Abstract:
Long term volatility of stock returns plays a major role in determining the weight of stocks in forward looking portfolios. This thesis investigates the long run stock return volatility in Turkey in two parts: i) unconditional statistics derived from rolling windows samples, ii) conditional statistics derived from a predictive variance analysis using a Bayesian Markov Chain Monte Carlo approach. Unconditional variance decreases with the investment horizon and it becomes more likely for stock returns to beat fixed interest returns. The predictive variance analysis also suggests that return volatility decreases with time. The risk-inducing effect of momentum is dominated by the risk-reducing effect of the negative correlation between the error terms of the current and expected return equations. Assuming a time-varying covariance matrix reduces the portion of predictive variance attributable to the identical and independently distributed risks. Having fewer observations increases the predictive variance estimate. Overall results suggest that it is more preferable from an investor’s perspective to make long-term investments in Borsa Istanbul.
