizala-Salces, A., 1985, Financial market liberalization: The case of Chile 1973-1982, Unpublished Ph.D. dissertation (U.C. Berkeley). Morandr, F., 1988, Domestic currency appreciation and foreign capital inflows, what comes first?: Chile, 1978-1982, Journal of International Money and Finance, December. Ovando, H., 1988, Personal communication. Reisen, H. and H. Ybches, 1993, Time-varying estimates on the openness of the capital account in Korea and Taiwan, Journal of Development Economics 41. Schmidt-Hebbel, K., 1988, Consumo e inversion in Chile (1974-1983): Una interpretacion 'real' del boom, in F. Morand6 and K. Schmidt-Hebbel, eds., Del auge a la crisis de 1982 (Instituto Interamericano de Mercados de Capital, Santiago). Schmidt-Hebbel, K. and L. Serven, 1994, Successful structural adjustment consumption and growth in Chile, Mimeo. (World Bank) August. Sjaastad, L., 1983, Failure of economic liberalism in the core of Latin America, World Economy, March. Vial, J. and M.B. Matin, 1986, Series monetarias Chilenas: 1960-1985, Estudios de Economia (Universidad de Chile) Abril.