The Feldstein–Horioka puzzle in South Africa: A fractional cointegration approach

Abstract

The Feldstein-Horioka puzzle (FH), that is the strong correlation between saving and investment in a world where obstacles to capital mobility are limited, has been studied extensively since it was exposed in 1980. Even though the theoretical and empirical literature has examined many of its potential causes, the puzzle persists. This paper aims at shedding further light on the issue by investigating the relationship between saving and investment in South Africa since 1946 using fractional integration and cointegration techniques to account for high persistence in the series.We find evidence of fractional cointegration between saving and investment, indicating some degree of persistence in the gap between the two variables. We also find a structural break in saving and investment ratios to GDP around 1980, which roughly coincides with the start of a financial deregulation process in South Africa. While fractional cointegration holds before the break, it does not thereafter. In other words, while the FH puzzle is observed before the start of financial deregulation, it subsequently disappears. This suggests financial deregulation may have loosened the link between saving and investment.JEL Codes: C22, C32, E21, E22, F41.

The Feldstein–Horioka hypothesis for African countries: Evidence from recent panel error‐correction modelling

“…Their results using the FMOLS and DOLS panel cointegration estimations showed only a slightly higher degree of capital mobility in African countries due to economic integration. Similar to the studies of Adedeji and Thornton (2007) and Bangake and Eggoh (2012), Gil‐Alana et al (2016) found that the saving retention coefficient in South Africa decreased after financial deregulation process, which was considered as an observed structural break around 1980.…”

“…It is noteworthy to observe that some attempts in testing the Feldstein–Horioka hypothesis for African countries have been made in the last decade and a half. Some important empirical investigations in this regard are, the studies conducted by Cooray and Sinha (2007), De Wet and Van Eyden (2005), Murthy (2005), Adedeji and Thornton (2007), Cyrille (2010), Bangake and Eggoh (2011, 2012), Kumar, Sen, and Srivastava (2014), Barros and Gil‐Alana (2015), Gil‐Alana, Andre, Gupta, Chang, and Ranjbar (2016), Raheem (2017). Adedeji and Thornton (2007) employed FMOLS and DOLS estimators allowing for heterogeneity and cross‐sectional dependence effects.…”