Dynamic Capital Mobility in Pacific Basin Developing Countries

The changing degree of capital mobility in several Pacific Basin countries that have pursued financial liberalization is estimated empirically. Tracing the impact of the liberalization process on the capital account, the paper also examines the implications for monetary policy operating in this changing economic environment. Empirical estimates support an overall finding of increased capital mobility in the region over the past decade. However, country experiences, with the exception of Singapore, have been more episodic than uniform, oscillating between periods of high and low financial openness. [JEL E43, F21, G15]

Abstract

The changing degree of capital mobility in several Pacific Basin countries that have pursued financial liberalization is estimated empirically. Tracing the impact of the liberalization process on the capital account, the paper also examines the implications for monetary policy operating in this changing economic environment. Empirical estimates support an overall finding of increased capital mobility in the region over the past decade. However, country experiences, with the exception of Singapore, have been more episodic than uniform, oscillating between periods of high and low financial openness. [JEL E43, F21, G15]

Estimation and Policy Implications

The changing degree of capital mobility in several Pacific Basin countries that have pursued financial liberalization is estimated empirically. Tracing the impact of the liberalization process on the capital account, the paper also examines the implications for monetary policy operating in this changing economic environment. Empirical estimates support an overall finding of increased capital mobility in the region over the past decade. However, country experiences, with the exception of Singapore, have been more episodic than uniform, oscillating between periods of high and low financial openness. [JEL E43, F21, G15]

Over the past decade developing countries in the Pacific Basin have experienced a continuing process of financial market liberalization and growing financial flows. Although individual countries have taken different measures, the region as a whole has moved toward the deregulation of domestic financial markets and the relaxation of restrictions on international capital flows. As liberalization continues in the region, the impact of these developments on the level of integration between local and world financial markets becomes an important consideration. In particular, measuring the changing degree of capital mobility—defined broadly in this context as the degree of linkage between domestic and foreign interest rates—is central to our understanding and assessment of financial liberalization and its consequences.

Most of the empirical work done in this area has examined capital mobility from a static perspective, estimating either an offset or an openness coefficient. For example, Edwards and Khan (1985) developed an analytical framework for interest rate determination in developing countries, wherein the prevailing interest rate represents a weighted average of open and closed economy rates that would have existed otherwise. Estimation of this weight indicates the relative importance of interest rate parity and domestic monetary factors in determining national interest rates, thus capturing the degree of financial openness. More accurately, however, this approach estimates the average degree of openness for a given country over a sample period. Hence, this line of research essentially ignores the changing degree of capital mobility and, consequently, the changing roles of stabilization and exchange rate policies over time.

To assess the dynamic impact of financial market liberalization on the mobility of capital and the role of policy, this paper applies autoregressive conditional heteroscedasticity (ARCH) estimation to interest rate differentials between selected Pacific Basin developing countries and Japan.1 The purpose is to determine the extent to which these rates have become increasingly interrelated. Specifically, a sustained narrowing in deviations of interest rate differentials from an appropriate measure of interest rate parity indicates an increased level of capital mobility. Using time-series estimation, this paper details the timing and magnitude of these changes resulting from financial liberalization.

There are many possible explanations for deviations from covered or uncovered interest rate parity. Various frictions, such as information and transactions costs, preclude strict equalization in rates of return across markets. Instead, differentials between exchange rate-adjusted interest rates may exist within some neutral band, indicating a range over which arbitrage opportunities are nonexistent. More generally, the width of the band should reflect all observable and unobservable costs involved in relocating arbitrage funds. For example, consider the case of capital controls and risk-averse investors. Beyond the “certainty” or direct costs associated with existing controls, capital restrictions exert “uncertainty costs” concerning future controls (see Otani and Tiwari (1990)). Capital controls, differential taxation, foreign exchange controls, and so on, introduce elements of country-specific risk that may drive a wedge between various market rates of return, as captured by a risk premium. Moreover, country risk also reduces the sensitivity of capital supply that would eliminate return differentials beyond the risk premium, leading to a still broader range of speculative inactivity.

Financial market liberalization, which involves changes in various adjustment costs and asset risk, directly affects the size of the neutral band within which return differentials may systematically deviate from interest rate parity. Specifically, liberalization measures that effectively increase financial openness will narrow the implicit band. The effectiveness of these measures ultimately depends on the reputation of the authorities and the credibility of their commitment to liberalization, as perceived by market participants. Moreover, these reputation and credibility effects are likely to be established only gradually over time, conditional on past credibility, reputation, and efforts to open up domestic financial markets.

As liberalization takes hold and the band narrows, increasingly confined deviations from interest rate parity will reflect the reduced impact of domestic monetary factors in determining domestic interest rates, corresponding to a heightened responsiveness of internationally mobile capital. Furthermore, the centrality of the band may also be affected by increasing financial market integration, as indicated by a diminishing risk premium stemming from greater asset substitutability.

The paper is organized as follows. Section I implements standard autoregressive moving average (ARMA) estimation to represent the time-series behavior of interest rate differentials in the Pacific Basin. Next, the estimates are revised via the ARCH model to capture any changing variability in interest rate parity deviations. These time-series results are translated into a dynamic representation of capital mobility through the simulation of impulse-response functions using country parameter estimates. Section II offers conclusions.

I. Time-Series Estimation

Monthly data series on domestic money market interest rates were obtained for the Republic of Korea, Singapore, Malaysia, and Thailand.2 For purposes of comparison, the three-month London interbank offer rate (LIBOR) on Japanese yen deposits was chosen to represent the world rate of interest.3 The choice of an offshore rate as the reference point was intended to limit the influence of liberalization measures taken in Japan in order to isolate the impact of measures taken domestically on interest rate convergence.

Series on interest rate differentials were then constructed by subtracting the Japan LIBOR from the domesticrate, or Φt = (ii*)t, expressed in annual percentage terms. Assuming that exchange rates vis-á-vis the yen follow a random walk in the sample,4 one can relate interest rate differentials to a suitably defined interest rate parity condition, as a means of calibrating deviations from the latter as a measure of capital mobility.

ARMA Estimation

To analyze deviations from interest rate parity, the time-series behavior of interest rate differentials is characterized, using an ARMA representation of Φt including a constant and a nonlinear time trend. Introducing a constant allows for the presence of a long-run liquidity and/or risk premium, implying a nonzero long-run value of Φ. An ARMA specification with a time trend permits one to separate transitory fluctuations around interest rate parity, which reflect the degree of capital mobility, from movements in the parity value itself, which reflect changes in asset substitutability.5

An ARMA (p, q) model for de-meaned and de-trended interest rate differentials—representing deviations from interest rate parity—can be written generally as

A(L)(φtμtrendt)=B(L)ϵt,(1)

where μ is the constant, and A(L) and B(L) are the AR and MA lag polynomials of orders p and q, respectively. ARMA estimation results are reported below in Table 1.

Table 1.

ARMA Estimation: Interest Rate Differentials in the Pacific Basin

article image
Note: Two asterisks indicate significance at the 1 percent level; t-statistics are in parentheses.

Exponential trendt = e(–0.05t).

As shown in Table 1, the appropriate time-series representation for interest rate differentials in Singapore, Korea, and Thailand was an AR(1) specification, whereas for Malaysia, the ARMA(1,1) model was more appropriate. So, for example, the time-series behavior of interest rate differentials in Malaysia is best represented by the univariate model, Φt = 0.9373Φt-1+ ϵt – 0.4381ϵt-1. The Schwarz information criterion (SIC) was used to select the particular order of the model—that is, the number of AR and MA lags—for each country.6

Constant terms were found statistically significant for Thailand and Korea only and were omitted elsewhere. Also, a time trend was found to be significant for Korean interest rate differentials. The large estimates (close to 1) for the AR(1) coefficient in Table 1 suggest weak mean-reverting behavior in the data, especially in the case of Malaysia and Thailand.7

Figure 1 presents plots of interest rate differentials, Φ, and residuals from ARMA estimation for each country. Across all four countries, large residuals generally cluster together, as do small residuals. This observation strongly suggests that an ARCH framework—which models an underlying pattern to the changing variance of residuals—is appropriate in order to improve the initial estimates and to further illustrate the systematic changes in capital mobility in the Pacific Basin.

Figure 1.
Figure 1.

Interest Rate Differential versus Japanese LIB OR, September 1978 to December 1990

Citation: IMF Staff Papers 1992, 003; 10.5089/9781451973174.024.A009

ARCH Estimation

The purpose of this subsection is to test formally for the presence of ARCH residuals and to revise the original estimates through ARCH estimation. Maximum likelihood estimation (MLE) is applied to the data series to correct for conditional heteroscedasticity. Under an assumption of normality and ARCH errors, the log-likelihood function for the tth observation can be written as follows:

lt=0.5loght0.5ϵt2/ht;ht=α0+α1ϵt12+,,+αnϵtn2,(2)

where ht is the conditional variance function for ϵt. Maximizing equation (2) with respect to the vector of parameters α, λ, μ and a time trend coefficient generates the ARCH-corrected estimates. As discussed in Engle (1982), the MLE approach generates consistent estimation of the variance-covariance matrix and asymptotically efficient parameter estimates.

Implementation of ARCH estimation to Φ is accomplished in two steps. First, using the residuals obtained from the previous ARMA estimation, a Lagrange multiplier or ARCH test8 is applied to check for the presence of ARCH residuals and to determine the appropriate linear specification for ht. Second, once a suitable ARCH model has been selected, MLE estimates are computed through iterative search, using ARMA estimates as initial starting values. Under sufficient regularity conditions, the iterative solution will yield optimal MLE parameter values (see Engle (1982)).

ARCH estimation results are presented for each country in Table 2 below. ARCH residuals were found significant for every country, with the exception of Malaysia.9

Table 2.

ARCH Estimation: Interest Rate Differentials in the Pacific Basin

article image
Note: One asterisk indicates significance at the 5 percent level, and two asterisks indicate significance at the 1 percent level;t-statistics are in parentheses.

ARCH test statistic is distributed x2(k), where k equals the number of lags in ht.

To interpret these results in terms of dynamic capital mobility, note that the conditional variance function ht represents systematic changes in the variability of disturbances which underlie deviations from interest rate parity. Departures from interest rate parity, in turn, reflect the degree of linkage between interest rates. Consequently, the changing degree of capital mobility can be characterized by the underlying pattern of conditional heteroscedasticity in disturbances to interest rate parity.

Figure 2 summarizes the estimates of dynamic capital mobility in the Pacific Basin by tracing this underlying variance pattern. Using residuals from ARCH estimation, sample values of ht taken as annual averages are calculated. In viewing Figure 2, note that movement down along the vertical axis toward the origin indicates increasing capital mobility corresponding to decreasing variability in interest rate parity deviations.

Figure 2.
Figure 2.

Regional Interest Rate Parity Deviations

(Conditional variance averages)

Citation: IMF Staff Papers 1992, 003; 10.5089/9781451973174.024.A009

As Figure 2 indicates, Singapore appears to have experienced a uniform decline (increase) in conditional variance (capital mobility) over the 1980s, leveling off around 1983. Korea’s experience was similar to Singapore’s, except for an apparent reversal in direction in the latter part of the decade.10 Thailand showed several episodes of increasing and decreasing financial openness in the 1980s without a clear tendency toward either direction. As for Malaysia, using initial ARMA estimates, a moving average of the sample variance suggests an overall dampening in shocks to interest rate parity with one notable exception. During specific periods in the mid-1980s, Malaysia witnessed extraordinary volatility in interest rate differentials which overwhelmed the detection of any possible underlying ARCH pattern in interest rate parity deviations.

Impulse-Response Simulations

Translating the variability of deviations from interest rate parity as a measure of capital mobility makes implications for policy less readily apparent than, say, an openness or offset coefficient. To better understand policy implications based on this calibration of capital mobility, impulse-response functions were constructed from the time-series estimates in Section I to simulate the dynamic effects of domestic monetary shocks on interest rate parity deviations. Integrating the changing variability of interest rate parity disturbances with the estimated time-series structure completes the dynamic representation of capital mobility.11

Specifically, ARCH parameter estimates and the conditional variance functions are used to track the effects of a one-standard-deviation shock to interest rate differentials in 1980 and in 1990.12 (See Figure 3,) Simulated trajectories are interpreted as departures from trend or long-run interest rate parity values where amplitude is measured in annual percentage points and persistence is measured in months.

Figure 3.
Figure 3.

Impulse-Response Simulations with ARCH Estimates

Citation: IMF Staff Papers 1992, 003; 10.5089/9781451973174.024.A009

II. Conclusion

Financial market liberalization in the Pacific Basin has had a significant impact on the region’s economic and financial environment over the past decade. Time-varying estimation of return differentials provides strong support for the notion that liberalization measures have raised the level of integration between domestic and international financial markets. Consequently, the degree of linkage between regional and world interest rates has increased. With a higher degree of capital mobility, departures from interest rate parity induce increasingly responsive market forces to narrow existing return differentials. And as rates of return converge across markets, the ability of independent monetary policy to affect domestic interest rates in the Pacific Basin becomes further limited.

REFERENCES

  • Edwards, Sebastian, and Moshin S. Khan, “Interest Rate Determination in Developing Countries: A Conceptual Framework,” Staff Papers, International Monetary Fund, Vol. 32 (September 1985), pp. 377403.

    • Search Google Scholar
    • Export Citation
  • Engle, Robert F., “Autoregressive Conditional Heteroscedasticity with Estimates of the Variance of United Kingdom Inflation,” Econometrica, Vol. 50 (July 1982), pp. 9871007.

    • Search Google Scholar
    • Export Citation
  • Levich, Richard M., “Empirical Studies of Exchange Rates: Price Behavior, Rate Determination, and Market Efficiency,” in Handbook of International Economics, Vol. 2, ed. by Ronald W. Jones and Peter B. Kenen (New York; Amsterdam: North-Holland 1985).

    • Search Google Scholar
    • Export Citation
  • Otani, Ichiro, and Siddharth Tiwari, “Capital Controls, Interest Rate Parity, and Exchange Rates: A Theoretical Approach,” International Economic Journal, Vol. 4 (Spring 1990), pp. 2544.

    • Search Google Scholar
    • Export Citation
  • Reisen, Helmut, and Hélène Yéches, “Time-Varying Estimates on the Openness of the Capital Account in Korea and Taiwan [Province of China],” OECD Development Centre (unpublished; Paris: Organization for Economic Cooperation and Development, 1991).

    • Search Google Scholar
    • Export Citation
*

Hamid Faruqee is a Lecturer at the Woodrow Wilson School of Public and International Affairs, Princeton University. Work on this paper began as part of his summer intership in the Asian Department. The author would like to thank David Goldsbrough, Owen Evans, Ichiro Otani, Roger Kronenberg, and Ranjit Teja for helpful discussions and comments.

1

The methodology employed in this paper essentially follows that of Engle (1982), here applied to interest rate differentials for the Republic of Korea, Malaysia, Singapore, and Thailand.

2

The disposition between countries and interest rates is as follows: for Malaysia, the overnight interbank rate; for Korea, the daily rate on call money; and for Singapore and Thailand, three-month interbank rates.

3

Although the sample countries float vis-à-vis the yen, one can think of the Japanese offshore rate as a proxy for the rate of return and financial conditions in the “world” market. Given fixed exchange rates at some margin, a small open economy must import its inflation and money growth rates from abroad, fixing its risk-adjusted domestic interest rate in the long run. Using the U.S. LIBOR instead as the international rate of return does not change the findings presented here.

4

An uncovered version of interest rate parity with perfect capital mobility and asset substitulability can be written as

φt=itit*Δst+ke,

where ΔSt+ke is the expected rate of depreciation in the spot exchange rate over maturity, with s being the log spot rate, and k, the length of maturity for i and i*. Assuming the spot exchange rate follows a random walk, the last term in the expression is set to zero, equating rational with static expectations. Initial exchange rate estimates strongly support a random walk hypothesis in the data. See also Levich (1985).

5

In the case of time-varying risk, this process is approximated with an exponential time trend, representing a smooth, downward convergence in the risk premium to its long-run value, paralleling a process of progressive financial liberalization and integration.

6

Final selections of appropriate time-series models, as reported in Table 1, were made through computing the SIC for numerous ARMA(p,q) specifications: (RSS + log(NOBS)*SEE2*K)/NOBS, where RSS = residual sum of squares, NOBS = number of observations, SEE = standard error of estimate, and K = number of repressors.

7

As the AR(1) coefficient goes to unity (unit root in A(L)), the model will exhibit nonstationarity or no mean reversion, and the effect of shocks becomes permanent. Although modeling Φ as an integrated process may not be rejected statistically as an alternative representation, theoretically this specification is very unappealing, for it suggests a perfectly closed capital account and no long-run equalization of risk-adjusted returns or significant nonstationarity in the risk premium inducing a sizable stochastic trend component in return differentials.

8

The ARCH test is conducted by regressing squared residuals on their lagged values and a constant. The particular order of the ARCH process (form of the ht function) is determined from the test regression found to be statistically significant. The ARCH test statistic is computed by NOBS*R2.

9

The ARCH statistic for Malaysia has a p-value of 0.170 (significance level). Also, MLE estimates for the first-order ARCH process suggest that Malaysian interest rate differentials are not covariance stationary (estimate on α > 1).

10

The findings for Korea are consistent with the results obtained by Reisen and Yèches (1991), who used a time-varying approach to financial openness following Edwards and Khan (1985), even though two different interest rates—a domestic curb market rate for Korea and the U.S. LIBOR—were used.

11

Changing capital mobility affects both the impulse and propagation of shocks to interest rate parity. Note that although time variation has been allowed for with respect to the former, the propagation mechanism has been kept constant over time (that is, same AR and MA coefficients) as an approximation. This approximation becomes less appropriate when capital mobility is near zero, suggesting that allowing for time variation in both dimensions—the amplitude and the persistence of interest rate parity deviations—may sharpen the results.

12

For Malaysia, the impulse-response functions were simulated using the initial ARMA estimates. Instead of using the nonstationary conditional variance function, an 18-month moving average of sample variances was calculated to determine the size of a typical one-standard-deviation shock.

IMF Staff papers: Volume 39 No. 3
Author: International Monetary Fund. Research Dept.