The relative importance of capital inflows: some evidence from emerging market economies

ABSTRACT

This article examines the relative importance of the main components of capital inflows for a sample of emerging market economies. Does composition matter? Is there a nexus between capital inflow components? We assess, firstly, how each capital inflow component reacts to important macro and policy variables, and secondly, how the components themselves interact. We find that bank inflows appear the most sensitive to macro factors, institutions matter more for Latin America and external financial factors matter more for Asia. Further, for Latin America, capital inflows interact largely as complements, while for Asia, any expansion of bank inflows might crowd out FDI and portfolio flows.

KEYWORDS:

• Capital inflows
• emerging market economies

JEL CLASSIFICATION:

• F32
• F33

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplemental material

Supplemental data for this article can be accessed here.

Notes

¹ We concern ourselves with inflows in this article as outflows usually contain a different set of determinants.

² Firms would prefer to use retained earnings to finance their investments. However, financially constrained firms may need to resort to borrowing from abroad where FDI is preferred because local borrowers possess more information about foreign direct investors than if local firms issues debt or equity instruments or raise funds through bank financing. Bank-based finance is preferred to portfolio due to issuance costs. For more on Pecking order, see Myers (1984) and Daude and Fratzscher (2008).

³ While the literature on capital flow determinants is quite mature (see Ghosh et al. 2014), the cross effects among capital flows is a comparatively new concept, and policymakers are perhaps less aware of their consequences, especially in combination with macro and policy determinants of capital flows.

⁴ See Table A1 in the Supplementarty data for details. While higher-frequency data for the capital flows and some of the macro and finance variables are available, the data for institutional quality, financial liberalization, financial development and VIX are all annual observations.

⁵ See Table A1 for data details and Table A2 in the Supplemental data for descriptive statistics. This list is not exhaustive, but we feel that the variables included represent the extensive literature on the determinants of capital inflows and capture domestic and external economic effects, institutional characteristics and financial vulnerabilities (see Ahmed and Zlate 2014; Ghosh et al. 2014).

⁶ Lagged values of the first difference of the explanatory variables (CAPOUT, TRADE, FINDEV, IQ, GDP, IR, FINLIB). Three-stage least squares also takes account of the correlation between disturbances across equations in the system.

⁷ Testing was conducted on different orderings for these for little material difference. If we were to consider the A matrix for the variables in Y_t, then we effectively apply a recursive ordering to identify the contemporaneous relationships with the capital flow variables and within the controls and that the elements of the first three columns of the 4^th through to the 11^th row of A are zeros.

⁸ Even though they relate to lagged relationships and not contemporaneous ones, panel Granger Causality tests were generated as diagnostic tests to establish some basic patterns amongst the variables in Y_t.

⁹ The effect on the other inflows fall away relatively quickly, while they are lower but more persistent for FDI and Portfolio.

¹⁰ The conclusions drawn may be subject to the Lucas Critique. Care must be taken particularly with the interpretation of the cross effects results when one considers that the change in the inflow of one component may be due to a policy of financial repression or liberalization on the part of another.

¹¹ The nomenclature pertaining to the impulse responses is ‘impulse:response’. So, FDIL:OFL refers to the response of other inflows. Figures 1 and 2 present multiple responses to the one shock and, as such, does not show error bands as the graphs would be too cluttered. We provide the source impulse response graphs for all variables in the model in the Supplemental data.

¹² This underscores the importance of bank flows in the region during the sample period (See Ito 2000).