# Capital controls and deviations from proposed interest rate parity: Mexico 1982.

MARK M. SPIEGEL

An intervention analysis of the Mexican capital control policy of August 1982 shows the policy to have had a negative, but temporary, impact on the Mexican deviation from interest rate parity. Previous studies concerning the impact of capital control interventions have predicted positive effects of capital control policies. These stem from the effective tax capital controls place on foreign capital inflows. When capital controls are also designed to restrict domestic capital outflows, however, their net effect becomes an empirical, rather than theoretical, question.

I. INTRODUCTION

Dooley and Isard [1980] ascribe observed deviations from interest rate parity to a combination of political risks' and capital controls in place. Their empirical investigation into the effects of capital controls placed on non-resident earnings in Germany indicate a positive impact on the German deviation from interest rate parity. The Dooley and Isard evidence is cited by Melvin and Schlagenhauf [1985 a,b] in support of their estimation of a positive impact of Mexican capital controls on the deviation from Mexican interest rate parity in August 1982.

The Mexican case is distinct, however, from that studied by Dooley and Isard. While the German capital controls placed restrictions on foreign capital inflows, the Mexican capital controls primarily restricted domestic capital outflows. By making capital less scarce within the borders of Mexico, one would expect the capital controls to have a negative impact on the deviation from interest rate parity. Consequently, the Dooley and Isard framework seems to be inappropriate in its original form to predict the impact of the Mexican capital policy.

In this paper, I extend the Dooley and Isard model to one which can predict the effect of capital controls on both foreign capital inflows and domestic capital outflows. This model outlines the conditions which determine the net effect of the controls on the country's deviation from interest rate parity. I then conduct an intervention analysis, as in Box and Tiao [1975], to investigate the effect of the imposition of Mexican capital controls of August 1982. Results show a measurable negative impact of the capital control intervention on the deviation from interest rate parity in Mexico. However, this negative impact dies out in approximately six months. It is concluded that the Mexican policy was successful, for a limited time period, in lowering its deviation from interest rate parity via the use of capital controls policy.

II. THE MODEL

Following Dooley and Isard, let B denote the total stock of peso-denominated claims against the Mexican government. [B.sub.NR] represents the stock of peso-denominated claims of non-residents against the Mexican government, plus net non-resident claims on the Mexican private sector. [B.sub.M] denotes the stock of Mexican private sector claims on the Mexican government, plus net claims on non-residents held by the private sector. The market-clearing condition is:

[M.sub.R] [B.sub.NR] = B. (1)

The demand of the Mexican private sector for peso-denominated bonds depends on both the difference between the yield on pesos held in Mexico and the expected peso-equivalent yield on dollars, and the degree of severity of the capital controls restricting investment abroad. Let im denote the yield on claims against Mexican residents, while [[i.sub.US].sup*] and [[i.sub.M].sup.*] represent the Euro-dollar and Europeso yields respectively. Let E[carat over](P) represent the expected depreciation of the peso relative to the dollar. Assuming a linear specification, Mexican private sector demand can be specified as follows:

Substitute: [Mathematical Expression Omited]

The demand of non-residents for peso-denominated assets depends positively on the spread in the yield of peso-denominated claims on Mexican residents over the expected peso-equivalent yield on Eurodollars, as well as the yield spread of peso-denominated claims in Mexico over the Europeso rate. In addition, since capital controls impose a tax on foreign capital holders, the demand of non-residents will decline as the level of capital controls in place rises:

Substitute: [Mathematical Expression Omited] (3)

A positive exchange risk premium, r, is assumed to exist:

Substitute: [Mathematical Expression Omited] (4)

Combining equations (1), (2), (3), and (4), I obtain the deviation from interest rate parity as a function of country risk and capital controls:

Substitute: [Mathematical Expression Omited] (5)

where:

Substitute: [Mathematical Expression Omited] (5a)

Substitute: [Mathematical Expression Omited] (5b)

The sign of the capital control portion, (5b), becomes an empirical question when the effects on both foreign and domestic holders of Mexican assets are considered. Given that [a.sub.6] and [a.sub.2] are both positive, capital controls will discourage both domestic outflows and foreign inflows of capital. Capital control imposition will lower the deviation from interest rate parity if the domestic capital outflow effect exceeds the foreign capital inflow effect.

The coefficients in equation (5), however, need not be constant for the duration of the capital control policy. Domestic capital holders are likely to become more proficient at circumventing capital control regulations over time, leading to a declining influence of [a.sub.2]. On the other hand, the discouragement of foreign capital inflows is likely to hold for the duration of the policy. The net effect would therefore be expected to become less negative as the policy continues.

III. EMPIRICAL TESTING

Define the deviation from interest rate parity, [D.sub.t] as following:

Substitute: [Mathematical Expression Omited] (6) where [S.sub.t] represents the spot exchange rate and [F.sub.t,t+1] represents the forward exchange rate for pesos. The first term on the left can be interpreted as the interest rate differential between United States and Mexican securities, while the term on the right represents the forward exchange premium. When the United States and Mexican securities are issued under different legal regimes, a non-zero deviation from interest rate parity does not imply inefficiency in financial markets.

The deviation from interest rate parity will be affected by both political default risk in Mexico and the degree of capital controls present. I use the spread on long-term Mexican securities in the secondary bond market as an index of Mexican political risk. The bond market provides a cleaner measure than the secondary market for bank loans since the implicit insurance of depositors by their government hinders the usefulness of bank spread yields as a default risk measure. The bond spread series, given that the market is sufficiently deep," should provide a better default risk index. Using this index to capture default risk movements, the residual innovations attributable to capital controls should be observable.

Mexico introduced capital controls on 19 August 1982, subsequent to the reopening of the foreign exchange market, and tightened controls again on September 1. These controls were continued for the duration of the sample. I proxy the capital controls as beginning in September of 1982 in order to best match the beginning-of-month default risk proxy. The model is estimated using a general transfer function specification:

Substitute: [Mathematical Expression Omited] (7)

The hypothesis being tested maintains that the imposition of the capital controls had a negative impact on the deviation from interest rate parity by lowering the rate of domestic capital outflow. One might expect, however, that this impact would decay over time as methods for Circumventing the capital control regulations are discovered. On the other hand, the positive foreign capital inflow effect would be expected to be more permanent. The expected impact is therefore negative in the short run.

Estimation is done in terms of two alternative specifications. First, I estimate the capital control intervention via a "pulse variable," equal to zero for all months other than September 1982. This specification requires the effect of the imposition of capital controls to die out over time. Let [x.sub.5] and [x.sub.6] represent the estimated coefficients on the pulse capital control variable.

Estimation reveals that the univariate deviation from interest rate parity follows an AR(2) pattern. The bond spread index enters positively, with no dynamics, but with a three period lag. The capital control variable enters negatively, with a one-period lag, and a positive denominator indicating the predicted negative effect of the imposition of capital controls on the deviation from interest rate parity. Box-Pierce tests fail to reject the null hypothesis that the residuals are white noise at the five-percent confidence level.

The three-period lag on the bond spread index seems surprising since both the deviation from interest rate parity and the bond spread should represent current expectations of Mexican default risk. This entry indicates that if movements in and out of Mexican assets were costless, an uncovered arbitrage source would exist. Alternatively, the data may simply show ex-post that faster updating took place in the bond market during the specific period in question.

The negative impact of the capital controls' introduction in the pulse variable" specification is depicted in the impulse response function. The actual estimated impact of the capital controls on the deviation from interest rate parity is plotted both alone and in combination with the autoregressive parameters of the interest rate parity deviation series itself, referred to here as the cumulative impulse response function. Both functions show movements in interest rate parity deviations which can be attributed to the imposition of capital controls. The introduction of capital controls is estimated to have a large negative effect on the deviation from interest rate parity for approximately six months, after which the intervention effect converges to zero.

The pulse variable specification did not allow for permanent capital control effects. The addition of a "shift" variable to detect any permanent effect now allows both the permanent and temporary impacts of the capital controls to be estimated:

Substitute: [Mathematical Expression Omited] (8)

where C[C.sub.t].sup.p] and C[C.sub.t].sup.s] represent the pulse and shift effects of the capital controls respectively.

The estimation results with both the shift variable and the combination of shift and pulse variables. The estimates of the sign and significance of various parameters are robust to the introduction of the shift variable, C[C.sub.t].sup.s]. The highly significant denominator term on the shift variable reflects the lack of evidence concerning a permanent impact of the capital controls. Although the introduction of the shift parameter allows for the estimation of more complicated dynamics, the general pattern is quite similar to that of the simpler model. Box-Pierce tests also show the model to have reached white noise at a five-percent level of confidence.

Since the introduction of the shift parameter failed to indicate any measurable permanent effect, and since both specifications revealed similar estimated impulse response patterns, parsimony considerations would favor acceptance of the original specification. With either model, however, evidence shows a significant negative primary impact, lasting approximately six months, with no measurable permanent effect.

IV. CONCLUDING REMARKS

The Dooley and Isard [1980] model predicts the imposition of capital controls will have a positive effect on the deviation from interest rate parity. This result stems from the effective tax capital controls place on foreign investors. The predictions of the Dooley and Isard model are cited by Melvin and Schlagenhauf [1985 a,b] in support of their finding of a positive impact of capital controls on the deviation from interest rate parity in Mexico.

In this paper, I extend the Dooley and Isard model to allow the capital controls to affect both domestic and foreign capital outflows. The net effect on interest rate parity from these two offsetting impacts then becomes an empirical, rather than theoretical, question. The net effect will depend on the relative magnitudes of these two components.

An intervention analysis of the Mexican capital control policy's effect on the Mexican deviation from interest rate parity shows a significant negative impact which lasts approximately six months. This result is robust to a variety of specifications, including the introduction of a permanent shift parameter.

The implications for policy makers seem rather mixed. Capital controls policy appears to have been successful in the Mexican case in lowering the deviation from interest rate parity. This outcome was temporary, however, lasting only approximately six months. Afterwards, no measurable impact of the capital controls policy emerges from the data.

REFERENCES

Aliber, Robert Z. "The Interest Rate Parity Theorem: A Reinterpretation." Journal of Political Economy, November 1973, 1451-59.

Banco Nacional de Mexico. Review of the Current Economic Situation in Mexico, August/ September 1982.

Box, G. E. P. and G. C. Tiao. "Intervention Analysis with Applications to Economic and Environmental Problems." Journal of the American Statistical Association, March 1975, 70-79.

Dooley, Michael P. and Peter Isard. "Capital Controls, Political Risk and Deviations from Interest Rate Parity." Journal of Political Economy, April 1980, 370-84.

Edwards, Sebastian. "The Pricing of Bonds and Bank Loans in International Markets." European Economic Review, June 1986, 565-89.

Folkerts-Landau, David F. I. "The Changing Role of International Bank Lending in Development Finance." I.M.F. Staff Papers, March 1985.

Melvin, Michael T. and Don Schlagenhauf. "A Country Risk Index: Econometric Formulation and an Application to Mexico," Economic Inquiry, October 1985, 601-19.

Risk in International Lending: A Dynamic Factor Analysis Applied to France and Mexico." Journal of International Money and Finance, Supplement, March 1986, s31-s48.

Pick Publishing Corporation. Pick's Currency Yearbook, New York, 1984.

An intervention analysis of the Mexican capital control policy of August 1982 shows the policy to have had a negative, but temporary, impact on the Mexican deviation from interest rate parity. Previous studies concerning the impact of capital control interventions have predicted positive effects of capital control policies. These stem from the effective tax capital controls place on foreign capital inflows. When capital controls are also designed to restrict domestic capital outflows, however, their net effect becomes an empirical, rather than theoretical, question.

I. INTRODUCTION

Dooley and Isard [1980] ascribe observed deviations from interest rate parity to a combination of political risks' and capital controls in place. Their empirical investigation into the effects of capital controls placed on non-resident earnings in Germany indicate a positive impact on the German deviation from interest rate parity. The Dooley and Isard evidence is cited by Melvin and Schlagenhauf [1985 a,b] in support of their estimation of a positive impact of Mexican capital controls on the deviation from Mexican interest rate parity in August 1982.

The Mexican case is distinct, however, from that studied by Dooley and Isard. While the German capital controls placed restrictions on foreign capital inflows, the Mexican capital controls primarily restricted domestic capital outflows. By making capital less scarce within the borders of Mexico, one would expect the capital controls to have a negative impact on the deviation from interest rate parity. Consequently, the Dooley and Isard framework seems to be inappropriate in its original form to predict the impact of the Mexican capital policy.

In this paper, I extend the Dooley and Isard model to one which can predict the effect of capital controls on both foreign capital inflows and domestic capital outflows. This model outlines the conditions which determine the net effect of the controls on the country's deviation from interest rate parity. I then conduct an intervention analysis, as in Box and Tiao [1975], to investigate the effect of the imposition of Mexican capital controls of August 1982. Results show a measurable negative impact of the capital control intervention on the deviation from interest rate parity in Mexico. However, this negative impact dies out in approximately six months. It is concluded that the Mexican policy was successful, for a limited time period, in lowering its deviation from interest rate parity via the use of capital controls policy.

II. THE MODEL

Following Dooley and Isard, let B denote the total stock of peso-denominated claims against the Mexican government. [B.sub.NR] represents the stock of peso-denominated claims of non-residents against the Mexican government, plus net non-resident claims on the Mexican private sector. [B.sub.M] denotes the stock of Mexican private sector claims on the Mexican government, plus net claims on non-residents held by the private sector. The market-clearing condition is:

[M.sub.R] [B.sub.NR] = B. (1)

The demand of the Mexican private sector for peso-denominated bonds depends on both the difference between the yield on pesos held in Mexico and the expected peso-equivalent yield on dollars, and the degree of severity of the capital controls restricting investment abroad. Let im denote the yield on claims against Mexican residents, while [[i.sub.US].sup*] and [[i.sub.M].sup.*] represent the Euro-dollar and Europeso yields respectively. Let E[carat over](P) represent the expected depreciation of the peso relative to the dollar. Assuming a linear specification, Mexican private sector demand can be specified as follows:

Substitute: [Mathematical Expression Omited]

The demand of non-residents for peso-denominated assets depends positively on the spread in the yield of peso-denominated claims on Mexican residents over the expected peso-equivalent yield on Eurodollars, as well as the yield spread of peso-denominated claims in Mexico over the Europeso rate. In addition, since capital controls impose a tax on foreign capital holders, the demand of non-residents will decline as the level of capital controls in place rises:

Substitute: [Mathematical Expression Omited] (3)

A positive exchange risk premium, r, is assumed to exist:

Substitute: [Mathematical Expression Omited] (4)

Combining equations (1), (2), (3), and (4), I obtain the deviation from interest rate parity as a function of country risk and capital controls:

Substitute: [Mathematical Expression Omited] (5)

where:

Substitute: [Mathematical Expression Omited] (5a)

Substitute: [Mathematical Expression Omited] (5b)

The sign of the capital control portion, (5b), becomes an empirical question when the effects on both foreign and domestic holders of Mexican assets are considered. Given that [a.sub.6] and [a.sub.2] are both positive, capital controls will discourage both domestic outflows and foreign inflows of capital. Capital control imposition will lower the deviation from interest rate parity if the domestic capital outflow effect exceeds the foreign capital inflow effect.

The coefficients in equation (5), however, need not be constant for the duration of the capital control policy. Domestic capital holders are likely to become more proficient at circumventing capital control regulations over time, leading to a declining influence of [a.sub.2]. On the other hand, the discouragement of foreign capital inflows is likely to hold for the duration of the policy. The net effect would therefore be expected to become less negative as the policy continues.

III. EMPIRICAL TESTING

Define the deviation from interest rate parity, [D.sub.t] as following:

Substitute: [Mathematical Expression Omited] (6) where [S.sub.t] represents the spot exchange rate and [F.sub.t,t+1] represents the forward exchange rate for pesos. The first term on the left can be interpreted as the interest rate differential between United States and Mexican securities, while the term on the right represents the forward exchange premium. When the United States and Mexican securities are issued under different legal regimes, a non-zero deviation from interest rate parity does not imply inefficiency in financial markets.

The deviation from interest rate parity will be affected by both political default risk in Mexico and the degree of capital controls present. I use the spread on long-term Mexican securities in the secondary bond market as an index of Mexican political risk. The bond market provides a cleaner measure than the secondary market for bank loans since the implicit insurance of depositors by their government hinders the usefulness of bank spread yields as a default risk measure. The bond spread series, given that the market is sufficiently deep," should provide a better default risk index. Using this index to capture default risk movements, the residual innovations attributable to capital controls should be observable.

Mexico introduced capital controls on 19 August 1982, subsequent to the reopening of the foreign exchange market, and tightened controls again on September 1. These controls were continued for the duration of the sample. I proxy the capital controls as beginning in September of 1982 in order to best match the beginning-of-month default risk proxy. The model is estimated using a general transfer function specification:

Substitute: [Mathematical Expression Omited] (7)

The hypothesis being tested maintains that the imposition of the capital controls had a negative impact on the deviation from interest rate parity by lowering the rate of domestic capital outflow. One might expect, however, that this impact would decay over time as methods for Circumventing the capital control regulations are discovered. On the other hand, the positive foreign capital inflow effect would be expected to be more permanent. The expected impact is therefore negative in the short run.

Estimation is done in terms of two alternative specifications. First, I estimate the capital control intervention via a "pulse variable," equal to zero for all months other than September 1982. This specification requires the effect of the imposition of capital controls to die out over time. Let [x.sub.5] and [x.sub.6] represent the estimated coefficients on the pulse capital control variable.

Estimation reveals that the univariate deviation from interest rate parity follows an AR(2) pattern. The bond spread index enters positively, with no dynamics, but with a three period lag. The capital control variable enters negatively, with a one-period lag, and a positive denominator indicating the predicted negative effect of the imposition of capital controls on the deviation from interest rate parity. Box-Pierce tests fail to reject the null hypothesis that the residuals are white noise at the five-percent confidence level.

The three-period lag on the bond spread index seems surprising since both the deviation from interest rate parity and the bond spread should represent current expectations of Mexican default risk. This entry indicates that if movements in and out of Mexican assets were costless, an uncovered arbitrage source would exist. Alternatively, the data may simply show ex-post that faster updating took place in the bond market during the specific period in question.

The negative impact of the capital controls' introduction in the pulse variable" specification is depicted in the impulse response function. The actual estimated impact of the capital controls on the deviation from interest rate parity is plotted both alone and in combination with the autoregressive parameters of the interest rate parity deviation series itself, referred to here as the cumulative impulse response function. Both functions show movements in interest rate parity deviations which can be attributed to the imposition of capital controls. The introduction of capital controls is estimated to have a large negative effect on the deviation from interest rate parity for approximately six months, after which the intervention effect converges to zero.

The pulse variable specification did not allow for permanent capital control effects. The addition of a "shift" variable to detect any permanent effect now allows both the permanent and temporary impacts of the capital controls to be estimated:

Substitute: [Mathematical Expression Omited] (8)

where C[C.sub.t].sup.p] and C[C.sub.t].sup.s] represent the pulse and shift effects of the capital controls respectively.

The estimation results with both the shift variable and the combination of shift and pulse variables. The estimates of the sign and significance of various parameters are robust to the introduction of the shift variable, C[C.sub.t].sup.s]. The highly significant denominator term on the shift variable reflects the lack of evidence concerning a permanent impact of the capital controls. Although the introduction of the shift parameter allows for the estimation of more complicated dynamics, the general pattern is quite similar to that of the simpler model. Box-Pierce tests also show the model to have reached white noise at a five-percent level of confidence.

Since the introduction of the shift parameter failed to indicate any measurable permanent effect, and since both specifications revealed similar estimated impulse response patterns, parsimony considerations would favor acceptance of the original specification. With either model, however, evidence shows a significant negative primary impact, lasting approximately six months, with no measurable permanent effect.

IV. CONCLUDING REMARKS

The Dooley and Isard [1980] model predicts the imposition of capital controls will have a positive effect on the deviation from interest rate parity. This result stems from the effective tax capital controls place on foreign investors. The predictions of the Dooley and Isard model are cited by Melvin and Schlagenhauf [1985 a,b] in support of their finding of a positive impact of capital controls on the deviation from interest rate parity in Mexico.

In this paper, I extend the Dooley and Isard model to allow the capital controls to affect both domestic and foreign capital outflows. The net effect on interest rate parity from these two offsetting impacts then becomes an empirical, rather than theoretical, question. The net effect will depend on the relative magnitudes of these two components.

An intervention analysis of the Mexican capital control policy's effect on the Mexican deviation from interest rate parity shows a significant negative impact which lasts approximately six months. This result is robust to a variety of specifications, including the introduction of a permanent shift parameter.

The implications for policy makers seem rather mixed. Capital controls policy appears to have been successful in the Mexican case in lowering the deviation from interest rate parity. This outcome was temporary, however, lasting only approximately six months. Afterwards, no measurable impact of the capital controls policy emerges from the data.

DATA APPENDIX All data included in this study are monthly, from April 1980, to March 1984. U.S. Interest Rate Data Source: CITIBASE data tape Series: 1-month certificate of deposit rate Mexican Interest Rate Data Source: Subdireccion de Investigacion Economica, Indicadores Economicos Series: 1-month certificate of deposit rate Forward Discount Rate Data Source: Data Resources Inc. Series: 1-month forward discount Mexican Capital Controls Data Source: Pick's Currency Yearbook 1984 Series: Pulse Variable: A zero-one dummy equal to one only on September 1982. Shift Variable: A zero-one dummy equal to one beginning September 1982. U.S.-Mexico Spot Exchange Rate Data Source: International Financial Statistics Series: Monthly market rates, series rf. U.S.-Mexico Bond Spread Data Data Source: Folkerts-Landau 1982-1984, International Herald Tribune, 1980-1981. Series: World Bank, 10.25, June 1987; Mexico, 8.25, March 1987 All bonds are offshore, dollar-denominated. Bond data represent first Monday of corresponding month.

REFERENCES

Aliber, Robert Z. "The Interest Rate Parity Theorem: A Reinterpretation." Journal of Political Economy, November 1973, 1451-59.

Banco Nacional de Mexico. Review of the Current Economic Situation in Mexico, August/ September 1982.

Box, G. E. P. and G. C. Tiao. "Intervention Analysis with Applications to Economic and Environmental Problems." Journal of the American Statistical Association, March 1975, 70-79.

Dooley, Michael P. and Peter Isard. "Capital Controls, Political Risk and Deviations from Interest Rate Parity." Journal of Political Economy, April 1980, 370-84.

Edwards, Sebastian. "The Pricing of Bonds and Bank Loans in International Markets." European Economic Review, June 1986, 565-89.

Folkerts-Landau, David F. I. "The Changing Role of International Bank Lending in Development Finance." I.M.F. Staff Papers, March 1985.

Melvin, Michael T. and Don Schlagenhauf. "A Country Risk Index: Econometric Formulation and an Application to Mexico," Economic Inquiry, October 1985, 601-19.

Risk in International Lending: A Dynamic Factor Analysis Applied to France and Mexico." Journal of International Money and Finance, Supplement, March 1986, s31-s48.

Pick Publishing Corporation. Pick's Currency Yearbook, New York, 1984.

Printer friendly Cite/link Email Feedback | |

Author: | Spiegel, Mark M. |
---|---|

Publication: | Economic Inquiry |

Date: | Apr 1, 1990 |

Words: | 2334 |

Previous Article: | Why economic reforms fail in the Soviet system - a property rights-based approach. |

Next Article: | 1979-1982: rules or discretion in monetary policy? |

Topics: |