Appendix I Capital Mobility and Monetary Policy
- David Robinson, Ranjit Teja, Yangho Byeon, and Wanda Tseng
- Published Date:
- September 1991
During both 1989 and 1990, the growth of the monetary aggregates in Thailand was driven almost entirely by a rapid increase in private capital inflows, which the Bank of Thailand has found increasingly difficult to sterilize. This difficulty has, in part, reflected the limitations of the monetary instruments available to it, discussed in more depth in Section IV, But the scope for monetary policy has also been constrained by the authorities’ policy of fixing the value of the baht to a basket of foreign currencies, which has in practice been reflected in a close link between the baht and the U.S. dollar.
The degree to which the fixed exchange rate is, in fact, a constraint on monetary policy depends critically on the openness of the capital account. In general, the capital account in Thailand has been quite open in recent years (especially for inflows), although there have been important limitations in two areas. First, for prudential reasons the net open foreign exchange position of commercial banks is constrained, thus limiting the degree to which banks can adjust their foreign exchange positions to offset changes in monetary policy.39 Second, residents have until recently been unable to hold foreign-exchange-denominated deposits in domestic banks; at the same time, there have been controls on capital outflows. This, in turn, is likely to have limited the scope for arbitrage by residents in response to interest rate differentials on domestic and foreign assets. The degree of capital mobility may also be affected by a variety of other factors, in particular the degree of integration between foreign and domestic markets and between domestic markets themselves, even when no formal legal constraint exists: for exam-pie, while the largest corporations can reportedly borrow abroad quite easily, this is less likely to be so for other firms.40
This appendix assesses the openness of the capital account in Thailand, using a method first put forward by Edwards and Khan (1985), and an extension suggested by Haque and Montiel (1990), and thereby draws some conclusions on the likely effectiveness of monetary policy.
The Edwards and Khan approach attempts to measure the openness of the capital account directly by viewing the observed domestic interest rate at time t, it, as a weighted average of the interest rate that would obtain if the economy was completely open (which would be the foreign rate, adjusted for expected exchange rate changes, risk premiums, and transactions costs) and the interest rate that would exist if the economy was completely closed (which would equal expected inflation plus the real interest rate, the latter varying with domestic monetary conditions). Thus, they derive the following equation:
The parameter Ψ measures the degree of openness of the economy in the long term: if Ψ = 1, the economy is Hilly open. In practice, even in a fully open economy, domestic interest rates may only adjust to foreign interest rates with a lag, reflecting, for example, transactions costs and information lags. To allow for this, the model is extended so that the open economy-interest rate adjusts to foreign interest rates with a speed of adjustment Θ (if Θ = 1, then adjustment is instantaneous), allowing a distinction between the openness coefficient in the short run (Θ) and in the long term (Ψ). The degree of monetary disequilibrium, which determines the real interest rate in a closed economy, can be proxied in several ways. One, embodied in equation (2) below, is to define it as the ratio between real money supply m,t and money demand, the latter a function of real GDP (yt), the long-run real interest rate (assumed constant) and expected inflation (πe)A second approach, based on Makin (1982), is to postulate that nominal interest rates are affected in the short run by monetary “surprises.” Incorporating these additions, equation (1) can be written
where the dt are parameters. A detailed derivation of this equation, and expressions for the dt in terms of the underlying parameters of the system, can be found in Edwards and Khan, It is enough to note here that the openness coefficient Ψ is the sum of the parameters d1and d5 while the speed of adjustment Θ can be derived by dividing d1 by (d1 + d5).
This basic approach was adapted by Haque and Montiel for use when a domestic interest rate series either did not exist, or was not fully representative (because, for example, of interest rate controls or other forms of financial repression). Their approach is based on the assumption that the (unobservable) true domestic interest rate is an argument of a conventionally defined money demand function. This money demand function could then be used to substitute for the domestic interest rate in equation (1). The closed economy domestic interest rate could be derived in a similar fashion, assuming that the closed economy money supply mt equaled the observed money supply less the inflow of private capital during the period. On this basis, the following reduced form can be derived
Once again, the detailed derivation can be found in Haque and Montiel: it is sufficient here to note that the openness coefficient Ψ equals (1 – a2).
For Thailand, the key interest rates include the interbank and repurchase rates (which in practice move closely together and are the proximate target of monetary policy) and commercial bank deposit and lending rates. However, consistent data series are available only for the first two; lending and deposit rates are not only difficult to define consistently over time, but have also been distorted by the effect of lending and deposit rate ceilings,41 Therefore, the Edwards and Khan approach was applied to the interbank interest rate series, to assess how far the Bank of Thailand is able to influence the liquidity conditions in the banking system: the Haque and Montiel approach was used to test how far the (unobservable) true marginal cost of funds in the economy is influenced by foreign interest rates.
The two models were estimated using quarterly data for 1978–90; a quarterly GDP series was provided by the authorities through 1988 and extended through 1990 by applying the seasonal factors in the past to the annual data. Given the stability of the nominal exchange rate against the U.S. dollar over most of the period, the expected exchange rate change (against the U.S. dollar) was set at zero42: expected inflation was proxied by actual inflation. Changes in these specifications produced no substantive changes in the results.
The results of the estimations of the Edwards and Khan approach are set out in Table 17, using both equation (2) above and a formulation incorporating the effect of monetary surprises.43 Dummy variables were included for the first quarter of 1985 (in which the interbank rate rose very sharply partly as a result of the depreciation of the baht in November 1984) and for the third quarter of 1990 to take account of the Middle East crisis. Both specifications fit reasonably well, and a Chow test cannot reject the hypothesis of parameter stability, in each case, the lagged domestic interest rate and the foreign interest rate are correctly signed and highly significant; moreover, their values are almost the same in each specification. The long-run openness coefficient Ψ is very close to unity in each case (Table 18), which suggests that the limits on the net foreign position of commercial banks have not been a binding constraint over the period. However, adjustment is not instantaneous, with only about one half of the difference between domestic and foreign interest rates eliminated in each quarter, suggesting that information tags and other sources of friction in the system are important in the short term.
|1978: Q2 1990: Q4||Lagged Interest Rate||Foreign Interest Rate||Real GDP||Lagged Money Balances||Monetary Surprise||Expected Inflation||DW|
|Ordinary Least Squares|
|1. Standard specification||0.46||0.55||0.98||0.24||—||–0.09||0.76||2.0|
|2. Monetary surprise||0.43||0.54||—||—||–0.32||–0.16||0.78||1.83|
|Equation||Long Run||Short Run|
|Edwards and Khan (standard specification)||1.01||0.55|
|Edwards and Khan (monetary surprise)||0.97||0.54|
|Haque and Montiel||0.59||0.45|
The coefficients of the variables describing domestic monetary disequilibria arc generally less well specified, which is perhaps unsurprising given the high value of the openness coefficient noted above. In the standard specification, the coefficients on real GDP, lagged money balances, and inflation are all insignificant, and the coefficient on lagged money balances has the wrong sign. The monetary surprise approach gives generally better results: the coefficient on the monetary surprise is correctly signed and significant at the 5 percent level, although the coefficient on inflation is both insignificant and incorrectly signed.44
The results of the Haque and Montiel approach are set out in Table 19. The closed-economy money supply was defined as the money supply excluding all private capital inflows, except direct investment, which is unlikely to be as responsive to changes in interest differentials as other forms of capital inflows. Since a number of variables in the Haque and Montiel equation are nonstationary, the equation was estimated using error correction techniques. To this end, equation (3) was first estimated using ordinary least-squares techniques. The residual series from that estimation passed a standard Dickey-Fuller test for rejection of nonstationarity, and was then used (with a lag of one period) as an argument of an error correction model. The ordinary least-squares estimate can be interpreted as providing the long-run relationship, while the error correction mechanism shows the short-term dynamics. Thus, the long-run openness coefficient can be derived from the ordinary least-squares estimate, and the short-run openness coefficient from the error correction specification.
|Ordinary Least Squares|
|where rest is the residual from the ordinary least-squares regression, and D is a difference operator.|
The estimations in Table 19 fit well, with all variables correctly signed, and all but the foreign interest rate in the ordinary least-squares estimation significant at the 1 percent level: once again, a Chow test cannot reject the hypothesis of parameter stability. The long-run openness coefficient is approximately 0.6. while the short-run coefficient is about 0.5. The short-run openness coefficient is, surprisingly, only slightly lower than that derived using the Edwards and Khan approach, suggesting that those nonbanks that can adjust their portfolios in response to changes in foreign interest rates do so relatively quickly. In the long run, however, the openness coefficient derived using the Haque and Montiel approach is lower than that derived from the Edwards and Khan approach, which confirms the prior expectation that the interbank money market is the most open financial market in Thailand.
On the basis of the above, the following main conclusions can be drawn. First, the results confirm that the Thai capital account is quite open, especially as regards the interbank market. At the same time, both approaches used suggest that the differential between domestic and foreign rates is not eliminated immediately, so that considerable scope exists for an independent monetary policy in the short term. The degree to which this conclusion will hold in the future is, however, less clear. Capital account transactions were further liberalized in April 1991, including the introduction of foreign exchange accounts for corporations and individuals, and the authorities have stated that consideration will be given to eliminating the remaining restrictions on outflows. Moreover, as financial liberalization proceeds, domestic and foreign financial markets will become increasingly closely linked—an important objective of the liberalization program. In the long run, therefore, maintaining a relatively fixed exchange rate is likely to constrain increasingly the conduct of monetary policy.