Browsing by Author "Molefe, Thabo"

Filter results by typing the first few letters
Now showing 1 - 1 of 1
  • No Thumbnail Available
    Item
    Forecasting Stock Prices: Do Taylor Rule Fundamentals Matter?
    (Malikane, Christopher, 2024) Molefe, Thabo
    We build on studies by Ince et al. (2016), and Dladla and Malikane (2019) to explore the role of Taylor rule fundamentals in forecasting stock market returns in a small open economy. We investigate a mix of 11 countries from 1970 to 2022. We derive a stock return forecasting model using the dividend yield interest rate spread and introduce Taylor rule fundamentals into our model using Taylor’s (1993) rule. We compare the performance of our model to the autoregressive with lag-2 benchmark model and the historical average using Theil’s U, Diebold and Mariano’s (1995) MSE-T, and Harvey et al.’s (1998) ENC-T statistics to test for relative forecast accuracy, equal forecast ability, and model encompassing. Since the models are linear and nested, we follow McCracken (2007) in using asymptotic critical values for MSE tests and proceed to also follow Clark and McCracken (2007) in using asymptotic critical values for ENC tests. We split the data 50-50 and conduct an OLS regression on the in-sample portion of the data to obtain parameters. We obtain in-sample results that show the signs of the coefficients for most of our endogenous variables to be as per literature. The sign of the coefficient of inflation is negative in line with studies by Sousa, Vivian, and Wohar (2016), and Cooper and Priestley (2009) among others, and the output gap has ambiguous signs in line with Good News Case (GNC) and Bad News Case (BNC) by Blanchard (1981). However, the sign of the coefficient of the real effective exchange rate is positive for most countries in contrast to the reviewed literature by Wong (2017), and Hau and Rey (2006). We conduct out-of-sample forecasting using the recursive method due to its sensitivity to data structure changes and adaptability to data patterns. The results of the out-of- sample analysis show our model outperforming the autoregressive-lag 2 benchmark model for the 3-month horizon US and Brazil, and the 12-month horizon for most countries. It also outperforms the 3-month horizon Australia, and the 12-month horizon Poland historical average benchmarks