University of Groningen
Faculty of Economics and Business
MSc in International Economics and Business
Master Thesis
Financial fragility and growth loss post-financial crisis
Author: Koen Wolters Student number: S2693682
E-mail address: k.wolters.2@student.rug.nl Supervisor: Prof. dr. D.J. Bezemer Co-assessor: T.M Harchaoui, Ph.D.
Date: 1st February 2019
Abstract
This paper investigates whether the financial fragility of a country prior to the financial crisis of 2008 impacted the post-crisis recession severity (measured as growth loss). The sample contains 40 countries with data from 1998-2012, using A new international financial fragility
database1 to measure financial fragility. The results show that financial fragility significantly
increased the growth loss, alongside credit growth. Additionally, this paper analyzes how bank variables (such as bank capitalization) affect financial fragility.
Keywords: Financial fragility, growth loss, financial crisis
2
1. Introduction
The global financial crisis started after the collapse of the Lehman Brothers in late September of 2008, soon after the world experienced large economic downturns and recessions. The severity of the post-crisis recession differed per country, as countries that had more private debt suffered more severe recessions (Lane and Milesi-Ferretti; 2011, Claessens et al., 2010; Feldkircher, 2014). However, Zhang & Bezemer (2015) concluded that the composition of debt mattered as mortgage credit fueled the crisis. Since the crisis had detrimental effects, could it have been foreseen?
Scholars have identified variables that increased the severity of the post-crisis recession. The post-crisis recession in this paper refers to the recession that occurred in 2009, after the global financial crisis of 2008. Variables such as credit growth impact the severity of the post-crisis recession (Claessens et al., 2010). Could it be that the behavior and managerial decisions of commercial and investment banks within an economic system made a country prone to larger losses from the outset of the crisis, due to the fact that the decisions made the banking system initially more fragile?
This paper aims to shed light on the topic of financial fragility. Firstly, I examine the contribution of financial fragility to the severity of the post-crisis recession. Secondly, this paper discusses the impact that bank variables, such as bank capitalization have on financial fragility. This is relevant because the findings can potentially indicate whether financially fragile countries suffered more severe recessions. This can in turn, have policy implications, as countries will try and reduce their fragility to avoid losses in the next downturn. Additionally, the second objective of this paper is to assess the bank variables which impact financial fragility. Although the determinants of financial fragility have been investigated, the academic merits are that a different measure of financial fragility is used compared to the research done by Aspachs
et al. (2007) and Martínez-Jaramillo et al. (2010) who also investigate the determinants of
financial fragility.
This paper uses the New International Financial Fragility database by Andrianova et al. (2015), which is a panel dataset that contains a national aggregated measure of financial fragility and bank variables, from 1998-2012. The bank variable data ranges from traditional commercial banks to investment banks.
It is interesting to study if the severity of the crisis was impacted by financial fragility, because the crisis started due to an unpredicted shock, and financial fragility is, in essence, the vulnerability of a system to a shock. The effect of a shock causes more fragile systems to ‘break' and experience economic losses. It is important to understand that the fragility of the economy prior to the crisis did not cause the crisis. Fragility does not mean an economy will crash. A crash initiates when a fragile system experiences a shock, which they cannot handle (Berger and Pukthuanthong, 2016).
Therefore, the research questions I attempt to answer in this paper are:
-How did the financial fragility of a country, prior to the financial crisis, affect the
post-crisis recession severity?
3
The fragility of a country’s financial system is determined by multiple different factors. Aspachs et al. (2007), state that the bank’s default probability and its profitability determine the financial fragility of a system (Aspachs et al., 2007). Hence bank characteristics ranging from managerial decisions to a banks’ return on assets can impact the profitability. Financial fragility increases when the profitability of banks decreases. Borrowers failing to repay loans cause the borrowers’ indebtedness to increase. Persola (2011) identified that large indebtedness of borrowers increases the financial fragility of an economy. Additionally, shocks above a particular threshold level will cause severe losses to a fragile economy (Van Roye, 2014).
As the world economy is becoming more integrated, financial fragility plays a pivotal role, via the contagion channel. Financial integration increases the fragility of a system because shocks transmit and amplify through interconnected countries when the defaults of one country spread and start a downward spiral of events, as displayed in the crisis (Martínez-Jaramillo et
al., 2010 and Manz, 2010).
The results of this paper conclude that financial fragility prior to the crisis did in fact increase the severity of the post-crisis recession, alongside credit growth. Additionally, this paper finds that decrease in bank capitalization, return on assets, and managerial efficiency, increase financial fragility.
4
2. Literature review
2.1. Definition of financial fragility
To be able to understand financial fragility we have to define it. Table 1 summarizes the definitions of financial fragility according to multiple scholars. It is evident that there is not a clear cut single definition. A theme that reoccurs throughout the definitions is the ability to cope with a shock, as seen in the definitions from Allen and Gale (2004), Mishkin (1994) and Tsomocos (2003). Additionally, another theme is that fragile systems are unable to deal with the shock and this has negative consequences, such as an increase in defaults, a decrease in profitability and a decrease in productivity.
After consolidating the definitions in table 1, this paper uses the themes of the provided definitions to construct our own definition. A financial system is fragile when it is unable to
cope with an unexpected shock that leads to disproportionately large effects in terms of GDP growth.
Table 1: Definitions of financial fragility
Scholars Definition
Allen and Gale (2004) “A system is fragile when a small aggregate shock in the demand for liquidity leads to a disproportionately large effect in terms of default or asset-price volatility” Aspachs et al. (2007) “A combination of a high probability of default and low bank profitability”
Mishkin (1994) “A system is fragile when there is a shock that interferes with information flows so that the financial system is unable to channel funds to the most productive investment opportunities.”
Tsomocos (2003) “When there is a substantial default of a number of households and banks, without necessarily becoming bankrupt, occurs and the aggregate profitability of the banking sector decreases significantly.”
When discussing the topic of financial fragility, financial instability often arises. There is a distinction between the two terms. Financial stability refers to managing the variation surrounding the economic performance of a country, smaller variation means more stability. Hence instability would be when the financial performance of a country differs substantially per period. On the other hand, financial fragility refers to the ability to cope with shocks and make sure that the shock does not cause further adjustments in the economic performance. Therefore, stability regards the reduction of the variation of economic performance and fragility is the ability to cope with a shock that causes the variation of economic performance to be large. This implies that a country can be relatively stable but simultaneous be fragile.
2.2. Effects of financial fragility
5
or not. Pesola (2011) questioned whether bank loan losses were enhanced by an increase in financial fragility, measuring financial fragility by higher aggregate indebtedness ratio. This approach differs from the measurement of financial fragility by Aspachs et al. (2007), as they look at multiple2 factors that contribute to financial fragility. Although Pesola (2011) implements macroeconomic factors such as interest rate, he neglects the fact that there are other components to financial fragility such as default probability.
Aspachs et al. (2007) state that an economic system is most fragile when there is a combination of a high probability of default and low bank profitability. When there is a shock to a fragile economy, it will have larger negative consequences compared to a less fragile economy. They collected data from seven developed countries3 from 1990 to 2004. The data contained information on the episodes of financial stress a country experienced and the probability of default of the banking sector. Whereas Pesola (2011) conducted his research by compiling data on the annual aggregated banking sector losses of nine4 central banks. The sample period was from the early 1980s until 2004. He investigates what the effect of private indebtedness and macroeconomic shocks, such as changes in the interest rate and output, have on bank loan losses, where loan losses are the losses accrued when borrowers are unable to pay back their loans to the banks.
Aspachs et al. (2007) find that fragile countries that are not constrained by capital adequacy requirements were hit more severely in GDP. This implies that financial fragility is reduced when there are capital adequacy requirements, as the consequences of a shock are less in that condition (Aspachs et al., 2007). These results complement the findings of Pesola (2011), who found that the basic model, where the dependent variable was bank loan losses and the independent variables were real estate shocks, income surprise variable, and aggregate indebtedness ratio explained “70% of the observed variation in loan losses to relative total loans” (Pesola, 2011). Hence, the variation of loan losses is impacted to a certain extent by the indebtedness ratio, which was the measurement of financial fragility. This was also consistent with, Jappelli et al. (2013) who found that countries were more fragile when households are heavily indebted (Jappelli et al., 2013).
Furthermore, certain factors and conditions can impact the financial fragility of a country, however, it is important to note that having an inherent financially fragile country does not exclusively increase the significance of the likelihood of a crisis. The negative consequences of being a financially fragile country are triggered when there is a boom taking place at an unsustainable pace. Hence if a country is more fragile and there is a boom taking place, it makes that country more susceptible to a crisis according to Martínez-Jaramillo et al. (2010).
2.3. Reducing financial fragility
The financial fragility of a country can be contained when appropriately regulated. Institutions can play a vital role in their ability to reduce household fragility, as concluded by Jappelli et al. (2013). An example of this are reforms, pro-debtor reforms increase fragility,
2 Macroeconomic variables from the IMF and the OECD. Inflation from the CPI index and the interest rate is the
IFS call money rate. Residential property prices from the BIS
3 Finland, Norway, Sweden, North Korea, United Kingdom, Germany, and Japan
6
whereas pro-creditor reforms decrease fragility. While, Ahrend et al. (2012) identified that structural policies also help reduce financial fragility. Furthermore, a banks’ managerial inefficiency, return on assets and liquidity influence its fragility (Mazlan, 2015; Fielding & Rewilak, 2015; Degryse et al., 2013).
Pro-debtor reforms cause an increase in insolvency rates while pro-creditor reforms have the opposite effect. This means that once a country initiates pro-debtor reforms, such as the 2001 German reforms, where it was made possible for a bankrupt company not to be declared as bankrupt when the court approved a viable repayment plan, the insolvency rates increased. This is because pre-debtor reforms increase the fragility of an economy because investors would be willing to take on projects with more risks because the consequences of becoming insolvent are relatively less devastating. Another example of a pro-debtor reform, is the reform of 1978 in the United States, where it simplified the bankruptcy procedures, again increasing insolvency rates and consequently initiating a pro-creditor reform, to act against those who filed for bankruptcy to avoid debt repayment (Jappelli et al., 2013)
Structural policies can reduce financial fragility through two channels. The first channel is the external financial account structure and the second is the vulnerability to international financial contagion. The external account is important because they concluded that countries were most severely hit by the crisis via their external account and that the share of debt in total external liabilities increased drastically. Specifically, the risk of a crisis was increased by an increase in bank debt, currency mismatches and shorter banking debt maturities. The second channel is vulnerability to international contagion. This is important because structural policies can reduce the volatility or reduce risk of domestic banking crisis when an external financial shock occurs. The structural policies include domestic and externally orientated policies. External policies included targeted capital controls, and domestic policies include stricter macroprudential regulation (Ahrend et al., 2012).
Complementary, Ahrend et al. (2012) found that in order to reduce the risk of a financial crisis, there should be stricter information discloser rules, and capital requirements, while being governed by supervisory authorities. Moreover, they found that macroprudential regulations increase financial stability. The regulations should protect against excessive domestic credit growth. In addition, tax systems that favor debt finance over equity finance increase fragility, because it increases the share of debt in corporate financing (Ahrend et al., 2012).
7
2.4. Systemic risk and financial fragility
Systemic risk5 and financial fragility are intertwined. Throughout this paper, systemic risk will not be captured in the empirical model. This is because systemic risk lies beyond the scope of the objective of the paper. However, it is worthwhile to address, because the contagion channel is part of systemic risk, and this in turn affects financial fragility.
The relationship between systemic risk and financial fragility is explored by Acemoglu et
al. (2015) and Bluhm and Krahnen (2014). Acemoglu et al. (2015) investigate how systemic
risk increases financial fragility by differentiating between complete6 and incomplete7 financial networks. Bluhm and Krahnen (2014) analyze features of systemic risk that emerge during the financial crisis.
Acemoglu et al. (2015) find that a small magnitude shock will be absorbed in the complete financial network as the losses are shared over all the counterparties, hence the excess liquidity from non-distressed firms can flow to distressed firms. Additionally, a shock of a large magnitude will cause more financial distress in sectors that are more interconnected, because of the diverse lending, financial contagion will create a fragile system (Acemoglu et al., 2015). Bluhm and Krahnen (2014) found that systemic risk affects financial institutions through two channels, direct and indirect. Direct is through balance sheets, where financial institutions are connected through assets and liabilities, and indirect through non-liquid asset fire sales, such as portfolio correlations. They find that the direct channel is a dominant driver of systemic risk (Bluhm and Krahnen, 2014).
Noteworthy is the presence of the contagion channel. The channels identified amplify the exogenous shocks, through the contagion channel, this was also concluded by López-Espinosa
et al. (2013) and Acemoglu et al. (2015). These two channels increase the overall financial
fragility of an economy because they transmit the risks among institutions and increase the risk of the whole financial system (Bluhm and Krahnen, 2014).
Financial contagion can cause and worsen financial fragility, as explored by Manz (2010). He found that the failure of a firm can trigger a chain reaction of failures of other firms when investors realize these firms are all based on similar fundamentals within the same industry. Thus, fragility is increased in an economic system because information contagion takes place (Manz, 2010). Alongside information contagion, bank default probability, correlation, and the number of overexposed banks in the system increase the fragility of an economy (Martínez-Jaramillo et al., 2010).
2.5. Financial innovation and financial fragility
Financial innovation has increased the productivity at which we are able to work. Innovations as the credit card and ATMs have made that possible. However, financial
5 Schwarz (2008) defined systemic risk as an “economic shock or institutional failure [that] causes a chain reaction
of economic consequences, including market failures. These consequences could cause significant losses to financial institutions or substantial financial-market price volatility. The consequences of the economic shock impact financial institutions, the markets or both” (Schwarz, 2008).
6 Complete financial network is defined as; “where all the liabilities of each institution are held equally by all other
banks”, implying that the system is fully integrated and the risk is spread among all banks
8
innovation still brings certain risks towards the economy. As seen in the financial crisis of 2008, where collateral debt obligations – as a new financial innovation – severely increased the significance of the crisis. Consequently, financial fragility is exacerbated by financial innovation (Gennaioli et al., 2012). The demand for safer claims is higher than the supply, which stimulates financial intermediaries to manufacture “new claims out of risky cash flows that are perceived to be equally safe” (Gennaioli et al., 2010). Safer claims will be purchased more when the ex-ante belief is that claims are safe to invest in. Thus, financial innovation increases the fragility of the economy as the pressure to create new safe claims causes financial intermediaries to disregard certain risks. When risk are disregarded the assets are perceived as safe, while they are actually riskier assets, with a higher probability of default.
An example of this was the collateral debt obligations, where lower rated mortgages were packaged together and sold as a high rated mortgage. Individually these are not safe assets, because they carry a higher probability of default but they are sold as a package. This distinguishes them to be safer and less risky than they actually are, due to the fact that the low rated mortgages were packaged together with some high rated mortgages. This will eventually cause fire sales – once the risky debt materializes- to be larger (Gennaioli et al., 2010 & 2012; Shleifer and Vishny, 2010). The fire sales are enhanced by shadow banks and regular banks as investors demand more liquidity, however, the creation of excess liquidity increases the systemic risk due to increased leverage (Gennaioli et al., 2012).
2.6. Determinants of severity of a crisis
The depth and severity of the crisis have been explored by multiple prior scholars. Common determinants found among researchers are; credit growth, pre-crisis GDP growth, the current account, and share of mortgage credit growth. Cecchetti et al. (2011) analyzed the economic performance of countries prior to and during the crisis and saw that there was a large variation in a country's ability to recover. They found that certain characteristics made an economy more crisis resistant, such as a capitalized banking sector, lower loan to deposit ratio, a current account surplus, high foreign exchange reserves and low levels of growth rates of credit. Hence policy decisions affected an economy’s vulnerability to the crisis (Cecchetti et al., 2011). The level of vulnerability towards changes in determinants of the severity of a crisis has implications to its overall financial fragility. Being more vulnerable implies that the effects of the crisis had a larger impact. If the policy decisions undertaken to make a country less vulnerable were neglected the country would be more fragile once a crisis occurred and suffer larger losses. Interestingly, this could suggest that more fragile countries would potentially suffer deeper recessions. This is reiterated by Berkmen et al. (2012) who similarly found that countries with more leveraged domestic financial systems, high credit growth and more short-term debt were more vulnerable to the impact of the crisis and saw a larger impact on economic activity.
9
found that enterprise credit stimulates economic growth whereas household credit did not (Beck
et al., 2012). However, whether total credit growth, as a solitary debt measure, was a significant
determinant for the severity of the crisis was questioned by Zhang and Bezemer (2015). Zhang and Bezemer (2015) questioned whether credit growth caused the severity of the post-crisis recession. Using data from 51 countries from 2003/2005 until 2012 they measured the depth, duration, cumulative costs and output loss of the post-crisis recession. They concluded that mortgage credit growth has a significant effect on the severity of the post-crisis recession. They analyzed the effect of mortgage credit growth while controlling for other explanatory variables, such as total credit growth. Interestingly, the severity of the crisis was caused primarily due to an increase in mortgage credit growth rather than total credit growth. Hence, they attributed the post-crisis recession to the increase in mortgage share growth and not total credit growth (Zhang & Bezemer, 2015).
2.7. Measuring the severity of post-crisis recession
There are various ways to measure the severity of a crisis. Zhang and Bezemer (2015) use indicators of duration, depth, growth loss and cumulative costs. Zhang and Bezemer (2015) base their measurement of the depth on Feldkircher (2014), stating that the “depth is the change of quarterly real GDP from peak to trough” (Zhang and Bezemer, 2015; Feldkircher, 2014). They measure the duration as “the number of quarters from peak to trough during the period of 2007-2012” and measure the cumulative costs as the depth multiplied by the duration divided by two.
Previously mentioned scholars, Lane and Milesi-Ferretti (2011) look at the fall in output, consumption and domestic demand in 2008, whereas Berkmen et al. (2012) take the output growth in 2009 and compared it to the pre-crisis forecasts for that year, which was also used by Zhang and Bezemer (2015). They defined growth loss as the “forecasted GDP growth minus actual growth in 2009”. This would indicate how much the crisis caused a deviation in forecasted growth. Cecchetti et al. (2011) measured the severity of the crisis by analyzing the post-crisis cumulative GDP gap, measured as real GDP growth in 2008-2009.
2.8. Effect of crisis in advanced and emerging economies
If the severity of the post-crisis recession differed per country was investigated by Lane and Milesi-Ferretti (2011), Didier et al. (2012) and Ahrend et al. (2012). Lane and Milesi-Ferretti (2011) concluded that advanced economies realized larger output losses and suffered more post-crisis than emerging economies. A potential explanation as to why advanced economies fared worse after the crisis was that advanced economies had a larger external debt share, whereas those of emerging economies were much smaller (Ahrend et al., 2012). This links to financial fragility as countries which were less integrated fared better during the crisis. Less external debt and reduced vulnerability to the contagion educed financial shock were decreased, hence they were more financially stable.
10
advanced economies, however, the emerging economies converge to their pre-crisis growth rate at a faster trend compared to advanced economies.
2.9. Summary and Hypothesis
From the literature review it is apparent that financial fragility has received quite some academic attention. Existing literature by the scholars Aspachs et al. (2007), Martínez-Jaramillo
et al. (2010) and Manz (2010) identified factors that contribute to the financial fragility of a
country. The main ones are low profitability, capital requirements and financial contagion. In addition to which factors contribute the financial fragility of a country, scholars such as Jappelli
et al. (2013) and Ahrend et al. (2012) in turn, investigated how to reduce the financial fragility
of a country. However, there is a gap in the literature, and this gap is going to be explored throughout this paper. This paper differentiates itself by exploring the potential negative effects of a shock for an already fragile economy, specifically focusing on the global financial crisis of 2008. Additionally, this paper adds on to the current literature by analyzing which bank variables, such as bank capitalization and return on assets contribute to the financial fragility of a country.
This paper uses the methodology presented in Zhang and Bezemer’s (2015) article. However, this differs from their work, as their work focuses on the effect credit composition has on growth loss, whereas this paper examines the extent to which financial fragility impacted growth loss, and how bank variables, such as bank capitalization and return on assets contribute to financial fragility.
Hence I hypothesize that:
H1: Countries that were more fragile based on the national aggregated financial fragility score,
suffered larger growth losses. Additionally, growth losses also increased due to credit growth, and current account deficits.
11
3. Methodology
In this section I explain the methodology of the paper. The aim of this paper is twofold. The first objective is to analyze the effect financial fragility had on the post-crisis recession severity, in section 3.1. Second objective is to investigate how individual bank variables, such as bank capitalization and return on assets impact financial fragility, in section 3.2.
The reasoning behind doing two analyses is to get a better understanding about financial fragility, what effect it has and how it is affected. The first analysis investigates the extent to which financial fragility impacts the post-crisis recession severity. Once the impact has been established, the second analysis investigates which bank variables impact financial fragility. The two analyses do not directly relate to each other, implying that they are separate empirical models.
Andrianova et al. (2015) published A New International Database on Financial Fragility, which is a panel dataset that compiled data on six8 different types of deposit taking institutions, ranging from commercial banks to investment banks, for 124 countries over a period of 14 years from 1998 until 2012. The database included a national aggregated financial fragility score, for each country annually. This national aggregated financial fragility score (z-score) will be the variable used to represent financial fragility in section 3.1.
The database also provides bank variables such as; bank capitalization, managerial inefficiency, return on assets, liquidity, illiquidity, and risk exposure. The effect of these variables on financial fragility are explored in section 3.2. The measure for financial fragility in the empirical model in section 3.2 is impaired loans. A different measure for financial fragility is used compared to section 3.1. This is because the z-score is constructed using the bank variables, hence using the z-score would mean that the dependent variable is defined by the independent variables. Therefore financial fragility is measured as impaired loans in section 3.2.
3.1. Empirical model: growth loss
To be able to test the effect of financial fragility on the post-crisis recession severity, this paper mimics the methodology of Zhang and Bezemer (2015). This is the preferred methodology, because the post-crisis recession severity is measured as GDP growth loss. Growth loss is used as the measure for the post-crisis recession severity, because the forecasted GDP values take growth determinants into account which are independent of a crisis occurring. Additionally, the independent variables used in their empirical model were chosen based on an extensive literature review. The independent variables included in their model significantly correlates in multiple papers to the recession severity. The dependent variable is growth loss, the independent variables are; financial fragility, credit growth, pre-crisis growth, GDP per capita, current account, trade openness, financial openness, deregulation and exchange rate regime.
8 The 6 types of deposit taking institutions are; commercial banks, investment banks, cooperative banks, Islamic
12
Growth loss is defined as the forecasted GDP growth minus actual GDP growth in 2009.
The forecasted values are taken from the IMF’s World economic outlook of April 2008, prior to the collapse of the Lehman Brothers. For example, in the Netherlands the forecasted GDP growth for 2009 was 1.55%, while the actual GDP growth of 2009 was -3.98%. Hence, the growth loss is 1.55%-(-3.98%) = 5.54%.
Financial fragility is measured as the z-score from Andrianova et al. (2015) in 2007. The
value for 2007 is used because it measures the fragility prior to the crisis. The z-score for each country is defined as follows;
𝑍𝑗𝑡 =
𝑅𝑂𝐴𝐴𝑗𝑡+ 𝑒𝑞𝑢𝑖𝑡𝑦𝑎𝑠𝑠𝑒𝑡𝑠𝑗𝑡 𝑗𝑡
𝜎𝑅𝑂𝐴𝐴𝑗
ROAA stands for the Return on Average Assets, which is a weighted average of a banks’ annual return, in country j, in year t. Equity is the total value of bank equity, and assets is the total value of bank assets. 𝜎 is the standard deviation of ROAA over time in country j. The z-score is measured in this particular way, because it allows to measure the risk of an aggregated banking sector. This score indicates the financial soundness of a country. To ease interpretation, the score is inverted. The fragility of a country increases as the z-score gets larger or less negative. For example, Argentina has a z-score of -3.1 while Australia has a z-score of -16, this means that Argentina is more fragile. This is because the z-score is computed at an aggregate national level so we can compare it between countries.
This is the preferred measure of financial fragility because it computes a score based on six various different types of deposit taking institutions. Only using a dataset that focuses on commercial banks can mislead the conclusions about the financial fragility. This is because investment, real estate and mortgage banks played a large role in the global financial crisis. Hence the advantage of using this measure is that it takes multiple deposit taking institutions into account. Additionally, the z-score can be used for countries where market based data is not available, hence it is compatible.
The causality between financial fragility and growth loss can be questioned, as an increase in financial fragility can cause larger growth losses but likewise increase in growth losses can cause an increase in the financial fragility of a country. To avoid problems of endogeneity, I used the value of financial fragility in 2007 (Berkmen et al., 2012). This way, financial fragility in 2007 may cause growth loss in 2009, but the reverse is not true.
Model (1) tests whether financial fragility prior to the crisis impacted the growth loss of country j. Using a dataset of 40 countries.
9𝐺𝑟𝑜𝑤𝑡ℎ 𝑙𝑜𝑠𝑠𝑗 = 𝛼 + 𝛽1𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑓𝑟𝑎𝑔𝑖𝑙𝑖𝑡𝑦𝑗+ 𝜀 (1)
9 Note: The variables present in equations (1), (2), (3), (4), (5) and (6) have only one observation, for country j.
13
The literature review concludes that credit growth played a large role in the severity of the crisis (Berkmen et al., 2012; Frankel and Saravelos, 2012; Ólafsson and Pétursson, 2010). Therefore, this explanatory variable is used in each model henceforth.
Credit growth is defined as the annual average change in the GDP to total credit ratio from 2003
until 2007. This period is chosen because it was when there was intense competition between securitizers and the worst performing loans were created (Simkovic, 2013).
In model (2), credit growth and financial fragility are the independent variables. This is done to test whether financial fragility prior to the crisis impacted the growth loss of a country while specifying for increases in credit growth.
𝐺𝑟𝑜𝑤𝑡ℎ 𝑙𝑜𝑠𝑠𝑗 = 𝛼 + 𝛽1𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝐹𝑟𝑎𝑔𝑖𝑙𝑖𝑡𝑦𝑗+ 𝛽2𝐶𝑟𝑒𝑑𝑖𝑡 𝐺𝑟𝑜𝑤𝑡ℎ𝑗 + 𝜀 (2)
In model (3), the control variables GDP per capita and pre-crisis growth are added. They are added because these variables were identified to increase the severity of the crisis too (Lane and Milesi-Ferretti; 2011 and Feldkircher, 2014).
Pre-crisis growth is defined as the annual average change in GDP from 2003 until 2007. This
period was chosen for the same reasoning as credit growth. GDP per capita is defined as the real GDP per capita in 2008, because it captures the GDP prior to the recession in 2009. 𝐺𝑟𝑜𝑤𝑡ℎ 𝑙𝑜𝑠𝑠𝑗 = 𝛼 + 𝛽1𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝐹𝑟𝑎𝑔𝑖𝑙𝑖𝑡𝑦𝑗+ 𝛽2𝐶𝑟𝑒𝑑𝑖𝑡 𝐺𝑟𝑜𝑤𝑡ℎ𝑗 +
𝛽3𝑃𝑟𝑒 𝑐𝑟𝑖𝑠𝑖𝑠 𝑔𝑟𝑜𝑤𝑡ℎ𝑗 + 𝛽4𝐺𝐷𝑃 𝑝𝑒𝑟 𝑐𝑎𝑝𝑖𝑡𝑎𝑗 + 𝜀 (3)
Additionally, more control variables are added to model (3). These control variables are added because they significantly correlated to the severity of the crisis in multiple prior papers, identified by Zhang and Bezemer (2015). The control variables are current account (4), trade
openness (5), financial openness (5), exchange rate regime (6), and credit deregulation (6). Current account is measured as the current account as a percentage of GDP, where larger
numbers indicate surpluses. This variable is added to the model because it has been established that current account deficits can lead to increased growth losses, as identified by Claessens et
al. (2010).
𝐺𝑟𝑜𝑤𝑡ℎ 𝑙𝑜𝑠𝑠𝑗 = 𝛼 + 𝛽1𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝐹𝑟𝑎𝑔𝑖𝑙𝑖𝑡𝑦𝑗+ 𝛽2𝐶𝑟𝑒𝑑𝑖𝑡 𝐺𝑟𝑜𝑤𝑡ℎ𝑗 +
𝛽3𝑃𝑟𝑒 𝑐𝑟𝑖𝑠𝑖𝑠 𝑔𝑟𝑜𝑤𝑡ℎ𝑗 + 𝛽4𝐺𝐷𝑃 𝑝𝑒𝑟 𝑐𝑎𝑝𝑖𝑡𝑎𝑗 + 𝛽5𝐶𝑢𝑟𝑟𝑒𝑛𝑡 𝑎𝑐𝑐𝑜𝑢𝑛𝑡𝑗 + 𝜀 (4) Trade Openness is defined as the sum of imports and exports divided by GDP. This variable is
14
Financial openness is defined as the net foreign assets divided by GDP. This variable is added
because the shocks negatively affected wealth and lead to a decrease in capital around the world, hence this can potentially explain increases in growth loss (Berkmen et al. 2012).
𝐺𝑟𝑜𝑤𝑡ℎ 𝑙𝑜𝑠𝑠𝑗 = 𝛼 + 𝛽1𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝐹𝑟𝑎𝑔𝑖𝑙𝑖𝑡𝑦𝑗+ 𝛽2𝐶𝑟𝑒𝑑𝑖𝑡 𝐺𝑟𝑜𝑤𝑡ℎ𝑗 + 𝛽3𝑃𝑟𝑒 𝑐𝑟𝑖𝑠𝑖𝑠 𝑔𝑟𝑜𝑤𝑡ℎ𝑗 + 𝛽4𝐺𝐷𝑃 𝑝𝑒𝑟 𝑐𝑎𝑝𝑖𝑡𝑎𝑗 + 𝛽5𝑇𝑟𝑎𝑑𝑒 𝑂𝑝𝑒𝑛𝑒𝑠𝑠𝑗 +
𝛽6𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑂𝑝𝑒𝑛𝑒𝑠𝑠𝑗 + 𝜀 (5)
Exchange rate regime takes a value of 1 to 4. Where 1 indicates pegged exchange rate, 2
indicates pre-announced peg in a band narrower than +/- 2%, 3 indicates pre-announced peg in a band larger than +/- 2% and 4 indicates floating exchange rate. Credit deregulation is defined as the change in credit market deregulations, this is the percentage change of credit market deregulation score between 2007 and 2000, where higher scores indicates less credit regulation. 𝐺𝑟𝑜𝑤𝑡ℎ 𝑙𝑜𝑠𝑠𝑗 = 𝛼 + 𝛽1𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝐹𝑟𝑎𝑔𝑖𝑙𝑖𝑡𝑦𝑗+ 𝛽2𝐶𝑟𝑒𝑑𝑖𝑡 𝐺𝑟𝑜𝑤𝑡ℎ𝑗 +
𝛽3𝑃𝑟𝑒 𝑐𝑟𝑖𝑠𝑖𝑠 𝑔𝑟𝑜𝑤𝑡ℎ𝑗 + 𝛽4𝐺𝐷𝑃 𝑝𝑒𝑟 𝑐𝑎𝑝𝑖𝑡𝑎𝑗 + 𝛽5𝐸𝑥𝑐ℎ𝑎𝑛𝑔𝑒 𝑟𝑎𝑡𝑒 𝑟𝑒𝑔𝑖𝑚𝑒𝑗 +
𝛽6𝐶𝑟𝑒𝑑𝑖𝑡 𝑟𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑗 + 𝜀 (6)
3.2. Empirical model: bank variables and financial fragility
The second aim of this paper is to investigate how individual bank variables impact the financial fragility measured as impaired loans. Financial fragility is defined as the weighted average of impaired loans over gross loans, in country j, in year t. This measure of financial fragility is used because Rinaldi & Sanchis-Arellano (2006) and Iftikhar (2015) have identified it to be an appropriate measure of financial fragility. The z-score is not used as the dependent measure for financial fragility because it is defined by some of the independent variables used in this model, such as return on assets.
The bank variables are retrieved from the databased published by Andrianova et al. (2015). The variables are bank capitalization, managerial inefficiency, return on average assets, illiquidity, liquidity and risk exposure. These variables were chosen based on the CAMELS rating system. The CAMELS rating system is a rating system used by supervisory banks, such as the FED, to measure a banks’ probability of failure (Caton, 1997). Each variable corresponds to a component of the acronym, except for asset quality. The CAMELS acronym stands for capital adequacy, asset quality, management, earnings, liquidity, and sensitivity to market risk. Prior to 1997, sensitivity to market risk was not included in the rating system, therefore I make a distinction that the core variables do not include the variable that falls under the sensitivity to risk component.
15 The core variables are;
Bank variable Definition10 CAMELS rating system component
Bank capitalization 𝐸𝑞𝑢𝑖𝑡𝑦 𝑇𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠 Capital adequacy (C) Managerial inefficiency 𝐶𝑜𝑠𝑡𝑠 𝐼𝑛𝑐𝑜𝑚𝑒 Management (M) Return on average assets (ROAA) 𝑁𝑒𝑡 𝑖𝑛𝑐𝑜𝑚𝑒
𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠
Earnings (E)
Illiquidity 𝑁𝑒𝑡 𝑙𝑜𝑎𝑛𝑠
𝑇𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠
Liquidity (L)
Additional variables included;
Risk exposure 𝑁𝑒𝑡 𝑐ℎ𝑎𝑟𝑔𝑒 𝑜𝑓𝑓𝑠
𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑔𝑟𝑜𝑠𝑠 𝑙𝑜𝑎𝑛𝑠
Sensitivity to market risk (S)
Liquidity 𝐿𝑖𝑞𝑢𝑖𝑑 𝑎𝑠𝑠𝑒𝑡𝑠
𝑇𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠
Liquidity (L)
In model (7), the core variables are tested with regard to financial fragility (impaired loans), using panel dataset over a period of 14 years (1998-2012), for country j, in year t. See appendix 2 for list of countries and Appendix 4 for dataset sample.
𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑓𝑟𝑎𝑔𝑖𝑙𝑖𝑡𝑦𝑗𝑡 = 𝛼 + 𝛽1𝐵𝑎𝑛𝑘 𝑐𝑎𝑝𝑖𝑡𝑎𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛𝑗𝑡+
𝛽2𝑀𝑎𝑛𝑎𝑔𝑒𝑟𝑖𝑎𝑙 𝑖𝑛𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦𝑗𝑡 + 𝛽3𝑅𝑂𝐴𝐴𝑗𝑡 + 𝛽4𝐼𝑙𝑙𝑖𝑞𝑢𝑖𝑑𝑖𝑡𝑦𝑗𝑡 + 𝜀 (7) Additionally in model (8), the core variable illiquidity is replaced with liquidity. This is done because there are two variables present in the dataset published by Andrianova et al. (2015), which assesses the (L)11 component of the CAMELS rating. The difference between liquidity and illiquidity is that measurement for liquidity implies that as the ratio of liquid assets over total assets increases, a bank see an increase in liquid assets, hence it able to convert liquid assets to meet financial obligation. Whereas as illiquidity is measured as net loans over total assets, an increase in this ratio, implies that a bank has less liquidity to finance debt obligations, hence this is inversely related to liquidity. There is a negative correlation between illiquidity and liquidity, they are replaced to see if this impacts the model.
𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑓𝑟𝑎𝑔𝑖𝑙𝑖𝑡𝑦𝑗𝑡 = 𝛼 + 𝛽1𝐵𝑎𝑛𝑘 𝑐𝑎𝑝𝑖𝑡𝑎𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛𝑗𝑡+
𝛽2𝑀𝑎𝑛𝑎𝑔𝑒𝑟𝑖𝑎𝑙 𝑖𝑛𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦𝑗𝑡 + 𝛽3𝑅𝑂𝐴𝐴𝑗𝑡 + 𝛽4𝐿𝑖𝑞𝑢𝑖𝑑𝑖𝑡𝑦𝑗𝑡 + 𝜀 (8) In model (9), risk exposure is taken into account. Risk exposure is not identified as a core variable because sensitivity to risk (S) was added to the CAMELS rating system in 1997.
10 All measures are weighted averages based on individual banks.
11 The (L) component stands for liquidity in the CAMELS rating system, but since there are variables named
16
𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑓𝑟𝑎𝑔𝑖𝑙𝑖𝑡𝑦𝑗𝑡 = 𝛼 + 𝛽1𝐵𝑎𝑛𝑘 𝑐𝑎𝑝𝑖𝑡𝑎𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛𝑗𝑡+
𝛽2𝑀𝑎𝑛𝑎𝑔𝑒𝑟𝑖𝑎𝑙 𝑖𝑛𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦𝑗𝑡 + 𝛽3𝑅𝑂𝐴𝐴𝑗𝑡 + 𝛽4𝐼𝑙𝑙𝑖𝑞𝑢𝑖𝑑𝑖𝑡𝑦𝑗𝑡 + 𝛽5𝑅𝑖𝑠𝑘 𝑒𝑥𝑝𝑜𝑠𝑢𝑟𝑒𝑗𝑡 + 𝜀
(9) Ahrend et al. (2012) stated that tax systems that favor equity over debt finance reduce the financial fragility of an economy. The debt to equity ratios of financial corporations will be the variable used to represent the tax systems in place. The debt to equity ratio is used to represent the tax system in place because the tax treatment between debt and equity financing differs and favors debt financing. This increases the reliance of companies on debt financing (European commission, 2015). To reflect different debt to equity ratios, this paper chose three ratios, which highlight the debt to equity ratio per country. The ratios are; between 1&3, 3&4, and >4. This can, in turn, reflect the tax system in place. Smaller ratios as such can reflect lower difference in tax treatment, while larger ratios can show larger difference in tax treatment. This is examined in the robustness test of this model. The data is reduced to 33 countries, due to data availability.
𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑓𝑟𝑎𝑔𝑖𝑙𝑖𝑡𝑦𝑗𝑡 = 𝛼 + 𝛽1𝐵𝑎𝑛𝑘 𝑐𝑎𝑝𝑖𝑡𝑎𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛𝑗𝑡 +
𝛽2𝑀𝑎𝑛𝑎𝑔𝑒𝑟𝑖𝑎𝑙 𝑖𝑛𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦𝑗𝑡 + 𝛽3𝑅𝑂𝐴𝐴𝑗𝑡 + 𝛽4𝐼𝑙𝑙𝑖𝑞𝑢𝑖𝑑𝑖𝑡𝑦𝑗𝑡 + 𝛽5𝐷𝑒𝑏𝑡 𝑡𝑜 𝑒𝑞𝑢𝑖𝑡𝑦𝑗𝑡 + 𝜀
(A)
4. Data
This section describes the datasets used in the two models specified in the last section. Section 4.1 describes the data used in the growth loss model specified in methodology section 3.1. Section 4.2 describes the data of the bank variable model specified in section 3.2. Additionally, the data is also checked for diagnostic problems for both models and their data.
4.1. Growth loss
The sample of countries used in the model is based on data availability. The sample contains data on 40 countries. The full list of countries can be found in Appendix 1.
Growth loss is the dependent variable. The data used to construct the variable comes from
the IMF World Economic Outlook. The value for the forecasted GDP growth comes from the April 2008 edition, prior to the collapse of the Lehman Brothers. The actual GDP growth value is taken from the April 2009 edition.
Credit growth is the annual average change in the credit to GDP ratio from 2003 until 2007.
The total credit to GDP ratio is downloaded from the Bank of International Settlements.
Financial fragility data is represented by the z-score from Andrianova et al. (2015) in 2007.
It is inverted to allow for easier interpretation, and formatted (divided by 100) so that it is compatible with the rest of the data.
Pre-crisis growth is the annual average change in GDP from 2003 until 2007. The data is
17
GDP per capita is the GDP per capita in 2008 in constant (2010$) US dollars, and the GDP
per capita is divided by 10.000, so the mean GDP per capita is 0.312, which is $31,213.47. The GDP values are taken from the World Bank.
Financial Openness measures the net foreign assets divided by GDP in 2008. Source was
the WorldBank.
Current Account is measured the sum of net exports of goods and services, net primary
income, and net secondary income, as a percentage of GDP in 2008. Source is the IMF.
Credit deregulation is constructed by taking the percentage change in the credit market
regulation score from 2000 to 2007. The credit market rating is a score between 0-10, where higher scores means less regulation. The source of the data is the EFW Index report 2018.
Exchange rate regime takes a value between 1 and 4, where 1 indicates pegged exchange
rate, 2 indicates pre-announced peg in a band narrower than +/- 2%, 3 indicates pre-announced peg in a band larger than +/- 2%, and 4 indicators floating exchange rate, in 2007. Source: Ilzetzki et al. (2017).
The descriptive statistics for this analysis are summarized in table 1. The descriptive statistics indicate that all countries within the dataset experienced growth loss. The least was 0.7% and the largest was 14.2%, with a standard deviation of 2.7%. The least fragile country recorded an inverted z score of -50, while the most fragile has a score of -0.9. The standard deviation is 11. For example, the z-score for Australia was -16. See Appendix 3 for sample of the dataset.
Table 1: Descriptive statistics
Variable Observations Mean
Std.
Dev. Min Max
Growth loss 40 0.055 0.027 0.007 0.142
Credit growth 40 0.035 0.054 -0.078 0.196
Z score (Financial Fragility) 40 -0.172 0.110 -0.500 -0.009
GDP per capita 40 0.312 0.217 0.012 0.909
Pre-crisis growth 40 0.041 0.024 0.009 0.119
Current account 40 0.002 0.072 -0.145 0.169
Financial openness 40 0.188 0.400 -0.447 2.035
Trade openness 40 0.983 0.820 0.273 4.416
Exchange rate regime 40 2.275 1.062 1.000 4.000
Deregulation 40 0.044 0.141 -0.130 0.660
4.1.1. Diagnostic tests growth loss model
18
Multicollinearity
To test for Multicollinearity the variance inflation factor (VIF) is used and the coefficient correlation matrix is checked. The mean VIF is 1.99 which indicates there is no multicollinearity. Additionally, the matrix shows that there is only one high correlation between trade openness and financial openness of 0.719, which is to be expected, as they both measure a degree of openness (see Appendix 5, for multicollinearity matrix).
Heteroscedasticity
There could be heteroscedasticity in the data. To check whether this is the case, the Breusch-Pagan test and white test were conducted. In both cases this paper fails to reject the null hypothesis that there is homoscedasticity, so standard errors are used, instead of robust.
Normality
To test the assumption of normality, the Jarque-Bera test was used. We fail to reject the null hypothesis of normality because the p-value is 0.4795. The data is normal, and not affected by non-normality.
Endogeneity
To check if we are allowed to use normal OLS over instrument variables regression, the data is checked for endogeneity. The Durbin score, and the Wu-hausman score show that we fail to reject that the variables are exogenous. Implying that there is no endogeneity.
4.2. Bank variables and financial fragility model
The sample contains a panel dataset on 121 countries from 1998 until 2012. The full list of countries can be found in Appendix 2. The dataset published by Andrianova et al. (2015) is used in this model.
The bank variables in the database from Andrianova et al. (2015) are variables which can be found on BankScope. Bankscope provides a code for each measure, for example, the Bankscope code for equity is 2055, and for total assets is 2060. This is a form of notation, the full composition of equity (2055) and all the other variables can be found on Bankscope. These variables are also national data constructed as weighted averages, using weights based on individual banks’ total asset values.
Financial fragility is the ratio of impaired loans (2170) to gross loans (2000+2070). Bank capitalization is the ratio of equity (2055) to total assets (2060).
Managerial Inefficiency is the ratio of costs (2090) to income (2080+2085).
Return on average assets is the ratio of net income (2115) to average total assets (average
2025).
Illiquidity is the ratio of net loans (2000) to total assets (2025). Liquidity is the ratio of liquid assets (2075) to total assets (2025).
Risk exposure is the ratio of net charge offs (2150) to average gross loans (2000+2070).
19
Albania is 172% and the next year it is 42%. However, the max is 382% for Greece in 2001, but this is major outlier.
4.2.1. Diagnostic tests bank variables and financial fragility model
There may be factors in the data that influences the reliability and interpretation of the results, therefore I run several diagnostics checks.
Endogeneity
The dataset is declared as a panel dataset, to test whether the fixed effects model or the random effects model would be appropriate. The Hausman test concluded that the fixed effects model was more appropriate. This is because we reject the null hypothesis stating that the model has random effects, and therefore use a fixed effects model. This offers a solution to the endogeneity problem without using instrumental variables.
Heteroscedasticity
Heteroscedasticity is present in the model, hence robust standard errors are used in the fixed effects model. This was concluded by the Breusch-pagan test and white standard error test.
Multicollinearity
To test for Multicollinearity the variance inflation factor (VIF) is used and the coefficient correlation matrix is checked. The mean VIF is 1.37 which indicates there is no multicollinearity. Additionally, the matrix shows that there is only one high correlation between illiquidity and liquidity of -0.609 which is to be expected, they inversely relate to one another (see Appendix 6, for multicollinearity matrix).
Table 2: Descriptive statistics bank variables and financial fragility
Variable Observations Mean Std.
Dev.
Min Max
Financial fragility (impaired loans) 1,493 7.517 8.426 .030 103.288
Bank capitalization 1,782 9.765 6.191 -41.584 85.370
Managerial inefficiency 1,764 60.719 21.473 3.810 382.171
Return on average assets 1,779 1.340 2.567 -47.430 21.794
Illiquidity 1,781 49.648 15.114 2.360 92.400
Risk exposure 1,212 1.066 2.643 -16.361 31.475
20
5. Results
Section 5.1, provides the results for the extent to which financial fragility impacts the post-crisis recession severity. Section 5.2, show which bank variables impact financial fragility.
5.1. Growth loss model
Table 3 indicates that the coefficient for credit growth is statistically significant and has a positive correlation with growth loss, in models (2) to (5). This is in line with the literature, countries that experienced larger credit growth suffered more severe growth losses, as shown by Frankel and Saravelos, (2012); Berkmen et al. (2012); Ólafsson and Pétursson, (2010). However, Zhang and Bezemer (2015) concluded that the composition of credit growth had a causal relationship with growth loss, stating that mortgage credit share significantly impacted growth loss rather than total credit. Hence it should be noted that there are different types of credit, which can impact growth loss, such as mortgage credit, and the measure used in this model takes total credit growth into account.
The coefficient for financial fragility has a positive sign and is statistically significant in model (1), (2), (4) and (5). This shows that countries which were fragile prior to the crisis experienced larger growth losses. The concept that financial fragility increases losses is in line with the literature of Pesola (2011) who found that financial fragility increased bank loan losses. Likewise, in model (2) both coefficients are statistically significant implying that credit growth also increases growth loss while holding the financial fragility coefficient constant. From those results, we can conclude that both credit growth and financial fragility prior to the crisis increased the growth loss of a country.
Contradicting the literature the coefficient for current account positively correlates with growth loss at a significant level. Lane and Milesi-Ferretti (2011) and Claessens et al. (2010) concluded that current account deficits caused a more severe crisis, and the results from this paper indicate that countries that had a surplus suffered larger losses. However, the coefficient for financial fragility is statistically significant in model (4). This could be the case that countries that had a large surplus as a percentage of their GDP, made them more vulnerable to exogenous shocks. This is because countries were most severely hit by the crisis via their external account (Ahrend et al., 2012). Shocks that negatively impact an economy can spillover to other countries, as countries are integrated through the current and capital accounts, and through the contagion channel. The contagion channel played a pivotal role in increase the fragility because it transmits risk among institutions (Bluhm & Krahnen, 2014; Martínez-Jaramillo et al., 2010; Manz, 2010).
21
22 Table 3: Growth loss in 2009 and financial fragility
Growth loss (1) (2) (3) (4) (5) (6) Financial fragility 0.077* 0.067* 0.062 0.073** 0.065* 0.034 (0.039) (0.038) (0.039) (0.035) (0.038) (0.037) Credit growth 0.153** 0.149** 0.251*** 0.172** 0.075 (0.072) (0.074) (0.084) (0.073) (0.074) Pre-crisis growth -0.098 -0.283 -0.151 -0.150 (0.197) (0.188) (0.204) (0.183) GDP per capita -0.017 -0.044** -0.026 -0.009 (0.022) (0.021) (0.021) (0.021) Current account 0.200*** (0.062) Trade openness 0.011 (0.007) Financial openness 0.001 (0.016) Credit deregulation 0.088*** (0.031)
Exchange rate regime -0.004
(0.004)
Constant 0.068*** 0.061*** 0.069*** 0.083*** 0.063*** 0.071***
(0.008) (0.008) (0.015) (0.014) (0.015) (0.017)
Observations 40 40 40 40 40 40
Adjusted R2 0.072 0.149 0.116 0.292 0.184 0.246
23
5.2. Bank variable and financial fragility model
Table 4 shows that the bank capitalization coefficient has a significant negative correlation in models (1) and (3). The coefficient is negative because as banks' equity increases, by an increase in retained earnings or reserves, they have more capital and this increases the financial health of the bank as they are able to withstand more losses, hence less fragile. This was also concluded by Cecchetti et al. (2011), stating that an economy with more capitalized banks, meaning that a bank has sufficient assets to finance debt obligations made a country more resistant to a crisis. Additionally, Ahrend et al. (2012) concluded that equity finance decreases fragility compared to debt finance, accompanying the findings that when banks obtain more equity it reduces fragility. Similarly, Ahmad and Mazlan (2015) found that higher capital asset ratios reduce the fragility of banks.
The coefficients for managerial inefficiency are positive and significant for all models. This is in line with the literature of Ahmad and Mazlan (2015), Poghosyan and Cihák (2009) and Oshinsky & Olin (2006), but managerial inefficiency is only statistically significant for a limited number of model specifications. Ahmad and Mazlan (2015), concluded that increases in managerial quality decreased the fragility of domestic banks. However, this was not the case for foreign banks. Interestingly, Poghosyan and Cihák (2009) from the IMF, found that managerial inefficiency increased the financial fragility of banks, during the global financial crisis. Suggesting, that banks where managerial inefficiency was higher in 2008, were more fragile during the crisis. Similarly, Oshinsky & Olin (2006), found that increase in managerial inefficiency decreased the likelihood of bank failure, for problematic banks in 1991. However, the majority of their results conclude that managerial inefficiency has an insignificant effect.
In line with the literature of Fielding & Rewilak (2015), the coefficients of return on average assets are statistically significant and negative. This implies that as return on bank assets increases financial fragility decreases. This is because increasing return on assets means a bank is becoming relatively more profitable. This in turn makes them less fragile as they can withstand fluctuations in liquidity, after a shock occurs. Fielding & Rewilak (2015) concluded that average return on assets should be greater than 1.5% to withstand fluctuations, making a bank less fragile.
24
Table 4: Impaired loans (financial fragility) and bank variables
(1) (2) (3)
Bank capitalization -0.194* -0.200 -0.142*
(0.116) (0.124) (0.084) Managerial inefficiency 0.052** 0.065*** 0.056***
(0.020) (0.014) (0.020) Return on average assets -0.572*** -0.553*** -0.555***
(0.206) (0.206) (0.209) Illiquidity -0.172*** -0.168** (0.057) (0.0843) Liquidity 0.070* (0.041) Risk exposure 0.024 (0.329) Constant 15.71*** 4.328** 14.31*** (3.712) (1.889) (5.202) Observations 1,491 1,491 1,155 R2 0.129 0.101 0.147 Number of countries 121 121 104
Country FE YES YES YES
25
6. Robustness tests
6.1. Tax systems model (A)
Ahrend et al. (2012) concluded that tax systems that favor equity over debt financing are less fragile. To test the robustness, a measure for a countries tax system is added to the core indicators of financial fragility. Table 5 shows the effect a tax system dummy has on the financial fragility of a country while taking the core variables into account.
The dummy takes a value of 1, for when the debt to equity ratio of the financial corporations in the country is larger or equal to the boundary of the dummy variable. For example,
Year Country Debt to equity ratio Dummy variable: Debt to Equity ratio between 1 &3
Dummy variable: Debt to Equity Larger than 4
2009 Netherlands 2.26 1 0
Interestingly, the coefficient for debt to equity ratios between 1 and 3 is negative and statistically significant. This implies that as the debt to equity ratio decreases, financial fragility decreases. Additionally, debt to equity ratios larger than 4, increase financial fragility however the coefficient is not statistically significant but it is noteworthy. This can suggest that countries that have a tax system that favors debt over equity are more fragile, as corporations favor debt financing over equity, which would be in line with Ahrend et al. (2012).
Interestingly, table 5 shows the coefficient for return on average assets is negative and significant, in all models. This means that as the earnings of banks increase the financial fragility decreases (Fielding & Rewilak, 2015).
26
Table 5: Financial fragility and tax systems dummy
(1) (2) (3)
Bank capitalization -0.002 -0.011 -0.008
(-0.096) (-0.094) (-0.090) Managerial inefficiency -0.002 -0.005 -0.004
(-0.031) (-0.030) (-0.029) Return on average assets -1.252** -1.271** -1.243**
(-0.572) (-0.601) (-0.603)
Illiquidity 0.005 0.007 0.007
(-0.044) (-0.045) (-0.044) Debt to equity ratio between
1 and 3 -1.193*
(-0.636) Debt to equity ratio between
3 and 4 0.519
(-1.150) Debt to equity ratio larger
than 4 0.478 (-1.118) Constant 5.153 4.985 4.707 (-4.393) (-4.594) (-3.907) Observations 403 403 403 R-squared 0.084 0.079 0.079 Number of countries 33 33 33
Country FE YES YES YES
27
7. Conclusion
The aim of this paper is to investigate the extent to which financial fragility had an impact on the severity of the post-crisis recession of 2009. The financial fragility indicator that is used, was taken from Andrianova et al. (2015), who published a new international financial fragility database which contains a national aggregated financial fragility score. The post-crisis recession severity is measured by the growth loss of a country in the year 2009.
From the analysis in this paper, we can conclude that higher levels of financial fragility prior to the crisis increased the growth loss of economies. Alongside financial fragility, credit growth and current account as a percentage of GDP also increased the severity of the crisis. This is in line with the literature, which reiterated that in countries with higher credit growth the crisis was more severe. This paper is the first, to my knowledge, that empirically shows a positive correlation between financial fragility prior to the crisis and growth loss. This is in line with research that suggests the vulnerability of an economic system, contagion, and integrated systems are related to the severity and likelihood of a crisis (Ahrend et al., 2012, Didier et al., 2012, Manz, 2010, Martínez-Jaramillo et al., 2010).
Secondly, this paper uses the database to establish the relationship between bank variable financial fragility indicators, such as bank capitalization and managerial inefficiency, and financial fragility measured as impaired loans. The dataset contains panel data from 1998 until 2012, which includes data on 121 countries. To test the robustness of the model, a measurement for tax system was included. The results show that bank capitalization and return on average assets decrease financial fragility, while managerial inefficiency increases it. Increases in bank capitalization have a significant impact on decreasing financial fragility, which is in line with the literature from Cecchetti et al. (2011) and Ahrend et al. (2012). Additionally, tax systems that favor debt finance over equity finance are more fragile, in line with Ahrend et al. (2012).
Throughout the paper, there are data limitations. The financial fragility measures are not the same in both models in this paper. This is because the z-score is more sophisticated due to the fact that it was comprised of the bank variables. Therefore it could not be used in the second model, hence impaired loans was the financial fragility measure for the second model. Likewise, the dataset for the growth loss model is limited to 40 countries, which makes it a relatively small sample, an increase in the sample size would help the accuracy of the results. The countries used here were selected on the basis of the available information, so there is a selection bias, however, this is eliminated to the best of my ability.
28
8. References
Acemoglu, D., Ozdaglar, A., & Tahbaz-Salehi, A. (2015). Systemic risk and stability in financial networks. American Economic Review, 105(2), 564-608
Allen, F., & Gale, D. (2004). Financial fragility, liquidity, and asset prices. Journal of the
European Economic Association, 2(6), 1015-1048.
Andrianova, S., Baltagi, B., Beck, T., Demetriades, P., Fielding, D., Hall, S. G., Koch, S., Lensink, R., Rewilak, P. & Rousseau, P. (2015). A new international database on financial fragility. University of Leicester Economics Working Paper, 15, 18.
Ahmad, N., & Mazlan, N. F. (2015). Banking fragility sector index and determinants: a comparison between local-based and foreign-based commercial banks in Malaysia. International Journal of Business and Administrative Studies, 1(1), 5-17.
Ahrend, R., A. Goujard and C. Schwellnus (2012), "International Capital Mobility: Which Structural Policies Reduce Financial Fragility?", OECD Economic Policy Papers, No. 2, OECD Publishing, Paris, https://doi.org/10.1787/5k97gkcv5z27-en.
Aspachs, O., Goodhart, C. A., Tsomocos, D. P., & Zicchino, L. (2007). Towards a measure of financial fragility. Annals of finance, 3(1), 37-74.
Berger, D., & Pukthuanthong, K. (2016). Fragility, stress, and market returns. Journal of
Banking & Finance, 62, 152-163.
Beck, T., Büyükkarabacak, B., Rioja, F. K., Valev, N. T., 2012. Who gets the credit? and does it matter? Household vs. firm lending across countries. The B.E. Journal of Macroeconomics 12 (1), 1–46.
Berkmen, S. P., Gelos, G., Rennhack, R., Walsh, J. P., 2012. The global financial crisis: Explaining cross-country differences in the output impact. Journal of International Money and Finance 31 (1), 42–59.
Bluhm, M., & Krahnen, J. P. (2014). Systemic risk in an interconnected banking system with endogenous asset markets. Journal of Financial Stability, 13, 75-94.
Caton, W. F. (1997). Uniform Financial Institutions Rating System. Federal Register, 6. Cecchetti, S., King, M. R., Yetman, J., Aug. 2011. Weathering the financial crisis: good policy or good luck? BIS Working Papers 351, Bank for International Settlements.
Claessens, S., Dell’Ariccia, G., Igan, D., Laeven, L., 04 2010. Cross-country experiences and policy implications from the global financial crisis. Economic Policy 25, 267–293.
Degryse, H., Elahi, M. A., & Penas, M. F. (2013). Determinants of banking system fragility: A regional perspective.
Didier, T., Hevia, C., Schmukler, S. L., 2012. How resilient and countercyclical were emerging economies during the global financial crisis? Journal of International Money and Finance 31 (8), 2052 – 2077, policy Implications and Lessons from Global Financial Crisis.
European Commission, Green Paper Building a Capital Markets Union, Brussels 18 Feb
2015 COM(2015)63 final (CMU Green Paper) available from
http://ec.europa.eu/finance/consultations/2015/capital-markets-union/index_en.htm
