What are banks paying/earning while trading-off profitability for liquidity Risk management? : a study on the effect of LCR implementation (Basel III) on banks’ performance

(1)

School of Economics and Business

MSc Business Economics, Finance track

Master Thesis

What are Banks Paying/Earning while Trading-off Profitability for Liquidity

Risk Management?

(A Study on the Effect of LCR Implementation (Basel III) on Banks’ Performance)

Xhenis Kapllani

11110708

July, 2016

Supervisor: Tanju Yorulmazer

(2)

Statement of Originality

This document is written by Xhenis Kapllani, who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is 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

.

(3)

Abstract

This study analyzes the implication of Liquidity Coverage Ratio introduction into bank’s liquidity framework. The research is performed for the time period 2013-2016 among 1684 and 1074 European and G20 banks, respectively. Research methodology involves a combination of conceptual frameworks, graphical representation and the causal relationship developed through the econometrics’ approach of panel data with fixed effects. Results reveal an overall adverse effect of LCR on performance, where this effect is more robust for G20 banks than for the European counterparties. Another important result is that of differences in the impact of LCR. The impact of LCR differs based on the funding sources of banks and construction of asset portfolio.

Keywords: LCR, effect, performance, liquidity framework, regulation, asset’s yield, funding

(4)

Introduction

The Recent financial crisis was the negative externality result of excessive leverage and lenient liquidity risk management. Externality, which through contagion effect spread assets’ fire sale and systemic risk across economies. Risk-taking behavior and incentives for excess returns in an economic environment, where systemic risk costs are not internalized by banks, call for macro-prudential policy and preemptive measures. Motivated by the liquidity risk management shortcomings, Basel Committee came with new rules on liquidity management in the Basel III Regulatory Framework.

On December 2010 Basel Committee published for the first time the Liquidity Coverage Ratio (LCR). LCR is expressed as the ratio between the High-Quality Liquid Assets (HQLA) and Net Cash Outflow (NO). Components1_{of HQLA include very liquid marketable securities such as sovereign bonds.} NO is expressed as the difference between cash outflows and inflows expected during a 30-day time frame, here including retail, wholesale and other sources of funding for financial intuitions.

LCR establishes a minimum liquidity requirement for financial institutions. On January 2013, the final revised draft was submitted and finally in January 2015, the official implementation of this ratio started. The objective of introducing this liquidity buffer was to endorse short-term resilience in bank’s liquidity risk BCBS (2013). LCR promotes liquidity by ensuring that a bank has enough high-quality liquid assets (HQLA) to cover liquidity shocks under a 30-day stress scenario, time which is assumed to be necessary to take corrective measures by bank’s management. The standard requirement is that the ratio between HQLA and NO remains above one, on an ongoing basis, unless in stress scenarios when in such a case it is allowed to fall below one. The implementation of this ratio started with a 60% requirement in 2015 and will be adjusted by 10% each year, until 2019, which Is the final implementation year.

As it was anticipated, LCR introduction has been followed by significant changes in the banking industry. Meeting the LCR intends simultaneous restructuring of the asset and funding portfolio. Regarding the asset portfolio, banks are phasing out investments on non-easy marketable securities, such as corporate/municipal bonds, swaps or residential securities. Part of this process is also the replacement of these less liquid assets with more active traded securities such as sovereigns’, central banks’ and multilateral development banks’ securities.

Different measures have also been introduced on the liability side of the balance sheet. Emphasis has been to significantly lessen the size of short-term funding which has a high run-away risk factor and is very volatile. Examples of these securities are uncommitted liquidity facilities, such as wholesale funding from corporates or credit institutions. JPMorgan Chase, HSBC, UBS and Credit Suisse were the first

1_{More in depth analysis of LCR components is given in Appendix A.1, Basel Committee Framework (2013), while in Methodology section is}

(6)

2

financial institutions to come public with the decision of entering into a selection process of the existing deposits. For instance, JPMorgan Chase announced that they would force out around 150 billion dollars in deposits, Jenkins (2016). Even though no specification was given to the type of deposits being left out, it was clear that they were no retail deposits, being considered as the least flighty ones due to the insurance scheme on them. Basel Committee estimated that for 2015, banks had a total of 22.2 trillion euros in corporate deposits Jenkins (2016), from which 16.6 million euros were non-operational deposits. These deposits are the ones that are penalized the most in LCR computation. Forcing out these specific deposits can be a provisionary solution, but not a definite one, as it can lead to a massive clientele loss and lack of diversity in funding sources. Banks have to find a middle ground solution, thus it has been claimed that they are introducing higher fees for deposits, which are considered as a burden in performance measures, but not liquidity requirement. On the other side, higher fees are also an enhancer in settling an equal playing field for banks. Before the financial crisis hit, more risk-averse banks were utilizing all types of deposits to speed up and increase the volume their lending. All the actions applied up to this point in time, discipline banks to internalize the costs of risk-taking behavior and lower the probability of systemic/systematic risk ex-ante. In other words, secure stability, sustainable financing, and returns in a fair competition environment.

Nevertheless, since the introduction of LCR, financial stakeholders have contradicted the expected positive spillovers of this liquidity buffer. One of the first concerns is related to the composition of the assets portfolio. Government and central bank securities are very safe and easily marketable but they come at the trade-off of lower returns for banks. Furthermore, in anticipation of a higher demand for such securities as HQLA, their yield is expected to go further down. This effect combined with more cash and reserves’ holdings, can put a downward pressure on assets’ returns. Nonetheless, one can expect that in markets where there is a shortfall in liquid assets, the central banks might facilitate the transition phase, such as the central bank of Australia and Denmark. In these countries, central banks promised to provide special treatment to covered bonds and provide special liquidity facility during the implementation phase of LCR Wellink, (2011). Lower yield on government bonds will put an upward pressure on the yield of other assets such as bonds that fall below investment grade or corporate/real-estate loans, meaning a more difficult task for banks to keep a diversified portfolio and a broad customer base.

Another criticism of LCR concerns the funding costs for banks. Competition for short-term retail deposits is expected to be reflected in their rates. However, banks do not necessarily need only increase the size of short-term stable funding, they can also scale down and re-shift its sources of financing to more long-term liabilities. On the other hand, this might be feasible for banks with a large network scale of customers and interbank relationships, but harder to be achieved for small-size, private or domestic-focused banks. Taking into consideration all the pros and cons is substantial that we observe the overall effect on

(7)

3

performance and understand if such liquidity buffer will help banks to have sustainable profits in the long-run. If one could generalize the overall effect of LCR, would the overall impact be a bonus or a burden to banks’ performance? In other words, what are banks paying/earning while trading off profitability for

liquidity risk management?

Being able to understand how banks are affected, will help to raise awareness and interest not only in between the banking circle but also boost the interest of regulators to facilitate and speed up the process of implementation. Assessing the impact of LCR is also important as it helps to visualize how sustainable is and will be the banking network and what kind of systemic risk/benefits imposes on the whole economy. A short time has passed since the first official endorsement of LCR, yet banks started with the informal implementation process as of 2013. From this date up to the first quarter of 2016, it is possible to analyze how banks are introducing this ratio to their liquidity management framework, are there any differences from expectations and most importantly what is the effect of this ratio on bank’s performance. Previous literature such as Härle et al. (2010) and Nwogugu (2014) base their conclusions on expectations of both financial stakeholders and regulators. Other studies such as (Banerjee & Mio 2015, De Haan & Van den End 2013, Duijm & Wierts 2014), predict an overall good effect of LCR such as in the Netherlands and the United Kingdom, which had already introduced a similar ratio to LCR. European Banking Authority and Basel Committee (EBA 2015, BCBS 2013, and BCBS 2016) have so far only released studies on balance sheet restructuring trends for a limited number of European banks.

So far, to the best of my knowledge, no study has taken the initiative to analyze the impact of this liquidity buffer on banks. Different to previous papers, which build on correlation matrixes or time-series analysis, this research paper will introduce a thorough econometrics investigation. More specifically, it will take into consideration interconnection of micro, macro and bank-specific factors while studying the causal effect of LCR on ROA, ROE, and Net Interest Margin separately. Another innovative contribution of this study stands in the fact that the impact analysis will be built on a diverse scale of countries and banks, further than EU and US. Let us not forget that the last financial crisis, was a global one. This implying that every bank’s liquidity condition is crucial in spreading either a positive or negative spillover across the entire financial network.

For this reason, the analysis of this paper is based on two different subgroups, namely G20 and EU countries (including Norway and Switzerland). We expect these countries to be more pro-active in the process of implementation. In the first part of the analysis, the focus will be to analyze: trends in implementation of LCR and development in balance sheet restructuring. Based on these foundations, the analysis will continue gauging the effect of LCR on bank’s performance, through a panel data analysis. Panel data helps to mitigate biases which result due to time-invariant differences across banks or countries. Due to limited information on liquid assets type and duration, the computation of LCR will be based on

(8)

4

approximation from S&P (Capital IQ) balance sheet data on a sample of approximately 1684 and 1074 EU and G20 banks. Complementary data on control variables, which are time-invariant across banks and countries will be retrieved from DataStream Thompson Reuters. The frequency of the data is on quarterly basis for the time period 2013-2016.

This paper is structured as follow. The first part will give an overview of the existing

studies on liquidity coverage ratio. The literature review will be complemented by the construction

of LCR and its computation in more detail and the theoretical background behind the choice of

control and dependent variables. The third and fourth section will introduce you to the paper’s

methodology, data, and descriptive statistics, respectively. Results and robustness check will be

introduced in section 5 and 6. Lastly, the paper will be finalized with a conclusion and discussion.

Literature Review

Even though the official implementation of LCR initiated only in January 2015, many banks started re-shifting balance sheet components since the official draft was submitted in January 2013. The banks that were pioneers in adhering to LCR, were G20 countries’ banks, according to BCBS (2013) and BCBS (2016). In a recent monitoring report from BCBS (2016), where a group of 92 randomly chosen banks were invited to participate in a monitoring exercise, results revealed that the average LCR was 144% and 140% by the end of December 2014 and June 2015, accordingly. One striking result was that the maximum cap for LCR was not 100% as expected, but 400% for some banks. There are two possible explanations for this results. One anticipated hypothesis of Perotti & Suarez (2011) is that LCR requirement was met before it was introduced. The other straightforward explanation is that national authorities are free to enquire higher LCR from banks when a higher market/liquidity risk is perceived from them.

Basel Committee and European Banking Authority (EBA) were the first institutions to analyze the first implications of LCR on bank’s balance sheet. In a monitoring report for a sample of 280 European banks, EBA (2015) concluded that HQLA makes up only 10% of total assets. Furthermore, the ratio of retail deposits to total liabilities is approximately 30% versus the ratio of non-secured wholesale funding to total liabilities which is 25%. In another monitoring experiment among a sample of 92 internationally active banks, BCBS (2016) pointed out that from all the components used in the computation of LCR, Level 1 Assets account for 82.3% of HQLA. More specifically 0% risk weight assets and cash/withdrawable Central Bank deposits account for 47.3% and 38% of HQLA. For the same sample, total outflows represented 21.1% of total liabilities versus 6% weight of total inflows, results which signal the high reliance of banks on wholesale funding. It is evident that wholesale funding is still widely used, despite the objective of Basel Committee to reduce the reliance on unsecured funding. Nevertheless, what we are missing from this

(9)

5

picture is a time development of the variables and ratios just mentioned. It is essential to have a dynamic view from the first moment of introduction up until today, in order to address conclusion on the progress of LCR implementation and the methods used.

Taking into consideration the previous examples and also the short time-frame since LCR was introduced, it is challenging to be more precise regarding the strategies implemented by banks in meeting the 100% standard of LCR. Existing literature is scarce and most importantly indecisive about a dominant method which overrules the others. Hartlage (2012) explains that when banks are confronted with three options in altering their capital, namely: 1) delivering (raising equity); 2) mismatch reduction (increasing the maturity of borrowing); 3) regulatory arbitrage (switching from high to low taxed borrowing), the majority of banks will opt for the third option. An example of the third option is direct banking, which is a form of remote banking, where banks compete directly on prices of retail deposits, while saving on infrastructure expenditures. The reason this is the most preferred option is the fact that the first option is perceived as a costly strategy since banks fear the high cost of issuing equity due to asymmetric information. The second option is both less feasible and costlier for average size and less creditworthy banks as they face difficulty in extending the maturity of their funding or switching to less risky but costlier financing. Nevertheless, one cannot use these conclusions as a generalization of the current situation. First central banks are subsidizing banks through special programs of covered bonds and liquidity facilities in order to exchange illiquid risky assets with government securities. In addition, not all the markets can be subject to a shortage in HQLA assets and not all the banks are exposed to the same way of financing. To corroborate on Hartlage (2012) supposition, it is substantial to evaluate the process of implementation of LCR in a longer-time frame by looking at the example of countries who already have implanted a ratio similar to LCR.

The LCR was inspired by the previous experience of the British Individual Liquidity Guidance (ILG), Dutch Liquidity Balance (LB) and South Korean Local Currency Liquidity Ratio (LCLR). These ratios hold similar characteristics to LCR, for example in the nominator (cash and marketable securities) and denominator (demand deposits). After the introduction of these ratios, Korea, UK and the Netherlands experienced different strategies and reactions from banks in meeting the liquidity requirement.

Banerjee & Mio (2015) present the effect of liquidity regulation across UK banks. Results showed that banks increased the size of high-quality liquid assets and simultaneously decreased intra-financial assets, while keeping the total assets intact. Regarding different methods of funding weight over total liabilities, banks shifted from wholesale funding to retail domestic deposits.

De Haan & Van den End (2013a) study the behavior of Dutch banks during the 2004-2010, period which coincides with an introduction and implementation of LB. Results point out a prudential liquidity risk management. Firstly, Dutch banks held more high-quality liquid assets than required and more

(10)

6

specifically they relied heavily on very liquid bonds. Secondly, even when the expected cash outflows were expected to drop, banks did not cut their holding of HQLA.

In contrast to the previous study, Duijm & Wierts (2014) found that Dutch banks adjustment on the liabilities side is more significant than on assets’ side. In the periods of high outflows, Dutch banks tended to increase the demand for stable deposits, which consequently enhanced competition. Dietrich et al. (2014) explain that customer deposits are less expensive than wholesale deposits when used to meet the liquidity requirements in periods of high demand.

The only negative remarkable experience was the one of Korea, where the pressure to meet the liquidity requirements, after the Asian Currency Crisis of 1997, backfired and led to instability. Banks took advantage in leveraging long-term mortgage investment on 1 and 2-year retail deposits. Being involved in a “snowball” borrowing process in the 1/2-year retail deposits market, brought a fierce competition and higher prices between Korean banks, leading to a spike in the spread between 1 year and overnight rates. Under these circumstances, the Financial Supervisory Board of Korea had to lower the LCLR requirement from 105% to 100% and later suspend this requirement in order to slow down the distortion in the Korean interbank market.

Nevertheless, the previous examples are a resemblance of country-specific experiences, based on the banking system characteristics of the respective country. A drawback of the previous results is the fact that we are dealing with developed countries. One still remains puzzled to observe the counter-reaction of banking system irrespective of the macro and microeconomic characteristics. More specifically would there be a difference on LCR effect across different countries or banks or would LCR implementation have the same implication? Are developing countries less prone to meet this standard?

Another important consideration is that the effect might differ also between banks. Based on banks’ characteristics, such as size, risk-taking incentives, credit opportunities, ownership status, market power or networking, the LCR implementation process can differ extensively. Perotti & Suarez (2011) take as example of bank characteristics credit opportunities and risk-taking behavior of banks. Authors discuss the effect of liquidity regulation, by analyzing how short-term financing in high credit growth can lead to negative externalities. Another point made by Perroti & Suarez (2011) is that in the case of differences in credit opportunities across banks, price rules such as taxes are more efficient in preventing risk and maintain credit quality. Quantitative restrictions such as LCR can be unfairly punitive as they might hamper growth and profitable opportunities for banks which are successful in liquidity management. Furthermore, LCR fails to control for procyclicality, meaning that banks can overstate performance in liquidity standards in good times while not being able to sustain this ratio in the long-term. On the other hand, BCBS (2010) support the experience of UK and the Netherlands, and strongly defend the main objective of LCR toward stability and lower probability bank failures from short-runs.

(11)

7

While (Banerjee & Mio 2015, De Haan & Van den End 2013, and Duijm & Wierts 2014) find evidence of re-scaling of balance sheet elements due to liquidity regulation, one should also take into account how the bank performance can possibly be affected.

As it was evidenced in the monitoring studies of EBA (2015) and BCBS (2016), banks primarily choose cash to lever up liquidity and sometimes even keep excessive amount of cash due to their prudential liquidity risk management, De Haan & Van den End (2013) and Merrouche & Archaya, (2012). Nevertheless, one should distinguish between holding cash to comply with LCR and hoarding excessive amounts as a disproportionate precaution measure. Cash can affect negatively bank’s return either directly or indirectly. A straightforward evidence from the direct effect is the fact that banks that hold lump amounts of cash can signal to investors a poor quality performance, Nwogugu (2014). Cash itself has no credit in adding value to the yield of bank’s assets. Thus hoarding cash can bring lower returns and profitability for banks. Furthermore, from agency theory cash is known as a catalysator in enhancing risk-taking behavior. In the general allotment of excessive liquid assets, might serve as a reassurance and masking mechanism for non-performing assets in the eyes of regulators. Hence, incentivizing loan officers to be more lenient in screening and loan underwriting.

Another practical concern regarding the LCR implementation and HQLA is the inclusion of corporate bonds and residential mortgage securities as Level 2 assets, Nwogugu (2014). Acceptance or not as Level 2 assets, is based on the publicly available rating of securities and loans. This can lead to a conflict of interest between issuers and credit rating agencies, the same conflict that enhanced the residential mortgage bubble. Banks can end up holding assets, which are overrated and be exposed to fire-sales in times of distress. Furthermore, HQLA as a whole can become a “fake positive” in good times and reveal the real value in periods of bank’s distress. This happens for the simple reason as in most of the cases HQLA assets are recorded either under the category of “Held to Maturity” or Available for Sale” assets in the balance sheet. What this means is that the declines in the prices of securities are not recorded and written down, resulting in bank’s credit quality and value of its assets being overstated and misleading.

Regarding the adjustment on the liabilities side of the balance sheet, Hartlage (2012) predicted a “snowballing” effect from the borrowing needs of US banks to meet LCR requirements. He feared that US banks would be locked in a circle of borrowing and ever growing borrowing requirements. According to the author, a bank can still be considered liquid, no matter the additional amount it has to borrow in excess to meet LCR target. Nevertheless, this can raise a red flag for the bank which would rely heavily on wholesale funding to satisfy their liquidity needs. During the latest crisis in the UK, accounting and equity losses came primarily from banks which relied the most on wholesale funding. Liquidity constraints came from banks’ solvency rather than the lack of funding in general, Merrouche & Archaya (2012). Precautionary cash holding requirements for banks led to an increase in interest rates in the entire interbank

(12)

8

market, volatility in the real economy and systemic risk. In an economic environment where banks are concerned about insolvency risk and asymmetric information, banks tend to behave like financially constrained institutions. Instead of charging to their counterparties rates close to the central bank’s base rates, they increase lending rates in order to charge for uncertainty. This finding goes along with previous studies of Bini (2010) and Blundell & Atkinson (2010). Both of the studies reported an increase in spreads between government bonds and less liquid maturities, higher operating costs for banks and an inconvenient lending environment for SME’s and small business. The Same experience was predicted for the Dutch market from Bonner & Eijffinger (2012), namely an increase in interest charged and paid for unsecured loans and deposits from the Dutch banks while trying to meet their liquidity requirements.

Nevertheless, authors also concluded that banks, which stand on the threshold of liquidity requirements, do not transmit their funding costs to the corporate lending market due to the lack of purchasing power. Banerjee & Mio (2015) supported the same idea for UK market. Even if ILG (UK) had an effect on intra-financial claims as it has been proclaimed by the banking industry, it would not have an impact on prices and quantity of retail/wholesale loans/deposits.

Taking into consideration all the predictions on the implications on asset and liability side of the balance sheet, the final goal is to give a conclusive answer on how Is LCR going to affect the overall performance of banks. Based on the overall argumentation above such as low yield assets, cash as a discouraging element of the asset portfolio, higher funding costs, increase in spreads among costs of funding, it is reasonable for one to expect a negative effect of LCR, at least in the short-run.

BCBS (2010), predicted that “tighter capital and liquidity requirements would lead to a 14bp higher lending rate and a decrease in lending volumes of 3.2%”. Härle et al. (2010) in a sensitivity analysis test study on Basel III, predicted that the effect of implementing the new requirements of Basel III would reduce the ROE by 4% in Europe and 3% in the US. Most importantly, they predicted that 0.9% decline will be attributed to the liquidity standards. Another supporting evidence comes from the research conducted from Raddatz (2010). The author claims that the return on US bank’s stock will be adversely affected by the increase in non-deposit wholesale funds. However, one needs to be careful in drawing any conclusion on these observations, as they are just predictions grounded on correlation analysis and trends in co-movement between bank’s return and asset/liabilities portfolio.

In a later study, Dietrich & Wanzenried, (2011) take into account such drawback and tries to build a causal relationship analysis on the possible effect of another Basel III liquidity ratio, NSFR (Net Stable Funding Ratio) on profitability measures (ROA, ROE and Net Interest Margin) in 7 European countries. Differently from LCR, this ratio is targeted to measure liquidity in the long-run, but shares similar characteristics to LCR when It comes to the composition of nominator (asset requirement). Preliminary results in this study revealed that European banks, which rely heavily on wholesale funding and have a high

(13)

9

ratio for loan to deposits, would be strongly affected. Differently from the previous studies Dietrich et.al (2014) conclude that NSFR would not significantly affect the measures of profitability. Here, it also important to emphasize that NSFR has not yet started to be implemented and banks are still on their first steps of adjustment toward meeting this ratio. More significant results are expected to be seen in the future. Past studies, give an overview on how banks could have possibly built their balance sheet restructuring strategy. For our analysis, it would be useful to use this information as a background for causal relationships between LCR and performance measure. More specifically, to examine if differences in strategies influence changes of LCR impact. In a post-crisis recovering economic scenario, it is expected that the major changes are reflected on the asset side of the bank’s portfolio, where banks are expected to leverage the low asset yield on higher fees for their clients. Nevertheless, one should not exclude the possibility that differences in the effect of LCR on the return can come also from the liability side, taking into consideration the low-interest rate environment and quantitative easing policy of central banks across different countries. Therefore, it would be interesting to analyze different banks, which have different funding opportunities. For instance, is there any difference on the banks that are privately or publicly owned? Have banks, on which government has a high stake, a smoother transition phase in LCR implementation? Do bigger size banks with more funding opportunities at a lower cost, have a significant competitive advantage compared to smaller counterparties in meeting the liquidity buffer?

With the contribution of the previously undertaken studies and the observations on the peculiarity of the current economic environment, the following hypotheses are drawn. Hypotheses, which are expected to assist us in analyzing the significance, magnitude, and difference of LCR effect on banking system performance.

Hypotheses

1. Sign of LCR: The overall effect of LCR on bank’s performance is negative. H0: β1/Europe and G20 Countries for Δ %LCR <=0

HA: β1/ Europe and G20 Countries for Δ % LCR >0

2. Comparison in the magnitude of the effect of LCR for EU and G20 sample. The negative effect is smaller for European countries compared to G20 ones.

H0: β1/Europe for Δ %LCR <=β1/G20 for Δ %LCR HA: β1/Europe for Δ %LCR > β1/G20 for Δ %LCR

3. Comparison in the sign and magnitude of the effect of LCR for different bank size

A- Sign: Effect is expected to be negative for smaller size banks and positive for banks with

significant market power.

H0: β1/Larger Bank Size for Δ %LCR >0 & β1/Smaller Bank Size for Δ %LCR<0 HA: β1/Larger Bank Size for Δ %LCR <0 & β1/Smaller Bank Size for Δ %LCR>0

(14)

10

B- Magnitude: For bigger size banks, the positive effect is stronger.

H0: β1/Larger Bank Size for Δ %LCR >=β1/Smaller Bank Size for Δ %LCR HA: β1/Larger Bank Size for Δ %LCR <β1/Smaller Bank Size for Δ %LCR

4. Effect of LCR for Non/Government Owned Entities: Banks with high government stake are less adversely affected from LCR. The effect of LCR on non-government owned banks is negative.

H0: β1/Government Owned Bank for Δ %LCR >=0 & β1/Non-Government Owned Bank for Δ %LCR <=0 HA: β1/Government Owned Bank for Δ %LCR <0 & β1/Non-Government Owned Bank for Δ %LCR >0

5. Effect of LCR for Private/ Public Entities: Public Owned banks are less affected compared to their privately owned counterparties. The effect of LCR on private-owned banks is negative.

H0: β1/Publicly Owned Banks for Δ %LCR >=0 & β1/Privately Owned Banks for Δ %LCR <=0 HA: β1/Publicly Owned Banks for Δ %LCR <0 & β1/Privately Owned Banks for Δ %LCR > 0

Methodology

The aim of this section is to establish a method which can assist in finding evidence on the significance of LCR effect on profitability. To analyze correlations and possible causal relations, the analysis will first focus on graphical representations to then be continued with panel data analysis.

i)

Graphical Representation

Before analyzing the significance of LCR effect, it is crucial to determine if there is a progress in the implementation process. To serve this purpose, there are built histograms2_{with observations’ frequency} for 4 group categories (LCR= 0-0.5; 0.6-1; 1.1-2; 2.1-5) along 3 respective years (2013 1st_{quarter until} 2015 4th_{quarter) for all countries.}

The next step in our analysis would be to determine which are the drivers behind meeting LCR. To relate these drivers to the LCR implementation is necessary to observe a dynamic representation of HQLA, NO and their components since the first official introduction of LCR in 2013. As the process of implementation of LCR is expected to be a gradual one, the growth rate of HQLA and Net Outflows and their components will be analyzed.

Via the final part of the graphical analysis, the research will try to reveal if there are significant changes in balance sheet restructuring due to LCR. To be able to determine this, the ratio of HQLA to Total Assets of the Bank will be observed from 2013 up until 2016. This ratio aims at representing the weight of HQLA on asset portfolio. Another ratio is one of Net Cash Outflows to Total Liabilities. Similarly, to the previous ratio, this ratio gives a dynamic overview of the outflows and inflows elements to Total Liabilities of the Bank. Having the dynamic development of these two ratios, helps us to understand which of these

(15)

11

two elements has a higher weight in implementation. Having this information as input, later on, we can predict how this development is related to the performance of banks.

To further predict the effect, one can also analyze correlation matrixes or possible trends between the main explanatory variable (LCR) and the profitability measures (ROE, ROA, Net Interest Margin).

It is noteworthy to mention that for all the above graphs; each observation is represented based on the weight of bank’s assets to total assets of the banks in the country where the bank is operating.

ii)

Panel Data Analysis

Panel data analysis will be based on a cross-country analysis of 1074 and 1684, G20 and EU banks in a time period 2013-2016 for 4 quarters, respectively. The reason why panel data, will be used for the analysis is the fact that it accounts for entities’ heterogeneity. Differently stated it allows to control bank specific variables that are time invariant, unobservable or hard to be identified. The specific feature that allows us to do so is called fixed effect. For our analysis, we will make use of the country and bank fixed effect. It is in the interest of this analysis to control for invariant characteristics of banks and countries, so we can filter out only the effect of LCR on performance.

From the fact that banks which operate in the same country share the same economic-legal-political environment, it can be intuitively implied that while we are controlling for time-invariant bank characteristics, we are simultaneously controlling for macro variables which can affect the performance of banks. Nevertheless, it remains crucial to control for the heterogeneous bank and macro indicators for banks. The following analysis introduces a detailed clarification behind the computation of LCR and the choice behind the dependent and control variables.

iii)

Liquidity Coverage Ratio

LCR= 𝐻𝑄𝐿𝐴

𝑁𝑒𝑡 𝐶𝑎𝑠ℎ 𝑂𝑢𝑡𝑓𝑙𝑜𝑤

LCR computation depends on the discretion of the national authorities. Moreover, at this point in time, banks are not required to make public their LCR results. For these reasons, LCR will be computed using as a backbone the recommendations of the Basel Committee Framework (Appendix A.1), while using a standardized approach based on the public available data from banks’ balance sheet. The advantage of following this approach helps to have a standardized measurement of LCR.

To quantity LCR, there are needed two components: High-Quality Liquid Assets (HQLA) and Net Cash Outflow (NO). The stock of HQLA is essential to cover cash outflows of the bank in case of a stress scenario along 30 days. The fundamental characteristics for the variables included in this category are low risk, the certainty of valuation, low correlation with risky assets, active, sizable market, and flight to quality, low volatility and to be listed on a recognized stock exchange BCBS (2013). What this basically means is

(16)

12

that these assets, which serve as a buffer, can be traded at their fair value in periods of distress and are central bank eligible. HQLA are divided into two more categories Level 1 Assets: the most liquid ones (cash and central bank bonds and notes) and less liquid ones Level 2 Assets (Table 1). Level 2 Assets are capped at 40% of the total HQLA.

The Denominator (Net Cash Outflow) is defined as the difference between Cash Outflows (Table 3) and Cash Inflows (Table 2). The only condition is that when computing Net Cash Outflows, Inflows cannot count more than 75% of the Outflows. Banks should consider as Cash Inflows, all inflows from performing contractual obligations which under no circumstance will default in a 30-day horizon. Cash outflows are more complex when it comes to interpreting and elaborating on them. In a rising scale for risk of withdrawal from short-term to long-term, from insured and secured up to unsecured sources of funding are listed in this category.

HQLA

If Level 1 Assets> Level 2 Assets:

HQLA=∑(𝐶𝑎𝑠ℎ + 𝐼𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡 𝑖𝑛 𝐺𝑜𝑣𝑒𝑟𝑛𝑚𝑛𝑒𝑡 𝑆𝑒𝑐𝑢𝑟𝑖𝑡𝑖𝑒𝑠 + 𝑇𝑜𝑡𝑎𝑙 𝑆𝑒𝑛𝑖𝑜𝑟 𝐵𝑜𝑛𝑑𝑠 &𝑁𝑜𝑡𝑒𝑠) ∗ 1) + (𝑇𝑜𝑡𝑎𝑙 𝐶𝑜𝑚𝑚𝑒𝑟𝑐𝑖𝑎𝑙 𝑃𝑎𝑝𝑒𝑟 ∗ 0.85) + (𝑀𝑜𝑟𝑡𝑔𝑎𝑔𝑒 𝐵𝑎𝑐𝑘𝑒𝑑 𝑆𝑒𝑐𝑢𝑟𝑖𝑡𝑖𝑒𝑠 ∗ 0.75)

If Level 1 Assets<Level 2 Assets:

HQLA=∑((𝐶𝑎𝑠ℎ + 𝐼𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡 𝑖𝑛 𝐺𝑜𝑣𝑒𝑟𝑛𝑚𝑛𝑒𝑡 𝑆𝑒𝑐𝑢𝑟𝑖𝑡𝑖𝑒𝑠 + 𝑇𝑜𝑡𝑎𝑙 𝑆𝑒𝑛𝑖𝑜𝑟 𝐵𝑜𝑛𝑑𝑠 &𝑁𝑜𝑡𝑒𝑠) ∗ 1) + (0.4 ∗ 𝐻𝑄𝐿𝐴 )

Table 1. High Quality Liquid Assets (HQLA)

Category Sub-Category Factor

H

Q

LA

Cash = Total Cash & Short Term Investment−Loans and Receivables from Credit Institutions−Investment in Government Securities Purchased under Agreement to Resell

Level 1 100%

Investment in Government Security Level 1

Marketable Securities in

Sovereigns

100%

Total Senior Bonds and Notes Level 1 100%

Total Commercial Paper Level 2-B 85%

Mortgage Backed Securities Level 2-B Qualifying

RMBS 75%

Note: Table 1 gives a rough representation of the variables which fall in each category, as in different legislation is advised to follow only the

general guideline recommended by Basel III (refer to appendix A.1). The variables for computation of LCR, which are mentioned in Table 1 have been retrieved from Capital IQ (S&P) database. In the first column is given the name of the standardized retrieved variable for all the banks. Second and third column give a broader and more specific category on which each component falls under the Basel III requirement for LCR. The last column represents the weight that needs to be considered for the computation of Level 1 and Level 2 assets. Weights are given based on liquidity transformation, certainty of inflows and risk of outflows during 30 days of liquidity distress.

(17)

13

Net Cash Outflow If Cash Outflow> Cash Inflow:

Net Cash Outflow= Cash Outflow- Cash Inflow

=∑(𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠 𝑑𝑢𝑒 𝑓𝑟𝑜𝑚 𝑑𝑢𝑒 𝑡𝑜 𝐵𝑎𝑛𝑘𝑠 ∗ 0.4) + (𝐶𝑜𝑟𝑝𝑜𝑟𝑎𝑡𝑒 𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠 ∗ 0.25) + ( 𝐷𝑒𝑚𝑎𝑛𝑑 𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠 ∗ 0.03) + (𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝐶𝑎𝑠ℎ 𝐵𝐸𝑎𝑟𝑖𝑛𝑔 𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠 ∗ 1) +

(𝑆ℎ𝑜𝑟𝑡 𝑇𝑒𝑟𝑚 𝐵𝑜𝑟𝑟𝑤𝑜𝑖𝑛𝑔 𝐷𝑒𝑟𝑖𝑣𝑎𝑡𝑖𝑣𝑒𝑠 ∗ 1) - ∑(𝑅𝑒𝑣𝑒𝑟𝑠𝑒 𝑅𝑒𝑝𝑜𝑠 ∗ 1) + (𝑆𝑒𝑐𝑢𝑟𝑒𝑑 𝐿𝑜𝑎𝑛𝑠 ∗ 0.5) + (𝑅𝑒𝑣𝑜𝑙𝑣𝑖𝑛𝑔 𝐶𝑟𝑒𝑑𝑖𝑡 ∗ 0.5)+(Level 1 Assets*0)+(Level 2 Assets * 0.15)

If Cash Outflow< Cash Inflow:

Net Cash Outflow= Cash Outflow- 0.75*Cash Outflow

=∑(𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠 𝑑𝑢𝑒 𝑓𝑟𝑜𝑚 𝑑𝑢𝑒 𝑡𝑜 𝐵𝑎𝑛𝑘𝑠 ∗ 0.4) + (𝐶𝑜𝑟𝑝𝑜𝑟𝑎𝑡𝑒 𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠 ∗ 0.25) + ( 𝐷𝑒𝑚𝑎𝑛𝑑 𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠 ∗ 0.03) + (𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝐶𝑎𝑠ℎ 𝐵𝐸𝑎𝑟𝑖𝑛𝑔 𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠 ∗ 1) +

(𝑆ℎ𝑜𝑟𝑡 𝑇𝑒𝑟𝑚 𝐵𝑜𝑟𝑟𝑤𝑜𝑖𝑛𝑔 𝐷𝑒𝑟𝑖𝑣𝑎𝑡𝑖𝑣𝑒𝑠 ∗ 1) - ∑(𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠 𝑑𝑢𝑒 𝑓𝑟𝑜𝑚 𝑑𝑢𝑒 𝑡𝑜 𝐵𝑎𝑛𝑘𝑠 ∗ 0.4) + (𝐶𝑜𝑟𝑝𝑜𝑟𝑎𝑡𝑒 𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠 ∗ 0.25) + ( 𝐷𝑒𝑚𝑎𝑛𝑑 𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠 ∗ 0.03) + (𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝐶𝑎𝑠ℎ 𝐵𝐸𝑎𝑟𝑖𝑛𝑔 𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠 ∗ 1) + (𝑆ℎ𝑜𝑟𝑡 𝑇𝑒𝑟𝑚 𝐵𝑜𝑟𝑟𝑤𝑜𝑖𝑛𝑔 𝐷𝑒𝑟𝑖𝑣𝑎𝑡𝑖𝑣𝑒𝑠 ∗ 1) *0.75

Table 2. Cash Inflow Components

Note: Table 2 gives a rough representation of the variables which fall in each category, as in different legislation is advised to follow only the

general guideline recommended by Basel III (refer to appendix A.1). The variables for computation of LCR, which are mentioned in Table 2 have been retrieved from Capital IQ (S&P) database. In the first column is given the name of the standardized retrieved variable for all the banks. Second and third column give a broader and more specific category on which each component falls under the Basel III requirement for LCR. The last column represents the weight that needs to be considered for the computation of Cash Inflows. Weights are given based on liquidity transformation, the certainty of inflows and risk of outflows during 30 days of liquidity distress.

C

ash Inf

low

Securities Purchased Under Agreements to Resell

(Reverse Repo)

Cash

Inflow Reverse Repo Secured by non-HQLA assets 100%

Secured Loans Cash

Inflows 50%

Revolving Credit Cash Inflows

Amounts to be received from non-financial wholesale counterparties, from transactions other

than those listed in above inflow categories.

50%

Level 1 Assets 0%

(18)

14

Table 3. Cash Outflow Components

C

ash O

utf

low

Deposits from/due to Banks Cash Outflow: Unsecured Wholesale Funding

Unsecured Wholesale Funding (Non-Financial Corporates, Sovereigns, Central Banks, Multilateral

Development Banks, and PSEs)

40% Corporate Deposits Cash Outflow: Unsecured Wholesale Funding

Operational Deposits Generated by Clearing, Custody, and Cash

Management Activities

25%

Demand Deposits Cash Outflow:

Retail Deposits Covered by Insurance Scheme 3%

Interest Bearing Cash Deposits

Cash Outflow:

Retail Deposits Stable/Less Sable Deposits 10%

Short Term Borrowing Derivatives

Cash Outflows Net Derivative Cash Outflows 100%

Note: Table 3 gives a rough representation of the variables which fall in each category, as different legislation is advised to follow only the general

guideline recommended by Basel III (refer to appendix A.1). The variables for computation of LCR, which are mentioned in Table 3 have been retrieved from Capital IQ (S&P) database. In the first column is given the name of the standardized retrieved variable for all the banks. Second and third column give a broader and more specific category on which each component falls under the Basel III requirement for LCR. The last column represents the weight that needs to be considered for the computation of Cash Outflows. Weights are given based on liquidity transformation, the certainty of inflows and risk of outflows during 30 days of liquidity distress.

iv)

Dependent and Control Variables Choice

3

The purpose of this section is to introduce the reader with the reasoning behind the choice of dependent variables, the construction the main control variables. Behind each variable selection stands an integration of past studies supporting evidence and the argumentation of the author.

Dependent Variables

As a performance measure for banks, 3 different indicators will be used, namely Return on Assets, Return on Equity and Net Interest Margin. These variables are not only the widest used by previous studies (Härle et al. 2010, and Dietrich et.al 2014), but also the most significant ones among all the studies.

One of the most criticized element of LCR is HQLA due to low yield assets and possible higher funding costs over time. Thus it is necessary to identify a ratio which can help us to analyze the impact of assets’ yield on banks’ profit. Return on Assets (ROA) is a ratio which is widely used by the banking industry to examine how efficiently the management is utilizing its assets.

Another concern is that LCR can affect return to investors’ equity. ROE is seen as an adequate performance measure for the equity investors for 3 reasons, simplicity in calculation, easy

(19)

15

access and feasibility in peer comparative analysis and comparison to the opportunity cost of equity. ROE is also preferred to ROA when the majority of bank’s income comes from off-balance sheet activities.

Lastly, Net Interest Margin helps to analyze if liquidity requirement is having any significant effect on funding spreads for banks and consequently on bank’s performance.

Control Variables

In order to identify a possible effect of LCR on banks, it is essential to control for other determinants of bank’s profitability. Past literature is rich in evidence about internal and external determinants. Athanasoglou et.al (2008) conducted a comparison on the main indicators on bank’s profitability, distinguishing between external and bank-specific determinants. Differently from bank-specific indicators, macroeconomic contributors represent the effect from the economic and legal environment of the country where the bank is operating. Stated differently, they are not related to the bank management and how the bank is being operated.

Research conducted on the effect of bank specific indicators on profitability is very broad. Nevertheless, up to date results reveal that the most significant indicators can be summarized as asset size, asset growth, ownership structure and credit rating of banks. One of the pioneer papers behind the asset size theory was the one presented by Demsetz (1973). In these papers, the author presented the “the efficient structure hypothesis”, which states that only firms with superior efficiency will earn supernatural profits. Demsetz (1973) viewed asset size and growth as the source of this superior efficiency, where banks can continuously gain from economies of scale. Later studies from Smirlock (1985) and Bikker & Hu (2002), were able to quantitatively prove the theory of Demsetz by showing that better-capitalized banks have the ability to fund themselves with less expensive capital and earn higher profits. Ravenscraft (1980) recognized the fundamentals of “the efficient structure hypothesis”, but at the same time concluded that supernatural profits are not only related to the reduction of costs in funding, but also to the ability to charge excessive prices. Excessive prices are attributed mostly to the market power of the bank which comes from product differentiation. While controlling for differences across financial and legal systems in different international banks, Demirguc-Kunt and Huizinga (1999), concluded that a higher ratio of bank’s assets to gross domestic product of a country was accompanied with higher profits and net interest margins. Nevertheless, what is missing in previous studies is taking into consideration the asset size growth and good performance of banks cannot be positively interconnected with each other indefinitely. One needs to consider also the neo-classical economic theory for firms. High growth it is not sustainable in the long-run due to diminishing marginal returns. To reflect the weakening

(20)

16

effect of asset growth on profitability, it is necessary to control for the quadratic term of this variable. Under this scenario, statistical results will reveal a positive effect of asset growth and the negative effect of its quadratic term.

While the literature on asset size was relatively conclusive, the same does not hold for ownership structure of banks. The first target of previously held research was to study the significance of ownership status on banks and then distinguish between private/publicly and government/non-government owned banks. Altunbas.et.al (1997) conclude that ownership status is an insignificant factor to bank’s profitability. Athanasoglou et.al (2008) revisit once more the ownership effect in banks ROA and reassure that there exists no such relationship. Nevertheless, in more recent studies, Iannotta et.al (2007) and Micco et.al (2007), point out that banks which are owned by the government have lower profitability compared to private banks. Private and mutually owned banks perform better due to better loan quality and lower insolvency risk, despite the lower costs advantage of public sector banks. Dietrich and Wanzenried (2011) in a study on Swiss banks distinguish between pre and post crisis period with regard to the relationship of ownership status and profitability. According to their study, opposite to previous studies results, state-owned banks were better performing during the crisis period. Throughout periods of turmoil, financial institutions owned by the government are seen as more stable and safe: not only can they be easily funded but also they run a lower risk for being overlooked to insolvency. Another determinant in the relationship of profitability and ownership dispersion is if the bank is public or privately owned. Existing literature, for instance, Iannotta et.al (2007), claims that banks which are publicly owned are affected by the market discipline mechanisms. Listed banks can be more profitable, but not necessarily more cost efficient. Dietrich and Wanzenried (2011) explain that listed banks are confronted with bigger pressure not only from their shareholders but also from regulators and thus the overall result cannot be clearly identified. The additional costs for meeting shareholders’ expectation, but at the same time the strict requirement of regulators due to the systemic risk they impose in the case of failure, can overcome the benefits of being listed. Following the results of previous work, one cannot deny the importance of ownership status on funding, but also the exposure to market risk. In order to conclude on the significance of ownership status of banks (private/publicly, government/publicly owned), it is essential to tackle the ownership dispersion effect in steps. More specifically, first contrast between ownership effect across countries and secondly analyze this effect across countries in the different time-frames (pre-post) crisis period.

The last variable which is perceived as essential to the banks’ performance is credit rating. The credit rating reflects how risky a bank is perceived from the investors and what the probabilities of default in case of insolvency are. All these expectations are directly related to the funding costs

(21)

17

of banks and hence to its profitability. Lepetit et.al (2008) takes a broad perspective on credit rating literature. Banks which are involved in risky investing earn a higher return on their investment activities but at the same time they trade-off earnings with their risk for insolvency. Lepetit et.al (2008) report that banks which are involved in non-interest earning activities, perceived also as more risky and higher probability of being insolvent, are the ones which have the highest return on assets and equity. Athanasoglou et.al (2008) takes this analysis further by studying the weight of both components for Greek banks. The results of his study reveal that not only has credit risk a significant effect on banks profitability but that the end result is an adverse effect on bank’s profits. Taking into consideration the results of previous studies, it is fundamental for this research to recognize and control for the effect of this variable on a bank’s profitability.

As it was earlier mentioned, banks’ profitability is not only determined by internal but also external factors which do not depend on how the bank is managed. Borio et.al (2001) explained that the financial system is excessively procyclical. The root of procyclicality is asymmetric information between credit seekers and lenders. When the economy is depressed and collateral values are low, even the most creditworthy clients face difficulty in financing themselves. This situation deteriorates further with tightening of the lending constraints from financial institutions and further amplification of this adverse cyclical effect into the economy. Thus, it is essential to analyze the significance of the economic cycle effect on bank’s performance. Different proxies have been chosen to represent the development of economic cycle, such as inflation, gross domestic product (GDP), business cycle, unemployment, on banks’ profits. Demirguc-Kunt and Huizinga (1999) and Bikker and Hu (2002), were one of the first authors to find a positive correlation between bank’s profitability and the business cycle. Demirguc-Kunt and Huizinga (1999) used GDP growth as a proxy for the business cycle impact, while Bikker and Hu (2002) implemented unemployment, inflation rate, GDP, the interest rate in their analysis. Based on the previous work of these authors, Athanasoglou et.al (2008) proved the significance of inflation and cyclical output on the performance of Greek banks. Other studies (Kaminsky and Schmukler 2002, Brooks et al. 2004, and Correa et.al 2014) analyzed how negative country externalities can spread into the banking system. These studies focused on the effect of a country’s credit downgrade on the bank’s stock return. Correa et.al (2014) results revealed a strong and significant negative effect of a credit downgrade on stock returns, especially for the banks that were expected to be subsidized by the government or the banks in advanced economies. One channel of transmission can be that in the case of a downgrade, the economy needs to borrow at higher costs. These higher costs are reflected in higher taxes and more frequent borrowing from the government that leads to frictions in the economy, lower private saving and investment. Hence, lower profits for banks. Another channel is

(22)

18

via the bank balance sheet. As many banks hold and will be required to hold more high-quality liquid assets such as government bonds, a credit downgrade will directly reduce the value of its assets and adversely affect the performance of banks. A last example from the sovereign-bank nexus is the fact that lower credit rating for countries can be translated into less reliable guarantees for domestic banks, where the government has committed to accept “too big to fail”.

The last variable that will be taken into consideration as a control variable is the interbank money market rate. This market is excessively sensitive to the contagion effect across banks. Interbank loans, differently from the retail deposits are not insured neither collateralized. The failure of one bank can light a chain of failure in the banking system, Rochet & Tirole (1996). Acharya and Yourulmazer (2008) explain that in the scenario of a systemic risk contagion, profit-maximizing banks herd with other banks. In periods of economic distress and high uncertainty, asymmetric information concerns in the financial system are reflected in the interbank market borrowing costs and hence performance of banks. As in the interest of this study is to analyze very short time frames from 1 day up to a max of 30 days (stress period of LCR), overnight borrowing and 30-day borrowing rate will be used as representatives of the effect of interbank rates (borrowing costs) on the performance of banks.

Based on the extensive analysis above, all the indicators discussed seem to have a direct/indirect association with the banks’ performance (Figure 1). For this reason, all the indicators will be used as a control variable for the regression analysis. Nevertheless, to prove the causal impact of LCR, endogeneity issues need to be restrictively excluded4_{. In addition to the statistical tests for endogeneity, another modest} approach will be used to observe the relationship between LCR and the control variables. For this reason, as a first step: bank and macro indicators will be regressed on the main explanatory variable (LCR changes). Only after demonstrating the insignificance of all performance influencers on LCR, we can proceed with the main regressions where LCR and all the control variables will be simultaneously integrated, as it is demonstrated in the below regressions. The three performance measures will be analyzed separately for all the hypotheses.

The regressions will be run first for EU countries, including Switzerland and Norway and then for G20 countries. Following this step, will help us to compare the effect for developed and developing economies and ultimately conclude if there is a significant difference in effect.

Another step in the analysis is to compare the effect of LCR for two different subgroups, namely private/public and government/non-government owned banks, to test the hypothesis on the difference between these two subgroups.

4_{Consult with the Robustness Check section.}

(23)

19

Dependent, independent and control variables will all be represented as natural logarithms. This choice is justified by the fact of time constraint in observing LCR developments and its possible externalities. Percentage change makes it easier to observe a possible effect caused by changes in balance sheet components to meet the 100% LCR standard.

What also need to be highlighted is the fact that all the variables’ observations are proportionally weighted based on total assets of the bank on the total assets of the banks in the respective country.

This allows a fair representation of the regression results based on the importance of the bank. Finally, to be able to further test significance strength in results, it is necessary first to compute the LCR for an extended period, namely before 2013. It is possible to do this step as we already have the data, and we can use the same approximations we used for the computations of LCR for the post-2013 period. The next step is to run regressions for two time periods before and after 2013. If the effect of LCR is significant for the post period 2013, but insignificant for the prior period, this strengthens further the significance of LCR influence since the moment it has been introduced.

Note: i represents observations for each bank for each quarter of the year; t represents the respective quarter of

(24)

20

Figure 1. Performance Indicators

Note: This figure gives an overview of the relationship between the dependent (ROA,ROE,NIM), control and explanatory variables.

While control variables are divided intro bank and country specific, their definite impact on performance is either positive/negative (+/-). The question mark between the control and the main explanatory variable it is used to represent the fact that we are not sure about the influence of these indicators on LCR and furthermore if they indirectly affect performance through liquidity regulation. What we expect to see is a negative effect of LCR on performance, represented by the (-) sign.

Data and Descriptive Statistics

i)

Data

Data has been collected from two main databases: S&P Capital IQ and DataStream Thompson Reuters on quarterly basis for the time period 2003-2016. The reason, we go so back in time is to be able to make a comparison when needed to validate the significance of LCR as it was explained extensively above. Another reason is also to observe dynamics of its components such as HQLA and NO.

The first database has been used to retrieve information on balance sheet/industry specific items, profitability/margin ratios, S&P rating for a list of 1074 and 1684 G20 and EU (including Norway and Switzerland) banks. Part of these samples are only commercial and saving banks. The reason we constrain the sample to these type of banks is due to the fact that for other banks, special rules apply and including them in our sample might lead to biased results.

DataStream provides information for Macro Economic Analysis. From this database, it was retrieved information on Gross Domestic Product GDP (Standardized and Growth), Credit Default Swaps (CDS) (6 months), Interbank Borrowing Rates (Overnight and 30 days’ period deposits) and Treasury Bill Rates (3-6 month). Even though we are using a broad range of control variable at the beginning, the regression analysis will include only the most significant ones and that make most economic sense.

From both databases was collected data on European Union countries (including here also Norway and Switzerland) and Group Twenty (G20) member countries. EU countries are part of a common political-economic environment that operate as a unified block. G20 is an international forum for government central

Bank Specific

Performance

Credit Rating Asset Growth Market Power GDP Growth CDS spread Overnig ht Rate 30 Days' Deposits Rate ROA ROE NIM

Macro

?

(25)

21

banks of 20 main global economies (19 individual countries and EU as an entity), which the main goal is to promote international financial stability. Here needs to be mentioned that due to lack and fragmentation of data, South African banks are left out of this study. Part of this exclusion is also EU, countries of which are studied separately.

As supported by the literature and the recently done studies from EBA and Basel Committee, counties of both EU and G20 have been the most active in the implementation campaign of LCR. By including in this list also developing economies can lead to biased results due to big discrepancies in economic and regulatory development compared to major economies.

EU countries’ similarity in fiscal-monetary-political policy will help to control for country-specific effect and focus specifically on the impact of our variable of interest. Meanwhile analyzing the G20 economies will help to do the same analysis from another perspective, one of dynamic economies with evolving regulatory environment.

In Table 4 and Table 5, the number of banks screened by government/non-government and public/private ownership for G20 and EU countries is represented respectively. A bank is considered to be owned by the government if the government owns more than or equal to 50% of the stake, while private ownership is established if the bank is listed on a stock exchange.

For 20 countries it is depicted an equal weight of privately and publicly owned companies as a cumulative of all the banks across countries. Nevertheless, the same cannot be concluded for the number of banks which are government and non-government owned. In this summary, the total number of banks on which the government has a high stake is underweighted compared to the number of banks owned by

private investors.

For EU countries, the sample is once more mainly represented by banks where the government does not have a high stake. The weight of privately and publicly owned companies is again relatively equally distributed.

(26)

22

Table 4. Composition for G20 Countries Table 5. Composition for European Countries

Privately Owned Government Owned

Country No Yes Grand

Total No Yes Grand Total Argentina 15 9 24 23 1 24 Australia 5 10 15 15 15 Brazil 17 7 24 21 3 24 Canada 5 7 12 12 12 China 27 16 43 38 5 43 France 21 17 38 37 1 38 Germany 13 13 26 25 1 26 India 15 7 22 10 12 22 Indonesia 20 14 34 32 2 34 Italy 26 25 51 48 3 51 Japan 8 6 14 12 2 14 Mexico 21 10 31 31 31 Russia 42 35 77 75 2 77 Saudi Arabia 8 4 12 1 11 12 South Korea 11 6 17 14 3 17 Turkey 20 12 32 30 2 32 United Kingdom 14 13 27 22 5 27 United States 247 328 575 574 1 575 Grand Total 535 539 1074 1020 54 1074

Private Owned Government Owned

Country Yes No Grand

Total No Yes Grand Total Austria 39 51 90 88 2 90 Belgium 13 10 23 21 2 23 Bulgaria 9 2 11 11 11 Croatia 27 24 51 44 7 51 Cyprus 4 8 12 12 12 Czech Republic 12 19 31 26 5 31 Denmark 37 37 74 68 6 74 Estonia 17 29 46 45 1 46 Finland 16 15 31 27 4 31 France 58 77 135 120 15 135 Germany 36 72 108 100 8 108 Greece 23 35 58 56 2 58 Hungary 5 15 20 20 20 Ireland 34 15 49 46 3 49 Italy 74 71 145 136 9 145 Latvia 11 14 25 24 1 25 Lithuania 13 20 33 29 4 33 Luxembourg 11 16 27 22 5 27 Malta 8 11 19 19 19 Netherlands 35 37 72 64 8 72 Norway 44 61 105 98 7 105 Poland 27 71 98 84 14 98 Portugal 20 13 33 30 3 33 Romania 10 4 14 14 14 Slovakia 6 12 18 17 1 18 Slovenia 17 22 39 33 6 39 Spain 47 58 105 95 10 105 Sweden 29 22 51 50 1 51 Switzerland 29 28 57 56 1 57 United Kingdom 49 55 104 98 6 104 Grand Total 760 924 1684 1553 131 1684