examine the Ljung–Box (LB) statistic of the interest rate fluctuationsand the LB statistics of the squared interest rate fluctuations. Table 5lists the results. Except for the deposit interest rate of the Philippines,all other interest rates are auto-correlated in the level value or in thesquares. This tells us that interest rate volatility has the characteristicof heteroskedastic variance problem.4.3. The estimation results of the EC-EGARCH (1, 1)-M modelSince the traditional ECM model could not deal with theheteroskedastic variance problem, we use the EC-EGARCH modelunder the structure ofMTAR to analyze the interest rate adjustment inthe short-run. Basing on the test results in Table 4 and the basicstatistics in Table 5, we use the asymmetric EC-EGARCH (1, 1)-Mmodel to examine the deposit interest rate of Hong Kong, Malaysia,the Philippines, Singapore, and Taiwan, and the lending rate of HongKong, the Philippines, and Taiwan. For the lending rate of Singaporeand the deposit interest rate of the U.S., we use the symmetric ECEGARCH(1, 1)-M model. As for the countries that do not have thecointegration relation, we estimate with the EGARCH (1, 1)-M model.Table 6 lists the estimation results of the EC-EGARCH (1, 1)-Mmodel.We apply the following tests to conduct the diagnostic test: theserial correlation test with statistics Q12 (uh−½) and Q12 (u2h−1), theEngle and Ng (1993) sign bias test (SB), the negative size bias test(NSB), the positive size bias test (PSB), and the joint test. The testresults show that there is no asymmetric effect in the standardizedresiduals, which indicates that the model specification is appropriate.Before we discuss the estimation results in Table 6, we would liketo explain the parameters listed in Table 6 and the meanings of them.Parameter m indicates the effect of the changes of the money marketrates on the retail interest rates. Parameters s denotes the effect of thevolatility of the interest rates on the retail interest rates. Parameters η1and η2 represent the effect of the asymmetric error correction term onthe retail interest rates, and the test results of parameters η1 and η2could tell us whether the adjustment rigidity of the interest ratesfollows the collusive pricing arrangement hypothesis or the adversecustomer reaction hypothesis. Parameter η1 also represents the effectof the symmetric error-correction term on the retail interest rate.Parameter γ indicates the existence of the asymmetric adjustmenteffect of the conditional variance of the interest rates.In the following, we discuss the economic rationale of theparameters in Table 6 at 10% statistical significant level.• Parameter m: Except for the deposit interest rates of the Philippinesand Thailand, and the lending rate of Thailand, the effect represented
đang được dịch, vui lòng đợi..
