Is the Feldstein–Horioka puzzle valid in lucky seven countries?

ABSTRACT

In recent years, seven countries, called lucky seven have come to the fore due to improvements in governance quality and economic growth. Thus, examining the investment-saving nexus, which is an important factor of sustainable growth, is very important in the creation of economic policies. In this paper, the nexus between savings and investments in lucky seven countries are examined during the period 1990–2017. Westerlund (2006) and Westerlund (2007) cointegration test results denote that savings and investments are cointegrated. Common correlated effects mean group (CCEMG) and augmented mean group (AMG) estimation results showed that savings have a significant and positive effect on investments in Indonesia, Mexico, and Poland. Kónya (2006) panel causality test results indicate that there is causality running from savings to investments in India and Mexico. Asymmetric test results denote that there is causality from positive shocks of savings to positive shocks of investments in Colombia, India, Kenya, and Mexico. In terms of negative shocks, no causality is determined. The overall results show that the hidden Feldstein–Horioka puzzle is valid in some lucky seven countries.

KEYWORDS:

• Lucky seven countries
• Feldstein–Horioka puzzle
• bootstrap panel cointegration
• asymmetric causality

JEL CLASSIFICATIONS:

• C23
• E21
• F21

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Sinem Eyuboglu http://orcid.org/0000-0002-3525-9173

Umut Uzar http://orcid.org/0000-0003-3438-1422

Notes

1 Findings from the causality test may be sensitive to the lag structure, deciding the optimal lag length(s) is very important for robustness of findings. In the panel, equation- and variable-varying lag structure could result in a raise in the computational burden substantially. In order to solve this problem, we followed Nazlioglu, Lebe, and Kayhan (2011). They allow maximal lags to differ across variables, but to be the same across equations. Then they estimate the system for each possible pair of ly₁,lx₁, ly₂, and lx₂ respectively by assuming from 1 to 4 lags and then select the combinations that minimize the Akaike information criteria.