초록 close

Most research on estimating dynamic optimal choice model has been using conventional econometric methods such as Maximum Likelihood (ML) Method and Generalized Method of Moment (GMM). It is well known that when the model has many state variables, the estimation becomes difficult due to the well-known “Curse of Dimensionality”. In this paper, we applied a new estimation technique that is based on the Bayesian estimation to confront this issue. To show the effectiveness of the Bayesian dynamic programming estimation algorithm, we first simulate the model of entry and exit in the export market, and then estimate back the parameters of the model based on simulated data. It enables us to solve the dynamic programming problem and estimate the parameters simultaneously instead of sequentially. This new method makes the computational burden of estimating the dynamic programming model on the same order of magnitude as those of estimating static model. We estimated a simple model of entry and exit behavior in the export market. We used the Bayesian dynamic programming method suggested by Imai, Jain, and Ching (2001) and successfully estimate back the true parameter from the simulated data.


Most research on estimating dynamic optimal choice model has been using conventional econometric methods such as Maximum Likelihood (ML) Method and Generalized Method of Moment (GMM). It is well known that when the model has many state variables, the estimation becomes difficult due to the well-known “Curse of Dimensionality”. In this paper, we applied a new estimation technique that is based on the Bayesian estimation to confront this issue. To show the effectiveness of the Bayesian dynamic programming estimation algorithm, we first simulate the model of entry and exit in the export market, and then estimate back the parameters of the model based on simulated data. It enables us to solve the dynamic programming problem and estimate the parameters simultaneously instead of sequentially. This new method makes the computational burden of estimating the dynamic programming model on the same order of magnitude as those of estimating static model. We estimated a simple model of entry and exit behavior in the export market. We used the Bayesian dynamic programming method suggested by Imai, Jain, and Ching (2001) and successfully estimate back the true parameter from the simulated data.


동태적 적정선택모형에 관한 많은 연구가 최빈우도법(ML) 및 일반화적률 법(GMM) 등 일반적인 계량경제학적 접근법에 기초한다. 하지만, 상태변수 가 많아질수록 이와 같은 계량추정법은 규모의 저주(Curse of Dimensionality)로 알려진 문제에 직면한다. 본 고에서는 이러한 문제를 해결하기 위한 베이지언 추정에 기초한 새로운 추정기술을 적용한다. 베이지 언 동태프로그래밍 추정법의 효율성을 보여주기 위하여, 먼저 수출시장에서 의 진출입모형을 시뮬레이션한다. 그 이후, 시뮬레이션 데이터에 기초하여 모형의 변수를 재추정한다. 본 추정법은 규모의 저주 문제를 해결하면서 동 태 프로그래밍 문제를 풀 수 있도록 해주며, 단계별 추정이 아닌 동시 추정 을 가능하게 해준다.