The Changing Relationship between Banking Crises and Capital Inflows

• DOI: http://doi.org/10.1111/rode.12242
• Published date: 01 May 2016
• Author: Jason Rastovski
• Date: 01 May 2016

The Changing Relationship between Banking Crises

and Capital Inflows

Jason Rastovski*

Abstract

Different types of capital inflows have varied effects when predicting banking crises in emerging and

developing economies, and these relationships have meaningfully changed over time. In a sample of 29

developing and emerging economies over the period 1976–1991 increases in short-term debt inflows raised

the probability of a banking crisis while increases in inflows for long-term borrowing by the private sector

had the opposite effect. Conversely, over the period 1992–2007 increases in inflows for long-term borrowing

by the private sector and for equity investment both increased the probability of a banking crisis. The

findings suggest distinct optimal capital account liberalization policies between the two periods.

1. Introduction

Since the early 1970s banking crises in developing and emerging economies have

become more frequent and capital inflows have increased considerably. The

literature has found that increases in total capital inflows generally precede banking

crises. In this paper we regress a binary variable indicating the onset of a banking

crisis on different types of lagged capital inflows. In a sample of 29 countries over

1976–2007 the probability of a country experiencing the onset of a banking crisis in

any given year was 5.9%. Over this time period it is not clear whether or not the

types of capital inflows a country experiences matters in predicting banking crises;

however, splitting the sample into two equal periods shows that the type of inflow

does matter and that the relationship between inflows and banking crises has

changed over time.

From 1976–1991 to 1992–2007 the probability of the outbreak of a banking crisis

in a given country increased from 5.4% to 6.5%. During 1976–1991 a 5 percentage

point increase in the lagged 5-year growth rate of short-term external debt

(borrowed by the public and private sectors combined) increased the probability of

a banking crisis by 3.1 percentage points, whereas a 5 percentage point increase in

the lagged 5-year growth rate of long-term external borrowing by the private sector

decreased the probability of a banking crisis by 2.2 percentage points. Other types

of capital inflows are not statistically significant predictors in the most complete

regression specification. During the 1992–2007 period increases in the growth rates

of long-term external borrowing by the private sector and foreign equity investment

(portfolio and direct investment combined) are statistically significant in predicting

banking crises. In this period a 5 percentage point increase in the lagged 5-year

growth rate of long-term external borrowing by the private sector increased the

probability of a banking crisis by 1.1 percentage points, and a 5 percentage point

increase in the lagged 5-year growth rate of foreign equity investment increased

*Rastovski (Corresponding author): Centre College, Danville, KY, 40422, USA. Tel: +1-859-238-6504;

E-mail: jason.rastovski@centre.edu.

Review of Development Economics, 20(2), 514–530, 2016

DOI:10.1111/rode.12242

©2016 John Wiley & Sons Ltd

that probability by 2.0 percentage points. Given the high volatility of the growth

rates of these types of external liabilities

1

and the observed frequency of banking

crises these effects are economically large.

This paper contributes to the literature on banking crises, financial openness, and

the differentiation between gross and net capital flows by: being the first to tie

equity and capital inflows owing to short-term debt and different types of long-term

debt to banking crises in a way that their relative impact can be observed;

2

using a

de facto measure of financial openness in the volume of capital inflows (following

Kose et al., 2009; Lane and Milesi-Ferretti, 2007) when assessing the impact of

inflows on banking crises; emphasizing the role gross rather than net capital flows in

crisis analysis (following Obstfeld, 2010; Lane, 2000; Mendoza and Terrones, 2008;

Forbes and Warnock, 2011).

Because banking crises can be disruptive and expensive to remediate,

understanding their causes has meaningful policy and welfare implications.

Eichengreen and Rose (1998) cite the cost of resolving financial crises and

recapitalizing banking systems at 10% of gross domestic product (GDP) in Malaysia

(1985–1988), 15% of GDP in Mexico (1994–1997), 20% of GDP in Venezuela (1994–

1997), 30% of GDP in Chile (1981–1986) and 50% of GDP in Kuwait (1990–1991).

Reinhart and Rogoff (2008) find that banking crises typically lead to sharp declines in

tax revenues and significant increases in government expenditures, and that on

average government debt rises by 86% in the 3 years after a banking crisis. Given that

a banking crisis is frequently resolved by government nationalization of banks this

increase in sovereign borrowing is not unexpected. They also find that banking crises

tend to be much shorter in duration than sovereign debt defaults, presumably because

of it being more costly to incur the negative effects on trade and investment (from a

banking crisis) than for the government to avoid international creditors (following a

default). In addition to showing the importance of studying banking crises, another

strand of the literature ties banking crises to capital inflows.

Kaminsky and Reinhart (1999) find that financial liberalization often precedes

banking crises. In 18 of the 26 banking crises they study, the financial sector had

been liberalized within the preceding 5 years. They use a mixture of developing and

developed countries, but Eichengreen and Arteta (2002) point out that the

characteristics of banking crises in developing and emerging economies are distinct

from those in developed economies. They indicate that banks account for a larger

share of financial institutions in developing countries, that those banks’ liabilities

tend to have shorter maturities, they have less developed supervision and

regulation, and they have less ability to hedge external risks than their counterparts

in developed countries. For these reasons our analysis focuses on developing and

emerging economies. The second key difference between this paper and Kaminsky

and Reinhart (1999) is that they consider the effect of general domestic financial

liberalization (captured using the M2 multiplier, domestic credit to GDP, the real

interest rate and the lending–deposit rate ratio) on banking crises, where we

specifically look at de facto financial openness. The de jure financial openness index

constructed in Chinn and Ito (2008) could also be used to conduct analysis similar

to Kaminsky and Reinhart (1999). This index quantifies formal policy restrictions

on a country’s interactions with the rest of the world.

3

In this paper the focus is on

identifying the impact of changes in types of capital inflows, which may not be

captured by the de jure measure of capital account openness used as a component

of the Chin and Ito index. Others have analyzed the impact of capital inflows on

banking crises, but restrict their analyses to only consider inflow “bonanzas.”

BANKING CRISES AND CAPITAL INFLOWS 515

©2016 John Wiley & Sons Ltd