On Modeling Number of District Share Transactions Using TwoLevel Hierarchical Structure with Bayesian Approach

Abstract

The condition of the Indonesian economy tends to fluctuate. One of the methods that
can be used to observe the fluctuating development of the country's economy is looking at
the development of the capital market as a leading indicator of the economy. This
research intends to model the number of stock transactions in the capital market using a
two-level hierarchical Judging from Social Population and Economy. The first (micro)
level is the district with its characteristics of the region and the second (macro) level is
the province with the distinguishing variable characteristics. A Bayesian hierarchical
model approach coupled with the MCMC algorithm to the three possible data patterns of
these transactions, i.e. Log-Normal 2 parameters, Log-Normal 3 parameters, and LogLogistic distribution is proposed. The analysis shows that Bayesian hierarchy modeling is
better than the Bayesian one-level, and employing the Log-Normal 3 parameter
distribution is better than the others. This can be known through the smallest DIC value.
The variation of the micro regression coefficients between provinces proved to be
significantly influenced by the characteristics of the district and the characteristics of the
province. This approach has succeeded to demonstrate that the Bayesian hierarchy model
more able to illustrate the influence of socioeconomic and population factors at different
levels on the growth of the number of stock transactions. The condition of the Indonesian economy tends to fluctuate. One of the methods that
can be used to observe the fluctuating development of the country's economy is looking at
the development of the capital market as a leading indicator of the economy. This
research intends to model the number of stock transactions in the capital market using a
two-level hierarchical Judging from Social Population and Economy. The first (micro)
level is the district with its characteristics of the region and the second (macro) level is
the province with the distinguishing variable characteristics. A Bayesian hierarchical
model approach coupled with the MCMC algorithm to the three possible data patterns of
these transactions, i.e. Log-Normal 2 parameters, Log-Normal 3 parameters, and LogLogistic distribution is proposed. The analysis shows that Bayesian hierarchy modeling is
better than the Bayesian one-level, and employing the Log-Normal 3 parameter
distribution is better than the others. This can be known through the smallest DIC value.
The variation of the micro regression coefficients between provinces proved to be
significantly influenced by the characteristics of the district and the characteristics of the
province. This approach has succeeded to demonstrate that the Bayesian hierarchy model
more able to illustrate the influence of socioeconomic and population factors at different
levels on the growth of the number of stock transactions.