### Discussion Paper

## Abstract

Previous studies found that Islamic stock market index in Malaysia (KLSI), does not react, or react negatively to interest rate, although one of the main criteria of Islamic finance is to avoid business and activities that yield interest because of its prohibition in Islamic laws. On the other hand, studies of Islamic stock market index in the US (DJIMI) found that there is no impact of interest rate on DJIMI. These two stock market indices have different screening criteria and different composite of securities. This study aims at investigating the monetary policy variables impact, the effect of interest rate, and the use of stock market indices as a hedge against inflation. It also examines the volatilities of monetary variables, interest rates, and inflation rate on two Islamic stock market indices. Using time series analysis such as GARCH the results are as follows. It is found that in the variance univariate models of the conventional indices that M1, M3, inflation rate, and real growth in GDP are significant in influencing KLCI volatility, while M2, M3, inflation rate and interest rate affected DJINA volatility. On the other hand, in the Islamic indices, KLSI and DJIMI variance is influenced by M2, M3, and inflation rate. In addition, in the multivariate model, DJIMI is influenced by the interest rate and the inflation rate in the mean and variance equations. In contrast, KLSI is influenced commonly in the mean and variance equations by M3, and the inflation rate.

## Comments and Questions

I read your paper and I think it deals with an interesting issue. However, some empirical results are a bit surprising. First of all, I have not seen in your paper a test for heteroskedasticity supporting the need for GARCH terms (you check ex-post homoskedasticity but this might be present ...[more]

... from the beginning, why this test is not included in Tables 4.1 and 4.2?). At the monthly level empirical evidences of heteroskedasticity might be weak. Furthermore, if your focus is on the mean relation, and given the limited size of your sample (about 100 observations) why not simply using a robust esimator for the parameter variances?

My doubts on the need of GARCH are also supported by the negative coefficients you report in the tables, without checking the log-moment condition; I'm particularly worried about the negative intercepts.

When you include M1, M2 and M3 in the model, did you check for collinearity problems?

Dear Sir

the main goal of the paper is to reply some research which indicated that islamic investments are not affected by interest rates. so i did the analysis in both the mean level and the volatility level.

i did not get what is the problem with a negative ...[more]

... intercepts.

i have checked for collinearity problem and because of its existence you can see that in one time i included M1 and M2 but not M3.

thank you for your comments.

You are right, you avoid collinearity by excluding one of the money stock variables. However, two questions are still open:

- is there really a need for GARCH? You should convince the reader that the mean residuals (without GARCH) show GARCH effects. I thus suggest to report LM-ARCH tests for ...[more]

... the mean model.

- If you compute the unconditional variance of you variance model (assuming it is really needed), is this quantity positive? (this might not be the case is the exogenous variables in the variance equation have a small unconditional mean, given that the intercept of the GARCH is negative)

see attached file

See attached file