Real effects of financial market integration: does lower home bias lead to welfare benefits?

† This paper has been written during an academic visit at Universitat Autònoma de Barcelona, within the European Network for Training in Economic Research program.

Abstract

This paper proposes equity home bias as a proxy for financial integration in the ongoing empirical debate on the impact of financial integration on economic growth. In integrated markets, investors are expected to take full advantage of the potential for international diversification. The extent of equity home bias (i.e. overinvesting in domestic stocks and foregoing gains from international diversification) provides a relevant quantity-based measure of financial integration. Using different techniques to compute home bias, this paper investigates whether countries with lower home bias experience faster economic growth. Additionally, the analysis extends to the link between (decreasing) home bias and international risk sharing and income inequality. The results suggest that financial integration, proxied by the decreasing equity home bias, is positively associated with economic growth and international risk sharing. At the same time, it appears that higher financial integration pairs with higher income inequality.

Keywords:

• home bias
• international portfolio choice
• economic growth
• model uncertainty
• international risk sharing
• income inequality

JEL Classifications:

• F36
• G11
• G15

Notes

† This paper has been written during an academic visit at Universitat Autònoma de Barcelona, within the European Network for Training in Economic Research program.

1. This paper makes the additional simplifying assumption that currency risk is not priced. See Solnik (1977) for further discussion supporting this empirical choice.

2. http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html.

3. The ‘difference’ and ‘system’ Generalized Method of Moments (GMM) estimators of Arellano and Bond (1991) and Arellano and Bover (1995)/Blundell and Bond (1998) that have become the norm in dynamic panel data estimations are designed for ‘large N, small T’ panels and therefore not applicable in this case. In panels with ‘(relatively) small’ N and ‘(relatively) large’ T (as in this paper), the resulting large number of instruments against the overall number of observations makes the model always appear correctly identified, as the tests fail in 100% of the cases to detect the invalidity of the GMM estimators. See Roodman (2006, 2009) for a detailed analysis of this issue. The fixed-effects estimator is proposed in this case as the better alternative (Verbeek 2004; Kiviet and Niemczyk 2007; Doko and Dufour 2010).

4. To reduce the influence of large outliers, data that are positive by construction are used in logarithms (see Edison et al. 2002; Schularick and Steger 2006).

5. Methodologically, with moderate endogeneity and weak instruments, the least-squares estimation carries over the instrumental variables choices (Verbeek 2004; Kiviet and Niemczyk 2007; Doko and Dufour 2010).

6. For instance, in a study exploiting the time-series dimension of data, Arestis and Demetriades (1997) employ Granger-causality tests in a vector autoregression framework to find evidence of reverse causality between economic growth and financial development indicators for the USA.

7. For one of the most extreme criticisms of causality inferences in a regression framework, regardless of methodological choices, see Freedman (1991).

8. Subsample analysis searching for EU/EMU effects has been performed for all the models presented in the paper (results are available from the author on request). A few clear patterns emerged. They mark a certain difference between the wider EU and the more recently and strongly integrated Euro Area, especially when it comes to the impact of home bias on economic growth. A better integrated monetary union appears to be less exposed to the negative effects of (higher) home bias on economic growth, but at the same time, to benefit the most from integration in terms of consumption smoothing. No significant effect has been found for the income smoothing or income inequality equations.