Abstract
M.Com. (Financial Economics)
This minor dissertation revisits the relationship between stock price valuation and demographic changes with an empirical application to South African data over the period 1985 to 2016. Built from the life cycle hypothesis, the study reexamines the theoretical link between birth rate and the dividend-price ratio proposed by Geanakoplos et al. (2004) based on the US demographic and economic development in the 20th century. Calibrated results prove support to a non-monotonic relationship between middle to young ratio and stock price valuation; possibly suggesting the existence of multiple equilibria between stock prices and demographic changes.
The derived theoretical link is further assessed empirically using South African data. Particularly, the study considers the dividend yields from all share, financial and industrial categories. Results from both parametric (non-linear cointegration) and nonparametric (neural network) confirm the nonlinear association between demographic changes and dividend yields across selected asset categories. It is found that stock prices from all share category have a stronger response to negative than positive demographic shock. In the financial category, the asymmetric response is equally reported, not only in terms of magnitude but also in terms of sign with stronger effect from the positive shock. In contrast, the industrial category could not confirm the existence of asymmetric response to demographic changes.
However, despite the evidence of asymmetry in the response of asset prices to positive and negative demographic changes as vindicated by the nonlinear cointegration output, it appears that the type of nonlinearity remains unclear. Given the ability of neural network in modelling complex dynamics, the traditional feed forward backward propagation is then used to evaluate the predictability of demographic and economic factors in forecasting asset prices. Empirical findings suggest that using middle to young ratio and wage income to predict dividend yields results in a prediction accuracy of more than 72% across asset category. This is consistent with the nonlinear cointegration output that more than 75% variation in the dividend yield is explained by wage income and middle to young ratio across asset classes; hence confirming the importance of demographic changes in shaping the dynamic behavior of asset development.