Publication: Durum uzayı modellerine markov değişimi yaklaşımı ve ekonometrik uygulaması
Abstract
Bu tez, Markov Değişimli Durum Uzayı Modellerini, mevsimsel Fourier terimlerinin modele dahil edilmesiyle ele alan bir modelleme yaklaşımını tanıtmaktadır. Markov Değişimli Durum Uzayı Modelleri, Markov Değişim Modelleri ile Durum Uzayı Modellerinin birleştirilmesiyle geliştirilen modellerdir. Tahmin yöntem olarak Markov Değişim Modelleri ve Durum Uzayı Modellerinden farklı bir yapıya ihtiyaç duymaktadır. Kim (1994) filtresinin geliştirilmesiyle bu ihtiyaç ortadan kalkmış ve Markov Değişimli Durum Uzayı Modellerinin ekonometride kullanılması mümkün hale gelmiştir. Yazında, Markov Değişimli Durum Uzayı Modelleri nispeten yer alsa da Markov Değişimli Durum Uzayı Modelleri yapısına mevsimsel Fourier terimlerine eklemeye yönelik bir çalışmaya rastlanılmamıştır. Markov Değişimli Durum Uzayı Modellerine mevsimsel Fourier terimlerinin eklenmesiyle yazındaki bu eksiklik giderilmeye çalışılmış ve modelleme esnasında tanımlama hatalarının minimize edilmesi amaçlanmıştır. Tezin uygulama kısmında ise mevsimsel Fourier terimlere sahip Markov Değişimli Durum Uzayı Modelleri yapısı kullanılarak Türkiye’de enflasyon serisinden, 2003M1-2024M12 dönemi için, toplam enflasyon belirsizliği, iç politikadan kaynaklanan enflasyon belirsizliği ve dış politikadan kaynaklanan enflasyon belirsizliği serisi olmak üzere üç adet yeni seri elde edilmiş ve bu dört seri arasındaki nedensel ilişki analiz edilmiştir. Uygulama sonucunda, tüm serilerin birbirleriyle nedensel bir ilişki içinde oldukları tespit edilmiştir.
This thesis introduces a modeling approach that deals with Markov Switching State Space Models by including seasonal Fourier terms in the model. Markov Switching State Space Models are models developed by combining Markov Switching Models and State Space Models. It needs a different estimation method that is not used in Markov Switching Models and State Space Models. By the development of the Kim (1994) filter, this need has been eliminated and it has become possible to use Markov Switching State Space Models in econometrics. Although Markov Switching State Space Models are relatively included in the literature, there has not been a study aimed at adding seasonal Fourier terms to the structure of Markov Switching State Space Models. By adding seasonal Fourier terms to the Markov Switching State Space Models, this deficiency in the literature was tried to be eliminated and it was aimed to minimize specification errors during modeling. In the application chapter of the thesis, three new series, namely total inflation uncertainty, inflation uncertainty caused by domestic policy, and inflation uncertainty caused by foreign policy, were obtained from the inflation series in Turkey for the period 2003M1-2024M12 using the structure of Markov Switching State Space Models with seasonal Fourier terms, and the absence of a causal relationship between these four series has been tested by causality tests. As a result of the application, it has been determined that all the series are in a causal relationship with each other.
This thesis introduces a modeling approach that deals with Markov Switching State Space Models by including seasonal Fourier terms in the model. Markov Switching State Space Models are models developed by combining Markov Switching Models and State Space Models. It needs a different estimation method that is not used in Markov Switching Models and State Space Models. By the development of the Kim (1994) filter, this need has been eliminated and it has become possible to use Markov Switching State Space Models in econometrics. Although Markov Switching State Space Models are relatively included in the literature, there has not been a study aimed at adding seasonal Fourier terms to the structure of Markov Switching State Space Models. By adding seasonal Fourier terms to the Markov Switching State Space Models, this deficiency in the literature was tried to be eliminated and it was aimed to minimize specification errors during modeling. In the application chapter of the thesis, three new series, namely total inflation uncertainty, inflation uncertainty caused by domestic policy, and inflation uncertainty caused by foreign policy, were obtained from the inflation series in Turkey for the period 2003M1-2024M12 using the structure of Markov Switching State Space Models with seasonal Fourier terms, and the absence of a causal relationship between these four series has been tested by causality tests. As a result of the application, it has been determined that all the series are in a causal relationship with each other.