A behavioural asset pricing approach: Perspectives from the South African stock market

Date
2024-12
Journal Title
Journal ISSN
Volume Title
Publisher
Stellenbosch University
Abstract
The estimation of a company’s cost of equity has been the subject of extensive debate due to its subjective nature. Existing asset pricing models being employed to estimate the cost of equity (returns) assume investors have homogeneous beliefs and expectations pertaining to expected returns. There is, however, heterogeneity in investors’ expectations of returns. Due to this heterogeneity not being accounted for, the accuracy of existing asset pricing models is questioned. To improve its accuracy, sentiment stochastic processes can be incorporated into asset pricing models to account for the heterogeneity in the decision-making of multiple investors in the market, during different market conditions. Prior research in South Africa has concluded that investor sentiment should be considered an explanatory variable in asset pricing models; however, none of these studies have augmented existing asset pricing models such as the Carhart four-factor (CH4) model, with an investor sentiment variable. The primary objective of the current research was, therefore, to account for the human element in the South African stock market, by investigating the impact of investor sentiment on the cross-sectional variation of the cost of equity for securities listed on the Johannesburg Stock Exchange (JSE) for the period 2003 to 2019. To address this aim, the CH4-model was augmented with investor sentiment as an affective risk factor. One indirect measure (composite sentiment index) and three direct measures (survey and textual analysis-based measures) were empirically tested. By employing a judgement sampling technique, the final sample comprised of 321 securities that formed part of the FTSE/JSE All share index (ALSI) over the 17-year study period. Variations of regression analyses were employed for hypothesis testing. Ordinary least squares (OLS) time-series regressions revealed that a statistically significant relationship exists between returns and the direct measures of investor sentiment only. These results indicated that direct and indirect measures of investor sentiment contain unique information. A further finding was that the two direct proxies that resembled sentiment toward the economic and business conditions in South Africa, reported significant explanatory power for returns, of which the proxy measured by means of textual analysis was the most significant predictor. OLS regressions were also conducted for high and low sentiment periods, to assess the statistical significance of the CH4-model risk factors during different sentiment periods. The regressions uncovered contrasting results for only the momentum risk factor. During low sentiment periods, the relationship between the momentum factor and returns was negative and not significant, whilst high sentiment periods revealed a positive and statistically significant relationship. It was, thus, concluded that trading behaviour in the South African stock market, vary across high and low sentiment periods. This finding is in line with Lo’s (2004) adaptive market hypothesis. The Fama and Macbeth (FM) (1973) two-stage regression approach confirmed that investor sentiment does have a statistically significant impact on the cross-sectional variation of the cost of equity for JSE-listed securities. Similar to the OLS regressions, the two direct measures that reflect investors’ overall sentiment towards the country and the business climate within the country, were significant predictors of returns. Furthermore, contrasting relationships were reported between the use of constant and rolling betas. The difference in the direction of the cross-sectional association, confirms the argument that the association between sentiment and returns can be either positive or negative, depending on conditions within the market and economy. The current research makes an important theoretical contribution to asset pricing research and behavioural finance research in South Africa, by providing a theoretical justification for why investor sentiment should be incorporated into asset pricing models. A methodological contribution was also made by constructing a composite investor sentiment index for the South African stock market, and by investigating and comparing several diverse measures of investor sentiment. Educators, investors, financial managers and policymakers are encouraged to incorporate investor sentiment in their curriculums, investment strategies, cost of equity estimations and monetary policies, to reap the potential benefits from emphasising the individuals, whose collective behaviour ultimately drives markets.
Description
Thesis (PhD)--Stellenbosch University, 2024.
Keywords
Citation