Нелинейный анализ доходности облигаций и акций: данные из Турции

This study aims to implement the Markov Switching Vector Auto Regression model which is one of the main pars of Non-linear models for analyzing the Turkish stock and bond market relationships. It is crucial for an investor whether the assets in his/her portfolio will compensate for each other against a shock or in economic environment which is not beneficial. This can be confirmed via possible increases and decreases in the correlation coefficient between the stock and bond returns. A further progress has been made by using the excess return series which are calculated in excess of the risk-free rate. After the series had been investigated thoroughly, models with a number of regimes and lagged values have been compared. According to results from the Hannan-Quinn Information Criterion the model with two regimes and no autoregressive part have been selected. It has been shown that the correlation between the two asset returns tend to be higher during economic downturns. The results have shown that in the period of January 2008 and January 2019, stock market and the bond market have not been useful for portfolio diversification purposes and negative contagion was observed among the two markets. Markov Switching Vector Auto Regression model was chosen as main non-linear model for our research because of its less errors while estimation. Meanwhile, based on the estimation of Markov Switching Vector Auto Regression model which has been used by comparing this model results with linear model namely linear regression using the same dataset and it has been confirmed that non-linear model is much stronger for Turkish markets than any other markets in developing region and relative advantages of non-linear model in Turkey increased over time.