Bachelor Thesis Economics and Business
Specialization: Economics and Finance
Faculty of Economics and Business
Academic year: 2017-2018
International risk sharing in the Eurozone.
Empirical evidence on the Banking Union.
Student name: José Miguel Espí García
Student number: 11124784
Statement of Originality
This document is written by Student José Miguel Espí García who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
The aftermath of the Great Recession has highlighted the relevance of the international risk-sharing mechanisms available to smooth consumption in the case of asymmetric shocks among the EU. European Banking Union is a critical reform implemented to increase international risk sharing in the Eurozone. This paper will study the empirical evidence that would sustain undertaking such a fundamental policy. An output variance decomposition methodology has been employed to measure risk sharing in the credit channel. The results have confirmed an increase of risk sharing in the credit channel since the introduction of the first two regulatory pillars of the Banking Union. The further decomposition of the credit channel into government savings, corporate savings, and household savings allows drawing three conclusions. First, the whole bulk of risk sharing in the credit channel is achieved via government savings. Second, corporate savings have a negative impact in terms of risk sharing. Third, risk sharing achieved via household savings is nonsignificant, which points out the lack of credit markets integration and that need to implement the third pillar of the Banking Union.
Table of Content
1. Introduction... 4
2. Literature review ... 7
2.1.International risk-sharing channels ... 7
2.2.International risk-sharing literature in Europe ... 11
2.2.1.Before the Treaty on European Union ... 11
2.2.2.First and second stage of the Treaty on European Union ... 12
2.2.3.Third stage of the Treaty on European Union ... 13
2.2.4.The Great Recession ... 14
2.3.Hypothesis ... 15
3. Methodology ... 16
4. Dataset and sources... 23
5. Results and interpretation ... 25
5.1.Estimated risk sharing in the credit channel and its subcomponents... 25
5.2.Banking Union’s estimated effect in the credit channel and its subcomponents ... 27
5.3.Evolution of risk sharing in the credit channel and its subcomponents ... 29
6. Policy implication for the Eurozone ... 32
7. Discussion and limitations ... 35
8. Conclusion ... 36
References... 38
Appendix A ... 42
1. Introduction
European Banking Union is at the top of the EU political agenda and represents one of the most significant challenges and opportunities for the Eurozone. Such is the case that it has been at the center of debate and reforms conducted by our most influential European leaders. The five presidents' report (Juncker et al., 2015) on completing Europe's Economic and Monetary Union (EMU) stated that progress in the EMU should occur through four fronts. One of the fronts is the financial union, which encompasses the Banking Union and the Capital Markets Union. Jean-Claude Juncker, president of the European Commission, declared in a press release of the European Commission (2017a) "the need to encourage all Member States to join the Banking Union... to reduce the remaining risks in the (national) banking system". In the same press release, Valdis Dombrovskis, Vice-President of the European Commission in charge of FISMA (Financial Stability, Financial Services, and Capital Markets Union) stated that "a complete Banking Union is essential for the future of the Economic and Monetary Union and for a financial system that supports jobs and growth.”
The Banking Union is a standard regulation at the European level that aims to make the banking system more transparent, unified, safer, and to break the negative feedback loop between governments and national banking systems that arose during the Great Recession (European Central Bank, 2018; European Commission, 2017a). It is based on three pillars (Pagliacci, 2014): The Single Supervisory Mechanism (SSM), the Single Resolution Mechanism (SRM) and the Deposit Guarantee Schemes (DGS). The SSM shifts banking supervision from national central banks to the European Central Bank (ECB), avoiding national bias in supervision (European Central Bank, 2018; Pagliacci, 2014). The SRM ensures a communitarian management procedure in case of bank default (European Commission, 2017b). This management procedure is financed by the private sector to avoid the use of taxpayers' money (European Commission, 2017b). The DGS facilities a communitarian deposit
protection in the face of banking failure (European Commission, 2017c). The SSM and SRM are already functioning since the summer of 2013, while the DGS still has to be implemented (European Commission, 2018). The Banking Union regulation gained relevance in the aftermath of the Great Recession, which highlighted the limitations of the EMU and the need to conduct reforms at both national and communitarian levels (Juncker et al., 2015).
The limitations faced by the EMU were long established by Mundell (1996). The entrance of the 19 EU Member States to the Eurozone implied a de facto total loss of monetary policy independence, leaving Members States with no autonomy to address domestic shocks (European Commission, 2017d). The Optimal Currency Area (OCA) (Mundell, 1961) states that the EMU would be optimal if no welfare loss is derived from the loss of autonomous monetary policy. According to Mundell (1996), to assess if no welfare loss is generated and determine if the EMU is an OCA, two criteria are of relevance: the importance of asymmetric shocks and the existence of adjustment mechanisms. Concerning the first criteria, the recent financial and debt crisis of 2008 have highlighted the relevance of asymmetric shocks in the euro area and the difficulties of having Member States with different monetary policy needs but with one single Central Bank (Nikolov, 2016). The second criteria, the adjustment mechanisms, can take up both at a national level and international level. The adjustment mechanisms at a national level are limited by law. The Treaty on European Union (1992), also referred to as Treaty of Maastricht, and the Stability and Growth Pact (1997) have imposed several restrictions on domestic fiscal freedom (Eichengreen, 1997). The Maastricht convergence criteria (1992) establishes that Member States shall not run budget deficits higher than 3% of GDP and that Government debt shall not exceed 60% of GDP. Moreover, the preventive and a corrective arm of Stability and Growth Pact (1997) ensures the enforceability of measures to correct excessive spending.
Therefore, the EMU is not only characterized by relevant asymmetric shocks affecting Member States, but also by imposing restrictions that limit national fiscal policies to address local shocks. Consequently, it seems logical to emphasize the relevance of the international risk-sharing mechanisms that must take place to smooth consumption against asymmetric shocks for the well-functioning of the EMU. This is, in other words, the adjustment mechanism at a communitarian level mentioned by Mundell (1996). As highlighted in the five presidents' report (Juncker et al., 2015), to strengthen the EMU against unexpected shocks it necessary for Member States to undertake collective reforms. One of these reforms is the Banking Union, which is designed to increase international risk sharing in the Eurozone through the integration of financial markets (Juncker et al., 2015).
Thus, the following question arises: what is the empirical evidence that would sustain undertaking this fundamental policy in the Eurozone? In other words, is European Banking Union empirically justified from an international risk-sharing analysis perspective? More specifically, did the Introduction of the Single Supervisory Mechanism (SSM) and the Single Resolution Mechanism (SRM) increase international risk sharing in the Eurozone? To answer these questions, a model to measure risk sharing in the credit channel will be constructed based on literature review. The model will be empirically regressed to determine the change in risk sharing after the introduction of the first two pillars of the Banking Union.
This paper is organized as follows. The first section will review the international risk-sharing empirical literature in Europe. It will be followed by the methodology and dataset sections. Subsequently, the results will be presented and discussed. Moreover, the policy implications for the Eurozone will be discussed. Finally, the limitations and conclusions of this research will be analyzed.
2. Literature review
2.1. International risk-sharing channels
Risk sharing, as defined in Cimadomo, Hauptmeier, Palazzo, and Popov (2018) is the practice of insuring consumption from output fluctuations generated by the business cycle. On a perfect hypothetical insured risk sharing, the correlation between consumption and production would be zero (Cimadomo et al., 2018). This would mean that the level of consumption remains unchanged for any given production fluctuations and thus perfectly shielding consumption. In the Eurozone, as it can be observed in Figure 1, cross-country dispersion of consumption growth and GDP growth evolve closely together. This indicates that that perfect risk sharing is far from reality in the Eurozone.
Figure 1
Cross-country dispersion of output growth (GDP) and consumption growth (C).
Note. Both GDP and C are deflated, per capita and measured in standard deviation. The y-axis
According to Cimadomo, Furtuna, and Giuliodori (2017), Nikolov (2016), Poncela, Pericoli, Manca, and Nardo (2016), Asdrubali and Kim (2004) (henceforth known as AK) and Asdrubali, Sørensen, and Yosha (1996) (henceforth known as ASY), the four main international risk-sharing channels can be classified in two broad categories: private channels and public channels. All these studies also state that the private channel is composed of the capital market channel, the credit market channel and the cross-border labor compensation channel. On the other hand, the public channel is solely formed by the fiscal redistribution channel.
Following the methodology of ASY, based on the System of National Accounts, the channels can be identified from a variance decomposition of GDP. An example will be used to explain the functioning of each channel. It will be assumed that Germany suffers an unexpected shock that decreases its GDP, and meanwhile, Spain suffers another unexpected shock that increases its GDP. The asymmetry of the shocks between countries prevents the use of a common monetary policy as an effective response. Nonetheless, these unexpected shocks can be smoothed through the different international risk-sharing channels.
𝐺𝐷𝑃𝑡𝑖 = (𝐺𝐷𝑃
𝑡𝑖− 𝐺𝑁𝐼𝑡𝑖) + (𝐺𝑁𝐼𝑡𝑖 − 𝐺𝐷𝐼𝑡𝑖) + (𝐺𝐷𝐼𝑡𝑖− 𝐶𝑡𝑖) + 𝐶𝑡𝑖 (1)
The capital market channel or net factor income, (𝐺𝐷𝑃𝑡𝑖 − 𝐺𝑁𝐼𝑡𝑖), is the difference between gross domestic product and gross national income (Cimadomo et al., 2018). It captures transactions such as cross-border labor compensations and income transfers from foreign ownership of assets (Backus et al., 1992; Lewis, 1996; Poncela et al., 2016). For example, the income generated from German ownership of assets in Spain can smooth consumption against the negative output shock in Germany (ASY; Nikolov, 2016). On the other hand, cross-border labor compensations are what is known as the labor market channel. They account for
compensations to non-residents that live in the country less than a year (Poncela et al., 2016). Nikolov (2016) explains that the labor channel would insure against asymmetric output shocks if German residents would immigrate to Spain to seek additional stream revenues. In practice, income transfers from border asset ownership are very superior in size compared to cross-border labor compensations (Poncela et al., 2016). That is why in this paper the labor channel will be taken into account within the capital market channel and not as an independent channel. The fiscal channel or net international transfers, (𝐺𝑁𝐼𝑡𝑖 − 𝐺𝐷𝐼𝑡𝑖), is the difference between the gross national income and the gross disposable income. It captures transfers of tax streams between governments and transfers of remittances (Cimadomo et al., 2018; Poncela et
al., 2016). Continuing with the example, the transfer of tax revenues from Spain to Germany
would help to shield German consumption from the negative output shock.
The credit market channel or gross national savings, (𝐺𝐷𝐼𝑡𝑖 − 𝐶𝑡𝑖), is the difference between the gross disposable income and total consumption, which captures cross-border borrowing/lending from both private and public sectors (Cimadomo et al., 2018; Poncela et al., 2016). The credit market channel, as described by Nikolov (2016) and Asdrubali et al. (1996) can help insulate Germany’s consumption from the negative output shock by borrowing from the Spanish credit market. Borrowing and lending between countries are more likely to occur when, as argued in Sørensen and Yosha (1998) paper, the negative output shock is not persistent. Otherwise, the high credit risk will prevent Spain from lending to Germany. It is important to highlight that the credit channel does not include borrowing and lending within the national borders but only from cross-border credit markets and supranational organizations, so that inter-temporal smoothing of risk is not captured within the credit channel (Cimadomo
Furthermore, as explained by Milano (2017) and, Sørensen and Yosha (1998), the credit market channel or gross national savings can be decomposed into government gross savings (Equation 2) and private gross savings (Equation 3). The latter can be split up into corporate (Equation 5) and households gross savings (Equation 4) (Milano, 2017; Sørensen & Yosha, 1998). Hence, the credit channel can be separated into government savings, household savings and corporate savings (Equation 6). The three subdivisions of the credit channel are derived from the System of National Accounts and presented as follows:
𝐺𝑜𝑣. 𝐺𝑟𝑜𝑠𝑠 𝑆𝑎𝑣𝑖𝑛𝑔𝑠 𝑡𝑖(𝐺𝑆) = 𝐺𝑜𝑣. 𝐺𝑟𝑜𝑠𝑠 𝐷𝑖𝑠𝑝𝑜𝑠𝑎𝑏𝑙𝑒 𝐼𝑛𝑐𝑜𝑚𝑒 𝑡 𝑖(𝐺𝐼) − 𝐺𝑜𝑣. 𝑇𝑜𝑡𝑎𝑙 𝐶𝑜𝑛𝑠𝑢𝑚𝑝𝑡𝑖𝑜𝑛 𝑡 𝑖(𝐺𝐶) (2) 𝑃𝑟𝑖𝑣𝑎𝑡𝑒 𝐺𝑟𝑜𝑠𝑠 𝑆𝑎𝑣𝑖𝑛𝑔𝑠 (𝑃𝑆) 𝑡𝑖 = 𝐺𝑟𝑜𝑠𝑠 𝑁𝑎𝑡𝑖𝑜𝑛𝑎𝑙 𝑆𝑎𝑣𝑖𝑛𝑔𝑠 (𝐺𝑁𝑆) 𝑡 𝑖 − 𝐺𝑜𝑣. 𝐺𝑟𝑜𝑠𝑠 𝑆𝑎𝑣𝑖𝑛𝑔𝑠 (𝐺𝑆) 𝑡 𝑖 (3) 𝐻𝑜𝑢𝑠𝑒ℎ𝑜𝑙𝑑𝑠 𝐺𝑟𝑜𝑠𝑠 𝑆𝑎𝑣𝑖𝑛𝑔𝑠(𝐻𝑆)𝑡𝑖 = 𝐻𝑜𝑢𝑠. 𝐺𝑟𝑜𝑠𝑠 𝐷𝑖𝑠𝑝. 𝐼𝑛𝑐𝑜𝑚𝑒 (𝐻𝐼) 𝑡 𝑖 − 𝐻𝑜𝑢𝑠. 𝑇𝑜𝑡𝑎𝑙 𝐶𝑜𝑛𝑠𝑢𝑚𝑝𝑡𝑖𝑜𝑛 (𝐻𝐶) 𝑡 𝑖 (4) 𝐶𝑜𝑟𝑝𝑜𝑟𝑎𝑡𝑖𝑜𝑛𝑠 𝐺𝑟𝑜𝑠𝑠 𝑆𝑎𝑣𝑖𝑛𝑔𝑠 (𝐶𝑆)𝑡𝑖 = 𝑃𝑟𝑖𝑣𝑎𝑡𝑒 𝐺𝑟𝑜𝑠𝑠 𝑆𝑎𝑣𝑖𝑛𝑔𝑠 (𝑃𝑆) 𝑡 𝑖 − 𝐻𝑜𝑢𝑠𝑒ℎ𝑜𝑙𝑑𝑠 𝐺𝑟𝑜𝑠𝑠 𝑆𝑎𝑣𝑖𝑛𝑔𝑠(𝐻𝑆) 𝑡 𝑖 (5) 𝐺𝐷𝐼𝑡𝑖− 𝐶 𝑡𝑖 = (𝐺𝐼𝑡𝑖− 𝐺𝐶𝑡𝑖) + (𝐻𝐼𝑡𝑖− 𝐻𝐶𝑡𝑖) + [(𝐺𝐷𝐼𝑡𝑖− 𝐶𝑡𝑖) − (𝐺𝐼𝑡𝑖− 𝐺𝐶𝑡𝑖) − (𝐻𝐼𝑡𝑖− 𝐻𝐶𝑡𝑖)] (6)
The government savings channel, (𝐺𝐼𝑡𝑖− 𝐺𝐶𝑡𝑖), captures government’s borrowing and lending in international credit markets and supranational organizations (Milano, 2017). The latter includes loans and transfers from EU institutions (Milano, 2017). The household savings channel, (𝐻𝐼𝑡𝑖− 𝐻𝐶𝑡𝑖), captures borrowing and lending from households in the international credit markets (Milano, 2017). Similarly, the corporate savings channel, [(𝐺𝐷𝐼𝑡𝑖− 𝐶𝑡𝑖) − (𝐺𝐼𝑡𝑖−
𝐺𝐶𝑡𝑖) − (𝐻𝐼
𝑡𝑖− 𝐻𝐶𝑡𝑖)], captures corporate’s cross-border borrowing and lending (Milano, 2017).
The extent to which the private sector, corporations and individuals, can smooth their consumption by borrowing and lending in the international credit market, depends on the degree of credit markets integration (Sørensen & Yosha, 1998).
Finally, as explained in AK and Cimadomo et al. (2018) it is necessary to make a distinction between ex-ante and ex-post insurance channels. Ex-ante insurance channels absorb
risk sharing through agreements made before the unexpected shock has taken place, such as the capital market channel (AK; Cimadomo et al., 2018). Ex post insurance channels, such as the credit market and its subcomponents, capture risk sharing due to agreements made after the unexpected shock (AK; Cimadomo et al., 2018). The fiscal channel can be both an ex-ante or
ex-post depending on its set up (AK). It could be that a fiscal distribution takes place at a
predetermined situation, in which case it would be ex-ante. Alternatively, it could be a single transfer agreed after the shock, in which case it would be ex-post (AK).
2.2. International risk-sharing literature in Europe
The literature on international risk sharing is profoundly influenced by the work of both ASY and AK. ASY, grouped all risk-sharing channels in a single framework, based on a GDP variance decomposition. This methodology allows to measure the quantity of risk sharing smoothed through the different channels, and it has been replicated by multiple authors. Nevertheless, the ASY empirical methodology left some relevant problems untreated, such as output endogeneity and the dynamic effects between channels. AK tackled these problems by using the same variance decomposition approach but introducing a panel VAR methodology instead of the static ASY model. These two empirical approaches have been the most influential frameworks in the international risk-sharing empirical literature around the globe. The evolution of international risk sharing in Europe is not only influenced by these two frameworks but also by the history of the European Union itself, where it can distinguish between the following periods.
2.2.1. Before the Treaty on European Union
Sørensen and Yosha (1998), applying the ASY empirical approach for the period 1966-1990, concluded that around 40% of output shocks in the European Community were absorbed
within one year. Half of it was due to government borrowing and the other half because of private borrowing. Not surprisingly, the absorption capacity through factor income was redeemed negligible and so it was the fiscal channel. The former fact was a reliable indicator of the lack of market integration in Europe that did not begin until 1990. The lack of market integration is also accredited by Kalemli-Ozcan, Sorensen, and Yosha (2004), who again using the ASY framework found that the capital channel smoothed only between 2% and -8% for the periods 1972-1982 and 1983-1992, respectively.
AK, using a panel VAR approach, studied the EU risk sharing for the period 1972-1990 in two sub-samples: before and after 1972. This division aimed to capture the end of the Bretton Woods and the implementation of the “snake in the tunnel” exchange rate regime (Verbeken & Rakić, 2018). AK results, in line with Sorensen and Yosha (2004), redeemed negligible the capital and the fiscal channel for both sub-samples. However, they concluded that there was a significant increase in the credit channel on the second sub-sample. These findings capture the market's developments in the 1980s, coinciding with the creation of the European Monetary System and the adoption of the Single Market (Verbeken & Rakić, 2018). They estimated that 43.4% and 23.6% of the GDP shocks were absorbed through the credit channels for impulse and cumulative responses, respectively.
2.2.2. First and second stage of the Treaty on European Union
The full liberalization of capital markets began the 1st of July of 1990 with the first stage of the Treaty on European Union (1992). From 1994 to 1998, the second stage took off aiming at the economic convergence between State Members. As is to be expected, these events had a substantial impact regarding international risk sharing. Indeed, the effects were captured by the research of Kalemli-Ozcan et al. (2004). They concluded that during the 1990s there was a very significant increase in risk sharing. Their research determined that a rose in foreign
ownership of assets generated a very significant increase of risk sharing in the capital market channel which rose to 9% for the period 1993-2000.
2.2.3. Third stage of the Treaty on European Union
The 1st of January of 1999 started the adoption of the euro, the third stage of the Treaty on European Union (1992). As argued by Cimadomo et al. (2018), the introduction of the single currency was expected to decrease transaction costs, enhance cross-border ownership of assets and thus increase the absorption capacity of the capital channel. Similarly, they argued that the elimination of currency risk would promote further credit market integration which would improve the credit channel. Contrary to what they expected, Cimadomo et al. (2018), using the AK framework, found that the credit market channel contribution was negative or inexistent for every year from 2007 to 2015. Moreover, Cimadomo et al. (2018) estimated for the same period that the fraction of unsmoothed shocks was between 80% and 60% and that the capital channel fluctuated between 15% and 40%. These results go in line with Cimadomo et al. (2017), where the risk sharing was estimated to smooth around 40% of the GDP shocks for the first years of the EMU.
Furthermore, Poncela et al. (2016) studied the period 1999-2914 using the AK methodology. In their research, the absorption capacity through the public channel was determined negligible. They also determined an increase in international risk sharing through the capital market channel and estimated that about 77% of shocks were unsmoothed. This result is superior to the previous 60% unsmoothed shocks determined by Sørensen and Yosha (1998). Hence, their research points out the fact that the unsmoothed fraction of output shocks is underestimated when the static model ASY is employed (Poncela et al., 2016).
Finally, Nikolov (2016), applying the ASY methodology, studied the period 2000-2015. He concluded that the absorption capacity of the capital, labor, and credit channel was
5.52%, 0.24%, and 18.15%, respectively. The fiscal channel was again redeemed negligible. Additionally, 75.74% of output shocks were unsmoothed, which is more than the results of Sørensen and Yosha (1998). The higher fraction of unsmoothed shocks could be explained by the use of the static ASY model. Alternatively, it could also be explained by the collapse in risk sharing in 2010 as it is argued in Kalemli-Ozcan, Luttini, and Sørensen (2014).
2.2.4. The Great Recession
In 2008, what started as a subprime mortgage financial crisis in the United States spread to European banks as the global interbank market collapsed (Ros & Spiegel, 2010). Since the SSM was not implemented, banking failure management remained a task for the national governments (Lane, 2012). This factor favored the negative feedback loop between the national baking systems and governments, and the financial crisis soon became also a sovereign debt crisis (Lane, 2012).
The financial and debt crisis of 2008 had a substantial impact on international risk sharing. Kalemli-Ozcan et al. (2014), using the ASY methodology, concluded that cross-border risk sharing collapsed in 2010 in Greece, Italy, Ireland, Portugal, and Spain as a direct consequence of the austerity programs carried out by national governments and their inability to borrow from the markets. Kalemli-Ozcan (2016) argued that for the recovery of risk sharing capacity after the 2008 crisis, the EU would benefit from implementing the Banking Union and from an increase in cross-border ownership of production assets. Indeed, the recent study by Cimadomo et al. (2017), showed that risk sharing in the Eurozone had increased from 40% at the beginning of the Eurozone to 60% after the crisis.
2.3. Hypothesis
The SSM and the SRM are expected to increase international risk sharing in the credit market channel. In Cimadomo et al. (2018) analysis, it is suggested that the European Banking Union is necessary for a positive absorption capacity of the credit channel. This argument is the line with Kalemli-Ozcan (2016) paper, where it is argued that the Banking Union would have a positive effect in terms of risk-sharing capacity.
One of the main reasons for low levels of international risk sharing in the Eurozone is the legal fragmentation between countries which causes less integrated credit markets (Kalemli-Ozcan, 2016). The Banking Union, by definition, is a standard regulation for the banking sector of the Eurozone, and thus it should increase the degree of credit market integration. As explained in Cimadomo et al. (2017) research, higher integrated credit markets can make credit supply more independent from idiosyncratic country shocks. They argue that international banks operating on the national credit market are less exposed to idiosyncratic national shocks than national banks. This allows to achieve a higher credit supply through the international credit market, and thus it enhances international risk sharing (Cimadomo et al., 2017). Therefore, the European Banking Union is expected to have a positive effect on the credit channel. Consequently, to answer the central question, the following hypotheses will be tested:
𝐻0: The fraction of GDP shocks smoothed through the credit channel after the implementation of the Banking Union equals the fraction of GDP shocks smoothed by the credit channel before the Banking Union.
𝐻1: The fraction of GDP shocks smoothed through the credit channel after the implementation of the Banking Union has increased when compared to the fraction of GDP shocks smoothed by the credit channel before the Banking Union.
3. Methodology
The econometric methodology developed by ASY, based on an output variance decomposition will be used to empirically assess the fraction of GDP shocks smoothed through the credit channel. From the system of National Accounts, Equation 7 can be formulated:
𝐺𝐷𝑃 =𝐺𝐷𝑃 𝐺𝑁𝐼 𝐺𝑁𝐼 𝐺𝐷𝐼 𝐺𝐷𝐼 𝐶 𝐶 (7)
As also explained in Sørensen and Yosha (1998) work, by taking logs, computing the log differences to obtain the growth rate of each of the variables, multiplying both sides by ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖, and taking expectations, Equation 8 can be obtained.
𝑣𝑎𝑟( ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) = 𝑐𝑜𝑣(∆𝑙𝑜𝑔𝐺𝐷𝑃 𝑡𝑖 − ∆𝑙𝑜𝑔𝐺𝑁𝐼𝑡𝑖, ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) (8) + 𝑐𝑜𝑣(∆𝑙𝑜𝑔𝐺𝑁𝐼𝑡𝑖 − ∆𝑙𝑜𝑔𝐺𝐷𝐼𝑡𝑖, ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) + 𝑐𝑜𝑣(∆𝑙𝑜𝑔𝐺𝐷𝐼𝑡𝑖 − ∆𝑙𝑜𝑔𝐶 𝑡𝑖, ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) + 𝑐𝑜𝑣(∆𝑙𝑜𝑔𝐶𝑡𝑖, ∆𝑙𝑜𝑔𝐺𝐷𝑃 𝑡𝑖)
If Equation 8 is divided by 𝑣𝑎𝑟( ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖), Equations 9, 10, 11 and 12 are easily obtained. The absorption capacity of the capital, fiscal and credit channels is measured by 𝛽𝑘, 𝛽𝑓, and 𝛽𝐶 respectively, and the unabsorbed fraction by 𝛽𝑢. Hence, the output shock is equal to the unabsorbed fraction plus the fraction smoothed through the capital, fiscal and credit channel (Equation 13). Nevertheless, this thesis will focus only on international risk sharing achieved through the credit channel, 𝛽𝐶,(Equation 11).
𝛽𝑘 = 𝑐𝑜𝑣(∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖 − ∆𝑙𝑜𝑔𝐺𝑁𝐼 𝑡𝑖, ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) 𝑣𝑎𝑟( ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) (9) 𝛽𝑓 = 𝑐𝑜𝑣(∆𝑙𝑜𝑔𝐺𝑁𝐼𝑡𝑖 − ∆𝑙𝑜𝑔𝐺𝐷𝐼 𝑡𝑖, ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) 𝑣𝑎𝑟( ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) (10) 𝛽𝐶 = 𝑐𝑜𝑣(∆𝑙𝑜𝑔𝐺𝐷𝐼𝑡𝑖 − ∆𝑙𝑜𝑔𝐶𝑡𝑖, ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) 𝑣𝑎𝑟( ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) (11) 𝛽𝑢 = 𝑐𝑜𝑣(∆𝑙𝑜𝑔𝐶𝑡𝑖, ∆𝑙𝑜𝑔𝐺𝐷𝑃 𝑡𝑖) 𝑣𝑎𝑟( ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) (12) 1 = 𝛽𝑘+ 𝛽𝑓+ 𝛽𝐶+ 𝛽𝑢 (13)
Equation 11 is precisely the slope of an ordinary least squares regression, where the independent variable is ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖 , and the dependent variable is ∆𝑙𝑜𝑔𝐺𝐷𝐼
𝑡𝑖 − ∆𝑙𝑜𝑔𝐶𝑡𝑖. Thus,
at practical terms, 𝛽𝐶 can be estimated as in Equation 14.
∆logGDIti- ∆logC t i = α𝐶+ μ𝐶𝑖 + λ 𝐶 𝑡 + β C∆logGDPt i + e C i,t (14)
Equation 14 measures the credit channel. The credit channel captures private and public borrowing and lending in international credit markets and EU institutions (Cimadomo et al., 2018; Milano, 2017). The functioning of 𝛽𝐶 can be derived from Equation 11. A positive value
of 𝛽𝐶 indicates a positive risk sharing contribution of the credit channel. This implies lending
when there is a positive output shock and borrowing when the output shock is negative (Cimadomo et al., 2018). For example, given a positive value of 𝛽𝐶, a negative output shock implies a decrease in gross national saving. Negative values of gross savings imply that consumption is higher than current gross disposable income and thus, the country is smoothing its consumption by borrowing from the international credit market. Moreover, a high value of
𝛽𝐶 implies that for small changes in GDP there is a high change in gross national savings. Therefore, the higher the 𝛽𝐶, the higher the fraction of output shock smoothed through the credit channel.
Additionally, 𝛽𝐶 can be further decomposed. Through Equation 6, it is known that gross national savings can be split up into gross government savings, gross corporate savings, and gross household savings. Based on this information, Equation 15 can be derived. Equation 16 is obtained by inserting Equation 15 in Equation 11.
𝐺𝐷𝐼 𝐶 = 𝐺𝐼 𝐺𝐶∗ 𝐻𝐼 𝐻𝐶∗ 𝐺𝐷𝐼 𝐶 𝐻𝐼 𝐻𝐶 ∗𝐺𝐶𝐺𝐼 (15) 𝛽𝐶 = 𝑐𝑜𝑣(∆𝑙𝑜𝑔𝐺𝐷𝐼𝑡𝑖 − ∆𝑙𝑜𝑔𝐶𝑡𝑖, ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) 𝑣𝑎𝑟( ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) = 𝑐𝑜𝑣 (∆𝑙𝑜𝑔 𝐺𝐷𝐼𝐶 𝑡 𝑖 , ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) 𝑣𝑎𝑟( ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) = 𝑐𝑜𝑣 (∆log 𝐺𝐼𝐺𝐶 ∗𝐻𝐶 ∗𝐻𝐼 𝐺𝐷𝐼 𝐶 𝐻𝐼 𝐻𝐶 ∗𝐺𝐶𝐺𝐼𝑡 𝑖 , ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) 𝑣𝑎𝑟( ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) = 𝑐𝑜𝑣 (∆𝑙𝑜𝑔 𝐺𝐼𝐺𝐶 𝑡 𝑖 + ∆𝑙𝑜𝑔 𝐻𝐼𝐻𝐶 𝑡 𝑖 + ∆𝑙𝑜𝑔 𝐺𝐷𝐼 𝐶 𝐻𝐼 𝐻𝐶 ∗𝐺𝐶𝐺𝐼𝑡 𝑖 , ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) 𝑣𝑎𝑟( ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) = 𝑐𝑜𝑣 (∆𝑙𝑜𝑔 𝐺𝐼𝐺𝐶 𝑡 𝑖 , ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) 𝑣𝑎𝑟( ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) + 𝑐𝑜𝑣 (∆𝑙𝑜𝑔 𝐻𝐼𝐻𝐶 𝑡 𝑖 , ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) 𝑣𝑎𝑟( ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) + 𝑐𝑜𝑣 (∆𝑙𝑜𝑔 𝐺𝐷𝐼 𝐶 𝐻𝐼 𝐻𝐶 ∗𝐺𝐶𝐺𝐼𝑡 𝑖 , ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) 𝑣𝑎𝑟( ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) (16)
Therefore, risk sharing of the credit channel, βC, can be split up into the risk sharing achieved via government savings, household savings, and corporate savings (Equation 17).
βC= βGS+ βHS+ βCS (17) βGS= 𝑐𝑜𝑣 (∆𝑙𝑜𝑔𝐺𝐶𝑡𝐺𝐼𝑖, ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) 𝑣𝑎𝑟( ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) = 𝑐𝑜𝑣(∆𝑙𝑜𝑔𝐺𝐼𝑡𝑖− ∆𝑙𝑜𝑔𝐺𝐶 𝑡𝑖, ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) 𝑣𝑎𝑟( ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) (18) βHS= 𝑐𝑜𝑣 (∆𝑙𝑜𝑔𝐻𝐶𝐻𝐼 𝑡 𝑖 , ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) 𝑣𝑎𝑟( ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) = 𝑐𝑜𝑣(∆𝑙𝑜𝑔𝐻𝐼𝑡𝑖− ∆𝑙𝑜𝑔𝐻𝐶 𝑡𝑖, ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) 𝑣𝑎𝑟( ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) (19) βCS= 𝑐𝑜𝑣 (∆𝑙𝑜𝑔 𝐺𝐷𝐼 𝐶 𝐻𝐼 𝐻𝐶 ∗𝐺𝐶𝐺𝐼𝑡 𝑖 , ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) 𝑣𝑎𝑟( ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) = 𝑐𝑜𝑣(∆𝑙𝑜𝑔𝐶𝑜𝑟𝑝𝑜𝑟𝑎𝑡𝑒 𝑆𝑎𝑣𝑖𝑛𝑔𝑠𝑡𝑖, ∆𝑙𝑜𝑔𝐺𝐷𝑃 𝑡𝑖)1 𝑣𝑎𝑟( ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) (20)
Equations 18, 19, and 20 are the slope of three ordinary least squares regressions where the independent variable is ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖, and the dependent variables are ∆𝑙𝑜𝑔∆𝑙𝑜𝑔𝐺𝐼𝑡𝑖 − ∆𝑙𝑜𝑔𝐺𝐶𝑡𝑖
𝑡 𝑖
, ∆𝑙𝑜𝑔𝐻𝐼𝑡𝑖− ∆𝑙𝑜𝑔𝐻𝐶𝑡𝑖, and ∆𝑙𝑜𝑔𝐶𝑜𝑟𝑝𝑜𝑟𝑎𝑡𝑒 𝑆𝑎𝑣𝑖𝑛𝑔𝑠𝑡𝑖, respectively. Thus, in practice, βGS, βCS, and βHS can be estimated through equations 21, 22, and 23.
∆𝑙𝑜𝑔𝐺𝐼𝑡𝑖 − ∆𝑙𝑜𝑔𝐺𝐶𝑡𝑖 = α𝐺𝑆+ μ𝐺𝑆𝑖 + λ𝐺𝑆𝑡 + βGS∆logGDPti + eGSi,t (21)
∆𝑙𝑜𝑔𝐻𝐼𝑡𝑖 − ∆𝑙𝑜𝑔𝐻𝐶𝑡𝑖 = α𝐻𝑆+ μ𝐻𝑆𝑖 +λ𝐻𝑆𝑡 + βHS∆logGDPti + eHSi,t (22)
∆𝑙𝑜𝑔𝐶𝑜𝑟𝑝𝑜𝑟𝑎𝑡𝑒 𝑆𝑎𝑣𝑖𝑛𝑔𝑠𝑡𝑖 = α𝐶𝑆+ μ𝐶𝑆𝑖 +λ𝑡𝐶𝑆 + βCS∆logGDPti + eCSi,t (23)
1 For simplicity the term 𝐺𝐷𝐼𝐶 𝐻𝐼 𝐻𝐶∗𝐺𝐶𝐺𝐼
Equation 21 measures the government savings channel. Hence, βGS measures the international risk sharing achieved through government borrowing and lending in international credit markets and EU institutions. Equation 22 and 23 capture the household savings channel and the corporate savings channel, respectively. Thus, βCSand βHS, are the fraction of GDP shocks smoothed through corporate and households’ borrowing/lending in the international credit market.
Finally, an interactive dummy variable has been added to Equations 14, 21, 22 and 23. The interactive dummy captures the effect of the Banking Union in the credit channels and its three subcomponents. Therefore, Equations 24, 25, 26 and 27 will also be regressed.
∆logGDIti- ∆logCti=α𝐶+ μ𝐶𝑖 + λ𝐶𝑡 + βC∆logGDPti + θCBU∆logGDPti+eCi,t (24)
∆logGIti- ∆logGCti=α𝐺𝑆+ μ𝐺𝑆𝑖 + λ 𝐺𝑆 𝑡 + β GS∆logGDPt i + θ GSBU∆logGDPti+eGSi,t (25)
∆logHIti- ∆logHCti=α𝐻𝑆+ μ𝐻𝑆𝑖 +λ𝑡𝐻𝑆+ βHS∆logGDPti + θHSBU∆logGDPti+eHS
i,t (26)
∆𝑙𝑜𝑔𝐶𝑜𝑟𝑝𝑜𝑟𝑎𝑡𝑒 𝑆𝑎𝑣𝑖𝑛𝑔𝑠𝑡𝑖=α
𝐶𝑆+ μ𝐶𝑆𝑖 +λ𝐶𝑆𝑡 + βCS∆logGDPti + θCSBU∆logGDPti + eCS
i,t (27)
In conclusion, Equations 14, 21, 22, 23, 24, 25, 26 and 27 will be estimated to answer the research question. Where, βC βGS, βCS and βHS represent the fraction of GDP shocks smoothed through the credit channel, government savings (GS), corporate savings (CS) and household savings(HS), respectively. BU is the dummy variable of the Banking Union. The first two pillars of the Banking Union were approved in the summer of 2013. Hence it was decided that, BU=1 for the years 2014-2017, and BU=0 for all the previous years. θC is the increase in international risk sharing in the credit channel since the introduction of the Banking Union, so that θC= β𝐶𝑝𝑜𝑠𝑡 𝐵𝑈− β𝐶𝑝𝑟𝑒 𝐵𝑈. Similarly, θGS, θCS, θHS captures the same effect but
for the government savings, corporate savings and household savings. Moreover, e𝑖,𝑡 is the error term and α the mean intercept, for each of the channels.
Additionally, the eight regressions include fixed time effects, μ𝑡, and fixed country effects, λ𝑖. Time-fixed effects capture specific yearly events that affect uniformly to all countries (Asdrubali et al., 1996). For instance, the time-fixed effect of 2008 will capture the repercussions of the financial crises in international risk sharing across the whole country set. On the other hand, country-fixed effects capture specific characteristics that are constant within countries through the entire period. For example, the country-fixed effects for peripheral and core countries of the Eurozone will differ capturing its credit market vulnerability and resilience, respectively (Cimadomo et al. 2017).
Time-fixed effects are a standardized practice in international risk sharing, and they have been used in all papers reviewed in this thesis with no exception, such as in Asdrubali et
al. (1996), Asdrubali and Kim (2004), Sørensen and Yosha (1998), and Kalemli-Ozcan et al.
(2014). Contrary, country-fixed effects are a less extended practice, yet they are also present in some studies as in Cimadomo et al. (2017), and Poncela et al. (2016). As it can be inferred from the literature review, the presence of both country and time effect are plausible from an economic theory point of view. In this research, to avoid omitted variable bias and following the usual practice in international risk-sharing literature, both time- and country-fixed effects are included.
In practice, the eight Equations to regress (14, 21, 22, 23, 24, 25, 26 and 27), will be estimated in the first place with a simple pooled ordinary least squares (OLS) corrected for robust cluster errors. Secondly, results will also be estimated with a generalized least squares (GLS) method. According to Nikolov (2016), GLS is a standardized econometric technique in international risk-sharing literature. GLS is more efficient than OLS for panel data with relative smaller cross-sections, T>N (Hepp & Von Hagen, 2012; Nikolov 2016). The time dimension
of the panel data of this study is indeed larger than the cross-sections, such that T>N. Note that Equation 17 only holds for OLS and not when using GLS.
In addition, also following the standardized econometric approach in international risk sharing, the residuals are corrected for heteroskedasticity and cross-sectional correlation. Moreover, the residuals are also allowed to follow an AR(1) process with a common coefficient to all countries because of the small sample size (Sørensen and Yosha, 1998). This is also the econometric approach followed in Asdrubali et al. (1996), Nikolov (2016), Poncela et al. (2016), and Sørensen and Yosha (1998).
In contrast with the standard practice, the pre-tests conducted in this research have revealed no evidence of the autocorrelation in the residuals for any of the regressions. Furthermore, the pre-tests also indicated that cross-sectional correlation is only present for equations 11 and 12. Finally, cross-sectional heteroskedasticity is present in all regressions to estimate. Due to the similarity of results between the standard econometric approach and the ones conducted following the pre-tests analysis, only the standardized approach is presented for reason of comparability. The pre-tests and the alternative estimations can be found in Appendix A and B respectively.
Finally, now that the models to estimate have been discussed, the hypothesis of this thesis can be formally rewritten as:
𝐻0: θi= 0
𝐻1: θi> 0
Where, θi= β𝑖𝑝𝑜𝑠𝑡 𝐵𝑈− β𝑖𝑝𝑟𝑒 𝐵𝑈, and i indicates each of the channels analyzed: credit
channel (i=C), government savings(i=GS), household savings(i=HS), and corporate savings (i=CS).
4. Dataset and sources
The data has been downloaded from the European Commission, Economic and Financial Affairs, AMECO database. The data retrieved covers the period of 2000-2017, at an annual basis, for 18 Members States of the Eurozone: Austria, Belgium, Cyprus, Estonia, Finland, France, Germany, Greece, Ireland, Italy, Latvia, Lithuania, Luxemburg, the Netherlands, Portugal, Slovakia, Slovenia and Spain. Malta is the only country of the Eurozone excluded due to the lack of information about private savings.
The process to elaborate the database has followed the procedure described at Poncela
et al. (2016). The following variables have been downloaded from AMECO: Gross Domestic
Product (GDP), Net Factor Income (NFI) Net International transfers (NIT), Total Consumption (C), Government Gross Disposable Income (GI), Government Final Consumption (GC), Households Gross Disposable Income (HI), Households Final Consumption (HC), Population (POP) and Harmonized Consumer Price Index (HCPI).
Based on the System of National Accounts, the variable gross disposable income has been constructed such that:
𝐺𝐷𝐼𝑡𝑖 = 𝐺𝐷𝑃
𝑡𝑖+ 𝑁𝐹𝐼𝑡𝑖 + 𝑁𝐼𝑇𝑡𝑖 (28)
The process followed to create the variables needed to estimate Equation 14 will be explained in detail hereafter. The same process will be repeated for the each of the variables needed to regress Equations 21, 22, 23, 24, 25, 26 and 27.
First, the nominal variables have been deflated using the Harmonized Consumer Price Index (HCPI) and expressed per capita diving by population. For Equation 14, the real per capita GDP, GDI, and C are calculated as follows:
After that, the yearly growth of each variable has been obtained by using log differences, such as:
𝑔𝑟𝑜𝑤𝑡ℎ 𝐺𝐷𝑃𝑡𝑖 = ln(𝑟𝑒𝑎𝑙 𝐺𝐷𝑃𝑡𝑖) − ln(𝑟𝑒𝑎𝑙 𝐺𝐷𝑃𝑡−1𝑖 ) (32)
𝑔𝑟𝑜𝑤𝑡ℎ 𝐺𝐷𝐼𝑡𝑖 = ln(𝑟𝑒𝑎𝑙 𝐺𝐷𝐼
𝑡𝑖) − ln(𝑟𝑒𝑎𝑙 𝐺𝐷𝐼𝑡−1𝑖 ) (33)
𝑔𝑟𝑜𝑤𝑡ℎ 𝐶𝑡𝑖 = ln(𝑟𝑒𝑎𝑙 𝐶
𝑡𝑖) − ln(𝑟𝑒𝑎𝑙 𝐶𝑡−1𝑖 ) (34)
Finally, the construction of the idiosyncratic shocks can simply be obtained by computing the difference between the national variable growth and the Eurozone average variable growth: ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖 = 𝑔𝑟𝑜𝑤𝑡ℎ 𝐺𝐷𝑃 𝑡𝑖 - 𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑔𝑟𝑜𝑤𝑡ℎ 𝐺𝐷𝑃𝑡𝐸𝑢𝑟𝑜𝑧𝑜𝑛𝑒 (35) ∆𝑙𝑜𝑔𝐺𝐷𝐼𝑡𝑖 = 𝑔𝑟𝑜𝑤𝑡ℎ 𝐺𝐷𝐼 𝑡𝑖 - 𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑔𝑟𝑜𝑤𝑡ℎ 𝐺𝐷𝐼𝑡𝐸𝑢𝑟𝑜𝑧𝑜𝑛𝑒 (36) ∆𝑙𝑜𝑔𝐶𝑡𝑖 = 𝑔𝑟𝑜𝑤𝑡ℎ 𝐶 𝑡𝑖 - 𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑔𝑟𝑜𝑤𝑡ℎ 𝐶𝑡𝐸𝑢𝑟𝑜𝑧𝑜𝑛𝑒 (37)
As recommended by Beyer, Doornik, and Hendry (2001) the average variable growth
for the Eurozone has been constructed with the weighted average of the cross-countries growth rates. The weights for each country are the relative size of their real GDP with respect to the real GDP of the Eurozone.
𝑅𝑒𝑎𝑙 𝐺𝐷𝑃𝑡𝑖𝑝𝑒𝑟 𝑐𝑎𝑝𝑖𝑡𝑎 = 𝑛𝑜𝑚𝑖𝑛𝑎𝑙 𝐺𝐷𝑃𝑡𝑖 𝐻𝐶𝑃𝐼𝑡𝑖 / 100 (29) 𝑅𝑒𝑎𝑙𝐺𝐷𝐼𝑡𝑖 = 𝑛𝑜𝑚𝑖𝑛𝑎𝑙 𝐺𝐷𝐼 𝐻𝐶𝑃𝐼𝑡𝑖 / 100 ∗ 1 𝑃𝑂𝑃𝑡𝑖 ∗ 1 𝑃𝑂𝑃𝑡𝑖 (30) 𝑅𝑒𝑎𝑙 𝐶𝑡𝑖 = 𝑛𝑜𝑚𝑖𝑛𝑎𝑙 𝐶 𝐻𝐶𝑃𝐼𝑡𝑖 / 100∗ 1 𝑃𝑂𝑃𝑡𝑖 (31)
𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑔𝑟𝑜𝑤𝑡ℎ 𝐺𝐷𝑃𝑡𝐸𝑢𝑟𝑜𝑧𝑜𝑛𝑒 = ∑ 𝑔𝑟𝑜𝑤𝑡ℎ 𝐺𝐷𝑃 𝑡𝑐𝑜𝑢𝑛𝑡𝑦 𝑖∗ 𝑊𝑒𝑖𝑔𝑡ℎ𝑡𝑐𝑜𝑢𝑛𝑡𝑟𝑦 𝑖 18 𝑖=1 (38) 𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑔𝑟𝑜𝑤𝑡ℎ 𝐺𝐷𝐼𝑡𝐸𝑢𝑟𝑜𝑧𝑜𝑛𝑒 = ∑ 𝑔𝑟𝑜𝑤𝑡ℎ 𝐺𝐷𝐼 𝑡𝑐𝑜𝑢𝑛𝑡𝑦 𝑖∗ 𝑊𝑒𝑖𝑔𝑡ℎ𝑡𝑐𝑜𝑢𝑛𝑡𝑟𝑦 𝑖 18 𝑖=1 (39) 𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑔𝑟𝑜𝑤𝑡ℎ 𝐶𝑡𝐸𝑢𝑟𝑜𝑧𝑜𝑛𝑒 = ∑ 𝑔𝑟𝑜𝑤𝑡ℎ 𝐶𝑡𝑐𝑜𝑢𝑛𝑡𝑦 𝑖 ∗ 𝑊𝑒𝑖𝑔𝑡ℎ𝑡𝑐𝑜𝑢𝑛𝑡𝑟𝑦 𝑖 18 𝑖=1 (40) 𝑊𝑒𝑖𝑔𝑡ℎ𝑡𝑐𝑜𝑢𝑛𝑡𝑟𝑦 𝑖 = 𝑟𝑒𝑎𝑙 𝐺𝐷𝑃𝑡𝑖 𝑡𝑜𝑡𝑎𝑙 𝑟𝑒𝑎𝑙 𝐺𝐷𝑃𝑡𝐸𝑢𝑟𝑜𝑧𝑜𝑛𝑒 (41)
5. Results and interpretation
5.1. Estimated risk sharing in the credit channel and its subcomponents
Table 1 displays the estimated percentage of GDP shocks that are smoothed through the credit channel. Furthermore, the risk sharing of the credit channel is decomposed into risk sharing attained via government savings, corporate savings, and household savings. The GLS estimation indicates that about 14% of GDP shocks in the Eurozone are smoothed through the credit channel. Recall that positive risk sharing implies lending when there is a positive output shock and borrowing when the output shock is negative (Cimadomo et al., 2018). Thus, about 14% of negative GDP shocks are smoothed through the credit channel by borrowing in international credit markets and/or EU institutions.
A richer analysis can be obtained by breaking down the credit channel into public, corporate and household savings. This analysis shows that the whole bulk of risk sharing in the credit channel is solely achieved through public savings. Governments smooth 44% of GDP shocks by borrowing/lending in international credit markets and/or EU institutions. This points out the fact that governments borrow about half the magnitude of negative output shocks to
maintain public consumption. In contrast, the corporate sector unsmooths 29% of GDP shocks. A negative absorption capacity indicates that corporations borrow during the economic boom of the business cycle and payback during the downturns (Cimadomo et al., 2018). Finally, the amount of risk sharing through household savings is small and not significantly different from zero. The extent to which international credit markets are integrated determines the access of households to international credit (Kalemli-Ozcan, 2016). The more international banks are located within national borders, the higher the ability of households to borrow or lend through cross-border markets. Hence, the non-significant risk sharing via household savings reflects a lack of international credit market integration at the individual’s level.
The econometric estimation from the OLS analysis gives relatively similar results. National governments smooth about 46% of GDP shocks through public savings meanwhile the corporate sector rather unsmooths about 28% of GDP shocks. Again, the consumption smoothing through household savings is found to be non-significant. However, there is a relevant distinction, the absorption capacity of the credit channel is redeemed statistically non-significant. This is probably due to the opposite effects of the public and corporate savings, which offset each other.
Table 1
Estimated coefficients of the credit channel (C) and its subcomponents: government savings (GS), corporate savings (CS) and household savings(HS)
OLS GLS βC 12,53 14,31*** (9,66) (1,35) βGS 45,77*** 44,40*** (11,28) (3,31) βCS -27,58** -29,07*** (14,23) (4,63) βHS -5,66 -4,62 (3,80) (1,64) Observations 378 378 Countries 18 18
Country FE YES YES
Time FE YES YES
Note. Standard errors are between brackets. The OLS estimation has been corrected for cluster
errors. The residuals in the GLS estimation have been corrected for heteroskedasticity, cross-sectional correlation and AR(1).
*p.1. **p.05. ***p.01.
5.2. Banking Union’s estimated effect in the credit channel and its
subcomponents
Table 2 displays the estimated change in the absorption capacity of the credit channel since the introduction of the European Banking Union. It also decomposes the change in risk sharing of the credit market channel into the changed attained via government savings, corporate savings, and household savings. The GSL estimation indicates that there has been a significant increase in international risk sharing within the credit channel since the introduction
of the SSM and SRM. The fraction of GDP shocks absorbed through the credit channel has increased by 47% since the introduction of the Banking Union regulation.
The Banking Union had an impact of diverse magnitudes for each of the savings channels. The public savings channel absorbed on average 13% more of every GDP shock after the introduction of the Banking Union with a 10% significance level. Stronger statistical evidence is found indicating that the corporate sector has benefited the most from the communitarian banking regulation. Since the introduction of the SSM and the SRM, the fraction of GDP shocks smoothed through the corporate savings has increased on average by 22%. These results reject the null hypothesis for the credit channel, government savings channel, and corporate savings channel. The null hypothesis stated no change in risk sharing after the implementation of the SSM and SRM. Therefore, it can be concluded that the Banking Union has increased credit markets integration among Eurozone members, increasing risk sharing through the credit channel and more specifically via government savings and corporate savings.
However, not surprisingly, the household savings channel has not recorded a significant improvement, and the null for household savings is not rejected. The results from the analysis seem to reflect the uncomplete structure of the Banking Union. The no implementation of the third pillar, the DGS, has left deposit insurance (household savings) as a matter of national regulation. The analysis from the OLS estimation indicates that there has been a significant increase of roughly 47%. However, no significant increase is captured for government, corporate and household savings. Therefore, according to the OLS estimation, the null hypothesis is rejected only for the credit channel. Thus, it can be concluded that there has been an improvement of risk sharing in the credit channel since the introduction of the Banking Union.
Table 2
Estimated coefficients of the interactive dummy variable of the credit channel (C) and its subcomponents: government savings (GS), corporate savings (CS) and household savings(HS)
OLS GLS θC 46,82** 47,29*** (9,62) (3,92) θGS 15,98 13,37* (15,29) (8,58) θCS 27,39 22,45** (24,86) (12,91) θHS 3,45 2,93 (7,04) (4,31) Observations 378 378 Countries 18 18
Country FE YES YES
Time FE YES YES
Note. Standard errors are between brackets. The OLS estimation has been corrected for cluster
errors. The residuals in the GLS estimation have been corrected for heteroskedasticity, cross-sectional correlation and AR(1).
*p.1. **p.05. ***p.01.
5.3. Evolution of risk sharing in the credit channel and its subcomponents
Figure 2 shows the evolution of risk sharing in the credit market channel for the period 2001-2017 using a rolling regression of 5 years. Two central facts can be derived from this graph. First, in 2008 coinciding with the financial and debt crisis, it can be observed the collapse of risk sharing in the credit channel in line with the results of Kalemli-Ozcan et al. (2014). Second, starting from 2013, the year of the approval of the SSM and the SRM, there is a very significant increase of international risk sharing in the credit channel. This confirms the expected positive effect of the Banking Union in the credit channel.
Figure 2
Evolution of risk sharing achieved via Gross National Savings (GNS) in the Eurozone
Note. Risk sharing is estimated using a GLS rolling regression of 5 years, including fixed time
and country effects. The residuals are corrected for heteroskedasticity and allowed to follow an AR(1) process The y-axis is the percentage of GDP shocks smoothed via GNS. The x-axis represents time.
Figure 3 decomposes the credit channel into its subcomponents, gross government savings (GN) and gross corporate savings (CS). Household savings have been omitted due to its small value and non-significance in figure 1. A key finding can be derived from this graph: there is an asymmetric response of government savings and corporate savings to the same stage of the business cycle. Taking 2008 as a reference point the following argument can be derived. The idiosyncratic GDP shock in 2008 was negative for the average of the Eurozone due to the financial crisis. Moreover, the risk sharing achieved via government savings took a positive
value in 2008. On the contrary, the 2008 risk sharing achieved via corporate savings had a negative value. The comparison of Equations 42 and 43 can provide a useful insight.
𝛽𝐺𝑁 = 𝑐𝑜𝑣(∆𝑙𝑜𝑔𝐺𝑜𝑣. 𝑆𝑎𝑣𝑖𝑛𝑔𝑠𝑡𝑖, ∆𝑙𝑜𝑔𝐺𝐷𝑃 𝑡𝑖) 𝑣𝑎𝑟( ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) (42) 𝛽𝐶𝑆 = 𝑐𝑜𝑣(∆𝑙𝑜𝑔𝐶𝑜𝑟𝑝. 𝑆𝑎𝑣𝑖𝑛𝑔𝑠𝑡𝑖, ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) 𝑣𝑎𝑟( ∆𝑙𝑜𝑔𝐺𝐷𝑃𝑡𝑖) (43)
The denominator of both equations is always the same. In 2008, when the shock is negative, 𝛽𝐺𝑁 took a positive value. Thus, from Equation 42 it must be the case that the covariance, cov(∆logGov.Savingsti, ∆logGDPti), is positive. A positive covariance implies that
government savings and GDP move in the same direction. This means that government savings decrease for a negative output shock. Thus, governments save less or spend more when there is a negative output shock, which helps to smooth consumption.
On the other hand, since 𝛽𝐶𝑆 took a negative value in 2008, it must be the case that 𝑐𝑜𝑣(∆𝑙𝑜𝑔𝐶𝑜𝑟𝑝. 𝑆𝑎𝑣𝑖𝑛𝑔𝑠𝑡𝑖, ∆𝑙𝑜𝑔𝐺𝐷𝑃
𝑡𝑖) is negative. This implies that corporate savings and the
output shock move in opposite directions. Thus, for a negative GDP shock, corporate savings increase, or corporations consume less. The increase in corporate savings indicates a stop in investment which further reduces consumption rather than smoothing it. Following similar reasoning, the opposite argument can be derived for a positive output shock. For a given positive output shock, government savings increase or consumption decreases. Conversely, for a given positive output shock, corporate savings decrease or corporate consumption increases.
To conclude, the asymmetric response of government savings and corporate savings to the same stage of the business cycle points out that, especially during economic downturns, risk sharing in the credit channel is uniquely achieved via government savings.
Figure 3
Evolution of risk sharing achieved via gross government savings (GN) and gross corporate savings (CS) in the Eurozone.
Note. Risk sharing is estimated using a GLS rolling regression of 5 years, including fixed time
and country effects. The residuals are corrected for heteroskedasticity and allowed to follow an AR(1) process. The y-axis is the percentage of GDP shocks smoothed via GN and CS. The x-axis represents time.
6. Policy implication for the Eurozone
Two policy implications can be derived from this study. First, the European Banking Union needs to be completed to increase and strengthen risk sharing attained through the credit channel. Second, since relying uniquely on the credit channel is unwise, further reforms must be conducted to enhance risk sharing through the capital and fiscal channel.
The facts derived from this research indicate that risk sharing through the credit channel in the Eurozone has significantly improved since the introduction of the Banking Union. However, it still is weak and incomplete. Risk sharing achieved through the credit channel is weak because it depends solely on the ability of national governments to borrow in international credit markets and/or EU institutions to smooth national consumption. Nonetheless, the ability of government to run deficits in the Eurozone is limited by the Stability and Growth Pact (1997). Thus, other international risk-sharing channels such as the capital or fiscal channel must be fostered to not rely solely on government borrowing for consumption smoothing.
As explained in the work of Gros (2016), the structure of the European Banking Union is incomplete due to two factors: the lack of deposit insurance regulation at a European level and the strong link between national governments and banks. The first issue is, in fact, the third pillar of the Banking Union, the DGS. The need to implement the DGS can be derived directly from the results of this research. The regulatory unification of the SSM and the SRM have positively contributed to the cohesion of credit markets and improved risk sharing. However, the lack of risk sharing via household savings points out to the fact that credit markets are still far from being perfectly integrated, especially at the individual’s level. The lack of risk sharing via household savings can be tackled with the implementation of the DGS. The DGS will guarantee the solvency of all deposits across Eurozone banks independently of the geolocation of the bank (Beck, 2016). To put it another way, the DGS ensures that the guarantee of deposits in case of bank failure does not ultimately depend on the solvency of national governments (Beck, 2016). This will create incentives for individuals to hold savings at cross-border depositary institutions and thus it will increase risk sharing via household savings. Therefore, the implementation of the DGS to complete the Banking Union must be a priority.
Concerning the second issue, the close link between banks and their national governments described by Gros (2016) is not directly observable from the results.
Nevertheless, it is equally problematic and essential to the conclusions of this research. The SSM and the SRM, are designed to prevent a contagion effect from the baking sector to national governments. Conversely, a contagion effect from national governments to the banking sector continues to be an issue. In case of a banking failure, the SRM ensures a communitarian management procedure independent from national governments and financed from the private sector (European Commission, 2017c). Hence, the SRM effectively prevents national governments from being responsible for bailing out defaulted banks, which would jeopardize national accounts and most likely increase the sovereign cost of borrowing (Gros, 2016). The SSM avoids national bias in supervision by shifting banking supervision from national central banks to the ECB (European Central Bank, 2018). As a consequence, the SSM and SRM are responsible for breaking part of the government-banks negative feedback loop and under this research, they are the primary cause for the increase in international risk sharing registered in the Eurozone since 2014. On the contrary, banks that hold significant amounts of debt of their national government could be affected by the health of the national accounts (Gros, 2016). In case of government failure, the loses from the national debt could wipe out their entire capital and put in jeopardy the whole banking system (Gros, 2016). The national concentration of debt held by banks needs to be diversified to further improve and strength international risk sharing through the credit channel (Gros, 2016).
Note that even if these problems were to be corrected, Banking Union would not be the solution to all risk-sharing problems (Kalemli-Ozcan, 2016). While it is true that the Banking Union can smooth out financial country-specific shocks, it can still lead to economic divergence (Kalemli-Ozcan, 2016). In the case of a financial country-specific shock, such as the default of a big bank, the credit supply does not to be interrupted because other Eurozone banks can channel funds to the country. However, in case of a real specific shock in the economy, such as a decrease in productivity, all foreign banks will stop the credit supply to
that specific country and will channel its fudns to countries where productivity is higher (Kalemli-Ozcan, 2016). Hence, this will generate a fall in the GDP and create economic divergence (Kalemli-Ozcan, 2016). Similarly, Cimadomo et al. (2017) argue that the Banking Union amplifies output shocks due to the pro-cyclicality of the credit supply. This in turn could lead to economic divergence (Cimadomo et al., 2017). Economic divergence does not necessarily have to be harmful if other channels are available to smooth out consumption (Kalemli-Ozcan, 2016). Thus, the arguments of Kalemli-Ozcan (2016) and Cimadomo et al. (2017) reinforce the need for enhancing alternative international risk-sharing channels, such as the capital or fiscal channel.
7. Discussion and limitations
The most important limitation encountered in this research is related to the methodology employed. The ASY static output variance decomposition assumes output to be exogenous and ignores channel dynamics. A panel VAR approach would have allowed for a more accurate measurement of the channels since it tackles the problems of output endogeneity and captures the dynamic effects between the channels. As pointed out by Poncela et al. (2016), the unsmoothed fraction of GDP shocks tends to be underestimated when employing the static ASY methodology. However, due to time limitations and the research scope of this thesis, the static ASY methodology had to be used.
Therefore, the repetition of this study employing a panel VAR would constitute interesting future research. The dynamic model will allow determining the effects of the European Banking Union in the capital channel, which was not included in this research. European Banking Union regulation should increase integration for both credit and capital markets. Hence, the hypothesis of this research would also apply to the capital market channel. In this thesis, a key finding has been made about the asymmetry response of government
savings and corporate savings to the same stage of the business cycle. A panel VAR would allow to study more in depth the dynamics between government and corporate savings and thus will represent compelling future research.
8. Conclusion
This research has investigated the level of risk sharing achieved among Eurozone members through the credit channel and particularly the risk sharing attained via each of its three subcomponents: government savings, corporate savings, and household savings. First of all, international risk sharing through the credit channel is low, absorbing about 14% of GDP shocks. Second, the decomposition of the credit channel allows identifying that the bulk of credit risk sharing is achieved via government savings which smooth about 44%. Third, corporate savings contribute negatively to risk sharing, unsmoothing about 29%. Fourth, risk sharing via household savings is small and nonsignificant.
Furthermore, this research has also studied the effect on risk sharing caused by the introduction of the first two regulatory pillars of the Banking Union. The SSM and the SRM had a positive impact on the credit market channel in terms of international risk sharing. However, the decomposition of the credit channel allows identifying that the improvement has mostly been felt via corporate savings, and to a lower extent at the government savings level. On the contrary, international risk sharing via household savings has not recorded a significant improvement.
Moreover, a key finding of this research is the asymmetric response of government savings and corporate savings to the same stage of the business cycle. For a positive stage of the business cycle, corporate savings decreases, and government savings increase. Conversely, for a negative stage of the business cycle, government savings decrease, and corporate savings increase.
The results obtained confirmed the expected positive effect of the SSM and SRM on the credit channel. Thus, it can be concluded that the Banking Union reforms are empirically justified from an international risk-sharing analysis. However, the analysis conducted have shown that the credit market channel still weak and incomplete. The lack of risk sharing via household savings evidences the uncomplete structure of the Banking Union. In addition, the decomposition of the credit channel and the asymmetric response of its subcomponents to an economic downturn prove that risk sharing is exclusively achieved via government savings. Finally, two main policy implications can be derived from the results. First, the Banking Union must be completed with the introduction of the DGS to stimulate risk sharing via household savings and further increase credit markets integration. Second, other international risk-sharing channels, such as the capital or fiscal channel, must be fostered to not rely exclusively on the ability of governments to borrow which is limited by law.
Finally, for future research, it is advised the repetition of this study employing a panel VAR to tackle the limitations of output endogeneity. A panel VAR approach would capture the substitutability or complementarity between channels, which would constitute a powerful insight for policy implication.
