Transformation Risk and its Determinants:

A New Approach based on the Basel III Liquidity Management Framework

Alain Angora, Caroline Roulet•

Université de Limoges, LAPE, 5 rue Félix Eboué, 87031 Limoges Cedex, France

This version: April 2011

Preliminary draft - Please do not quote without the permission of the authors

Abstract

Liquidity creation is one of the pre-eminent functions of banks but it is also a major source of their vulnerability to shocks. Considering US and European publicly traded commercial banks from 2000 to 2008, we consider the new measures of liquidity defined in the Basel III accords to estimate a level of liquidity creation beyond which a bank may not able to meet its liquidity requirements. Besides, as financial innovation provides new ways for banks to manage their liquidity, we investigate how transformation risk is impacted by the concentrations on loans that are potentially securitisable and on short term, potentially unstable market funding. On the whole, we show that transformation risk decreases with a higher concentration on loans that are potentially securitisable. However, transformation risk increases when banks are more concentrated on short term market debts. Thus by better understanding what factors significantly impact transformation risk, it can help banks to improve their risk management framework.

Keywords: Liquidity Creation, Transformation Risk, Bank Regulation

JEL classification: C23, G21, G28, G32

Corresponding author. Tel: +33-555-32-81-88, c.roulet@jplc.fr (C. Roulet).

1

1. Introduction

According to the theory of financial intermediation, an important role of banks in the

economy is to provide liquidity by funding long term, illiquid assets with short term, liquid

liabilities. Through this function of liquidity providers, banks create liquidity as they hold

illiquid assets and provide cash and demand deposits to the rest of the economy. Diamond and

Dybvig (1983) emphasize the “preference for liquidity” under uncertainty of economic agents

to justify the existence of banks: banks exist because they provide better liquidity insurance

than financial markets. However, as banks are liquidity insurers, they face transformation risk

and are exposed to the risk of run on deposits. More generally, the higher is liquidity creation,

the higher is the risk for banks to face losses from having to dispose of illiquid assets to meet

the liquidity demands of customers.

There is a large body of theoretical literature dealing with bank liquidity creation

(Bryant, 1980; Diamond and Dybvig, 1983; Holmstrom and Tirole, 1998, Kashyap et al.,

2002). Nevertheless, empirical studies are more recent and deal with the measurement

methodologies and the determinants of liquidity creation. Deep and Schaefer (2004) define

the “liquidity transformation gap” (also called, “LT gap”) as the difference of liquid liabilities

and liquid assets held by a bank, scaled by total assets. If the difference is positive, the bank

invests liquid liabilities into illiquid assets and performs a significant amount of liquidity

creation. Deep and Schaefer (2004) consider only the maturity to define the liquidity of bank

assets and liabilities. They consider as liquid all assets and liabilities that mature within one

year. Berger and Bouwman (2009) define the liquidity of bank assets and liabilities not only

based on their maturity but also by considering their category. In addition, their indicator

includes on and off-balance sheet items. Then, by considering the “liquidity transformation

gap” or the “liquidity creation”, several studies focus on the determinants of liquidity creation

(Deep and Schaefer, 2004; Rauch et al., 2008; Berger and Bouwman, 2009; Choi et al., 2009;

Pana et al., 2009; Chen et al., 2010). They consider several determinants such as bank capital,

profitability, credit risk, market power, the business cycle and the level of central bank policy

rate. All of these studies portray liquidity creation as an essential role of banks but they do not

deal with the liquidity pressures that banks may face and the possible excessive liquidity

creation. Indeed, the more banks create liquidity, the higher is their illiquidity and their risk to

face losses from having to sell some assets at fire sale prices to repay some debts claimed on

demand. However, liquidity creation is not likely to be damaging for a bank as long as it holds

adequate levels of stable funding to fund the amount of assets that cannot be monetised or that

2

cannot be pledged as collateral. In this context, the bank creates liquidity but it is able to repay

the liabilities claimed on demand by selling its liquid assets or using them as collateral.

Throughout the global financial crisis which began in mid-2007, many banks struggled

to maintain adequate liquidity. Unprecedented levels of liquidity support were required from

central banks in order to sustain the financial system and even with such extensive support a

number of banks failed, were forced into mergers or required resolution. Thus, banks have

experienced difficulties for managing their liquidity and face transformation risk, but the

problem is not solved yet. Following the Subprime crisis and in recognition of the need for

banks to improve their liquidity management, the Basel Committee on Banking Regulation

and Supervision has developed an international framework for liquidity assessment in banking

(BIS, 2009). Among the several guidelines, the Basel III accords include the implementation

of liquidity ratios concomitantly to capital standards in order to strengthen the stability of

banks1. Although banks face liquidity pressures and experience liquidity problems, financial

innovation enables them to manage their liquidity by mitigating the liquidity pressures

through new asset - liability management (ALM) framework. Following financial

globalisation and deregulation, banks have largely enhanced their market activities through

financial innovation (Shleifer and Vishny, 2009). On the liability side, banks modify their

funding structure and increase the share of market funding. On the asset side, they securitise

their loans. Such financial innovations enable banks to access to new sources of liquidity by

reducing their reliance on deposits (Mishkin, 2004) and by converting their illiquid loans into

cash (Loutskina, 2011).

Based on these facts, we suggest in this paper to extend the current literature on bank

liquidity creation in two directions.

The first objective of this paper is to assess the level of liquidity creation beyond

which a bank may not able to meet its liquidity requirements without borrowing money or fire

1 The Basel Committee on Banking Regulation and Supervision has developed two regulatory standards for liquidity (2009). The “net stable funding ratio” measures the amount of longer term, stable sources of funding used by an institution relative to the liquidity profile of the assets funded and the potential for contingent calls on funding liquidity arising from off balance sheet commitments and obligations. The standard requires a minimum amount of funding that is expected to be stable over a one year time horizon based on liquidity risk factors assigned to assets and off balance sheet liquidity exposures. This metric is intended to promote longer term structural funding of banks’ balance sheet, off-balance sheet exposures and capital markets activities. The Basel Committee also suggests the “liquidity coverage ratio”. This metric identifies the amount of unencumbered, high quality liquid assets an institution holds that can be used to offset the net cash outflows it would encounter under an acute short term stress scenario (i.e., over a 30 days time horizon) specified by supervisors. These proposals have been fully calibrated and agreed upon 12, September 2010 (Basel III accords).

3

selling its assets. In other words, we assess the level of liquidity creation a bank can perform

by being continuously able to face transformation risk. Although through liquidity creation

banks face transformation risk, the liquidity creation indicator suggested by Berger and

Bouwman (2009) does not indicate to what extent liquidity creation may become damaging

for a bank in terms of excessive liquidity creation and exposure to transformation risk (i.e.,

“how much is too much?”). Based on the Basel III accords, we consider the net stable funding

difference. It is computed as the difference of the required amount of stable funding and the

available amount of stable funding, scaled by total assets. It measures the amount of assets

that could be not monetised through the sale or the use as collateral in a secured borrowing

compared with the amount of longer-term, stable sources of funding used by an institution.

This indicator estimates the liquidity profile “at-risk” of banks from their liquidity creation

activities. It includes the liquidity unbalances of both sides of bank balance sheet. Besides, it

accounts for the impact of the liquidity of the financial markets, on the valuation of assets and

on the availability of funding, to assess bank exposure to transformation risk. If the difference

is negative or null, the required amount of stable funding is lower or equals the available

amount of stable funding. It means that the bank is not exposed to the risk of having to sell

some assets at fire sale prices to repay the liabilities claimed on demand. In the contrary to the

liquidity creation of Berger and Bouwman (2009), the net stable funding difference explicitly

shows a threshold beyond which a bank is likely to experience difficulties due to its inability

to face transformation risk. As we assume that bank illiquidity and transformation risk

increase with liquidity creation, and that the net stable funding difference is a measure of the

liquidity profile “at-risk” of banks, we show the similarity of these two indicators by doing a

statistical analysis. Then, we outline the advantages of the net stable funding difference

compared with the liquidity creation in order to estimate a level of liquidity creation for which

a bank is continuously able to meet its liquidity requirements with its own liquid assets (i.e.,

when the net stable funding difference is null). We call this level of liquidity creation the

“transformation risk neutral level of liquidity creation”. This issue seems relevant in order to

assess banks’ ability to face transformation risk when they create liquidity. Until this level of

liquidity creation, the bank has not to face losses as its holds enough assets that can be readily

monetised or that are pledgeable as collateral to meet the liquidity demands of customers. In a

regulatory perspective, the transformation risk neutral level of liquidity creation may be

useful, to evaluate from what level, liquidity creation may become excessive and damaging

for the stability of banks.

4

Although through their essential role of liquidity creation, banks face transformation

risk and may become fragile, financial innovation provides new ways for banks to manage

their liquidity and mitigate liquidity pressures. The second objective is to study how

transformation risk is impacted by the concentrations on loans that are potentially

securitisable and on short term, potentially unstable market funding. First, we question

whether the concentration on the loans that are potentially securitisable decreases

transformation risk. Indeed, one of the key issues in bank liquidity analysis is the liquidity of

assets. Cash, near cash items and trading assets are not problematic for bank liquidity. These

assets are liquid or can be easily monetised2. However, among the other assets, some assets

are totally illiquid and may lead to acute liquidity problems. Nevertheless, other assets even if

they are not directly saleable on financial markets may be sold through OTC transactions,

such as the loans that are securitised. Thus, we hypothesize that the concentration on the loans

that are potentially securitisable rather than on totally illiquid assets is likely to mitigate

liquidity pressures on banks and may decrease transformation risk. Second, we test the impact

on transformation risk of the concentration on short term, potentially unstable market funding.

Indeed, the stability of funding is another important issue for liquidity analysis in banking.

Short term debts are less stable than long term ones3. Besides, according to the BIS (2009),

short term deposits may be considered as more stable than short term market debts. Thus, the

more banks are funded by short term market debts, the greater is the potential instability of

their funding. Consequently, we hypothesize that the concentration on short term, potentially

unstable market funding rather than on short term, stable deposits is likely to increase

liquidity pressures on banks and transformation risk. However, banks may consider possible

liquidity shortages on funding markets (i.e., some market debts may be rolled-off at short

notice) to limit their liquidity creation. Thus, we conjecture that the concentration on short

term, potentially unstable market funding is likely to discourage banks for increasing their

liquidity creation that leads to lower exposure to transformation risk. The impact on

transformation risk of the concentration on short term, potentially unstable market funding is

ambiguous. The purpose is to point out the main factors that significantly impact

transformation risk in order to help banks to improve their risk management strategies.

2 As they are continuously traded on financial markets, it is possible to find a counterparty and sell these assets with no or relatively low discount.3 Long term debts are repayable by contract at their residual maturity which must exceed one year. Short term debts are due within one year or may be claimed at short notice.

5

Our results confirm the similarity of the liquidity creation indicator of Berger and

Bouwman (2009) and the net stable funding difference by considering US and European

publicly traded commercial banks over the 2000-2008 period. In addition, the net stable

funding difference enables us to assess a level of liquidity creation beyond which a bank may

not able to meet its liquidity requirements with its liquid assets. Moreover, our results show

that transformation risk decreases under high levels of concentration on loans that are

potentially securitisable. However, transformation risk increases when banks are more

concentrated on short term market debts.

The remainder of this paper is organised as follows. In section 2, we present the data.

In section 3, we describe our indicator of liquidity creation, the net stable funding difference

and we do a statistical analysis to assess the transformation risk neutral level of liquidity

creation. In section 4, we detail the determinants of transformation risk and we discuss the

regression framework. In section 5 and 6, we comment our regression results and perform

some robustness checks. Section 7 concludes.

2. Presentation of the sample

Our sample consists of US and European4 publicly traded commercial banks from

2000 to 2008. We focus on US and European banks because the required data are available on

standard databases to ensure an accurate representativeness of our sample of banks in each

country. Furthermore, we focus on listed banks because their balance sheet data are more

detailed which allows us to compute our indicators of liquidity that are our main variables of

interest.

Annual financial statements are extracted from Bloomberg. From 2000 to 2008, we

identify 870 listed commercial banks (645 in the US and 225 in Europe). However, the

breakdown for loans by category and the breakdown for deposits by maturity, which are

necessary to compute our proxies of liquidity, are not detailed in Bloomberg or in annual

reports for 71 US banks and 18 European banks. Thus, the final sample consists of 781

commercial banks (574 in the US and 207 in Europe). In table 1, we present the distribution

of banks by country. To deal with the issue of sample representativeness, we verify that on

4 We use data for European banks from the 20 following countries: Austria, Belgium, Cyprus, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Italy, Liechtenstein, Malta, Netherlands, Norway, Portugal, Spain, Sweden, Switzerland and the United Kingdom.

6

average from 2000 to 2008, the final sample constitutes over 66.4% of the banking assets of

US commercial banks and over 60.4% of the banking assets of European commercial banks.

[Insert Table 1]

Table 2 presents some general descriptive statistics of our final sample. By

considering several key accounting ratios, the data show that banks are on average focused on

traditional intermediation activities. Indeed, loans and deposits account for a large share of

total assets. The average share of total loans in total assets is 66.4% and the average share of

total deposits in total assets is 70.2%. In addition, on average, interest income accounts for

nearly three quarters of total income (72.3%). However, there is a high heterogeneity across

banks as shown by the high standard deviation and the extreme values of each ratio5.

Regarding the quality of bank assets, the average share of total provisions for loan losses in

total loans is 0.5%. In terms of profitability, the average return on assets is 0.9%. Lastly,

considering capitalisation, the average total risk weighted capital ratio is higher than the

minimum regulatory requirement at 13.2% and the average ratio of Tier 1 capital to total

assets is 8.2%.

[Insert Table 2]

3. Measuring bank liquidity and the transformation risk neutral level of liquidity

creation

3.1. Indicator of liquidity creation

Our indicator of liquidity creation is based on the liquidity creation measure in the

steps of Berger and Bouwman (2009). In the first step, all bank assets and liabilities are

5 We notice that the average share of total loans to total assets is significantly higher for US banks than for European banks (respectively, 67% and 66%). In addition, the average share of total deposits to total assets is significantly higher for US banks than for European banks (respectively, 77% and 51%). Besides, the average share of interest income is significantly higher for US banks than for European banks (respectively, 77% and 59%). These specificities may be explained by the differences in regulation in the US from Europe. Indeed, in the US, banking groups are submitted to requirements in terms of segmentation of their activities into several subsidiaries. In addition, US banking groups are allowed to carry out activities “closely related to banking”, such as investment banking and insurance, only if they are considered as “well capitalised” by the Federal Reserve (i.e., if they meet the Fed’s highest risk-based capital rating). It is the reason why most banking groups are focused on “banking business”, primarily issuing deposits and making loans. However, in Europe, banking groups are not submitted to such a regulation and can more easily develop their market activities.

7

classified as liquid, semi liquid or illiquid according to their maturity and their category.

Indeed, some assets are considered as easier to sell than others (such as the loans are

securitisable and the securities that are saleable on financial markets). Besides, some fundings

are considered as more volatile than others (such as commercial paper and short term

deposits). In the second step, both balance sheet sides are weighted in reference to the

liquidity creation theory suggested by Berger and Bouwman (2009). Table 3 shows the

weighting of bank balance sheet6 based on Berger and Bouwman (2009).

[Insert Table 3]

Liquidity creation (LC) is then calculated as follows (where all components are scaled by total

assets):

0.5 * illiquid assets + 0 * semi liquid assets - 0.5 * illiquid assets + 0.5 * liquid liabilities + 0 * semi liquid assets - 0.5 * illiquid liabilitiesLC = Total assets

All else equal, a bank creates one dollar of liquidity by investing one dollar of liquid liabilities

(such as transaction deposits) into one dollar of illiquid assets (such as business loans).

Similarly, a bank destroys one dollar of liquidity by investing one dollar of illiquid liabilities

or equity into one dollar of liquid assets such as treasury securities (i.e., the bank removes one

dollar of liquidity from the non bank public by replacing liquid treasuries with illiquid

liabilities or bank equity). The higher is the liquidity creation, the higher is bank illiquidity as

it invests more liquid liabilities into illiquid assets. In this context, the bank is at risk if some

debtholders claim their funds on demand when assets are saleable at fire sale prices.

3.2. The net stable funding difference

Although liquidity creation increases bank illiquidity and transformation risk, the

liquidity creation indicator suggested by Berger and Bouwman (2009) does not indicate to

what extent liquidity creation may become damaging for a bank in terms of excessive

liquidity creation and exposure to transformation risk. In this perspective and based on the

Basel III guidelines for bank liquidity assessment (BIS, 2009), we consider an alternative

6 In their model, Berger and Bouwman (2009) consider that bank off balance sheet positions can contribute to liquidity creation. However, we cannot obtain precise breakdown for off-balance sheet. Thus, in our study we only consider the liquidity created from on balance sheet positions.

8

indicator that shows to what extent a bank is unable to meet its liquidity requirements without

borrowing money or fire selling its assets. Thus, we compute the net stable funding difference

by calculating the difference of the required amount of stable funding and the available

amount of stable funding. Based on the definition of the BIS (2009), the required amount of

stable funding corresponds to the amount of a particular asset that could not be monetised

through the sale or the use as collateral in a secured borrowing. The available amount of

stable funding corresponds to the total amount of an institution’s: i) capital; ii) liabilities with

effective maturities of one year or greater; and iii) a portion of “stable” non-maturity deposits

and / or term deposits with maturities of less than one year that would be expected to stay

within the institution. To calculate the net stable funding difference, a specific required stable

funding factor is assigned to each particular type of asset and a specific available stable

funding factor is assigned to each particular type of liability. In appendix 1 (see table A.1), we

briefly summarize the composition of assets and liabilities categories and related stable

funding factor as defined in the Basel III accords. Table 4 shows the breakdown of bank

balance sheet as provided by Bloomberg and its weighting in accordance with the proposals

made by the BIS (2009) to calculate the net stable funding difference. On the asset side, we

consider the type and the maturity of bank assets in line with the definition of the BIS (2009)

to put the corresponding weights. On the liability side, we consider the maturity of the several

fundings to put the corresponding weights. However, as we have only the breakdown for

deposits according to their maturity and not according to the type of depositors, we consider

the intermediate weight of 0.7 for stable demand and saving deposits (including all deposits

with a maturity of less than one year).

[Insert Table 4]

The net stable funding difference (NSFD) is then calculated as follows (where all components

are scaled by total assets): 0 * (cash + interbank assets + short term marketable assets) + 0.5 * (long term marketable assets + customer acceptances) 0.7 * (demand deposits + saving deposits) + 0.85 * consumer loans + 0 * (short term market debts + other short term liabilities)

required amount of stable funding

available amount of stable funding

+ 1 * (commercial loans + other loans + other assets + net fixed assets)

+ 1 * (long term liabilities + equity)

total assets total assets total assets total assets-NSFD = = -

If the difference is positive, it means that the required amount of stable funding exceeds the

available amount of stable funding. Thus, the bank faces transformation risk and may

9

experience liquidity problems to repay the funding exigible on demand with the assets that

cannot be monetised or that are only saleable at fire sale prices.

3.3. Statistical analysis of the liquidity creation (LC) and the net stable funding

difference (NSFD)

As we assume that bank illiquidity and transformation risk increase with liquidity

creation and that the net stable funding difference is a measure of the liquidity profile “at-

risk” of banks, we do a statistical analysis to appreciate the similarity of our proxy of liquidity

creation (LC) and of the net stable funding difference (NSFD). The aim is to emphasize the

positive relationship that may exist between these two variables. First, we calculate Pearson’s

coefficient of correlation. We also present scatters to visualise the linear relationship that may

exist between these two indicators. We do this statistical analysis by considering all banks in

our sample. In addition, we separate US and European banks in order to examine whether the

results are driven by US banks alone as they account for a large share of our sample. In table

5, we present descriptive statistics of our two indicators and Pearson’s coefficients of

correlation.

[Insert Table 5]

We notice that the average LC of all banks in our sample is 31.6% of total assets and the

average NSFD is -7.9% of total assets (see table 5). Pearson’s coefficient of correlation

suggests a strong linear and positive relationship between LC and NSFD (i.e., this coefficient

being at 0.72 and significant at 1% level). Figure 1 illustrates our findings by showing the

linear and positive relationship between LC and NSFD.

10

Figure 1: Scatter of LC and NSFD (in percent of total assets), for US and European

commercial banks from 2000 to 2008

Considering separately US and European banks, we notice that the average LC and the

average NSFD of European banks (at respectively, 32.4% and -0.2% of total assets) are

significantly higher than those of US banks (at respectively, 31.3% and -10.8% of total assets,

see table 5). However, the difference in average NSFDs is higher (i.e., the mean test statistic

being at 28.77) than the difference in average LCs (i.e., the mean test statistic being at 3.08).

Besides, Pearson’s coefficients of correlation emphasize the strong linear and positive

relationship between LC and NSFD whatever the location of banks. Figures 2 and 3 illustrate

our findings.

Figure 2: Scatter of LC and NFSD (in percent of total assets), for US commercial banks

from 2000 to 2008

11

Figure 3: Scatter of LC and NFSD (in percent of total assets), for European commercial

banks from 2000 to 2008

To deeper understand the higher difference in average NSFD than in average LC between US

and European banks, we do an average comparison of the components of LC and NSFD. The

liquidity creation of a bank is positive when its illiquid assets exceed its illiquid liabilities

(i.e., some illiquid assets being funded by liquid liabilities). Based on the liquidity creation

theory of Berger and Bouwman (2009), we calculate the amounts of illiquid assets (IA_TA)

and of illiquid liabilities (IL_TA), scaled by total assets7. In addition, as the net stable funding

difference is the difference of two components (i.e., the required amount of stable funding and

the available of stable funding), we calculate them separately (RSF_TA and ASF_TA, each

component being scaled by total assets)8. The average values of these ratios and the mean test

statistics for the null hypothesis of identical means between US and Europeans banks are

shown in table 6.

[Insert Table 6]

The average differences between US and European banks are significant whatever the ratio

considered (see table 6). However, the higher average difference is for ASF_TA, the mean test

statistic being the greatest at -52.25. Thus, European banks hold on average significantly less

7 IA_TA corresponds to all illiquid assets, i.e., to totally illiquid assets and to the semi liquid assets that are illiquid. IA_TA is the weighted sum of all illiquid assets, scaled by total assets. The weights are defined in reference to the liquidity creation theory of Berger and Bouwman (2009). We assign a weight of 1 to all illiquid assets and a weight of 0.5 to all semi liquid assets. IL_TA corresponds to all illiquid liabilities, i.e., to totally illiquid liabilities and to the semi liquid liabilities that are illiquid. IL_TA is the weighted sum of all illiquid liabilities, scaled by total assets. We assign a weight of 1 to all illiquid liabilities and a weight of 0.5 to all semi liquid liabilities. For further details about the breakdown of assets and liabilities by liquidity categories, see table 3.8 For further details about the computation of RSF_TA and ASF_TA, see table 4.

12

available stable funding than US banks. Consequently, this gap enables us to understand why

European banks have on average significantly higher NSFD than US banks. The difference

between US and European banks in NSFD arises from differences on the liability side of bank

balance sheet. To further investigate the components that drive this difference in ASF_TA and

to understand why the difference in IL_TA is not as important as the difference in ASF_TA

between US and European banks, we do an average comparison of the liquid versus illiquid

liabilities in LC9 and of the stable versus unstable funding in NSFD10. All components are

scaled by total assets. The average values of these ratios and the mean test statistics for the

null hypothesis of identical means between US and Europeans banks are shown in table 7 and

8.

[Insert Tables 7 and 8]

US banks are largely funded by deposits (77.4% of total assets) that contribute to a large share

of their liquid liabilities in LC. Indeed, liquid deposits that account for 60.2% of total assets

drive liquid liabilities that account for 69.6% of total assets. However, European banks are

less funded by deposits (51.2% of total assets) but they are largely funded by market debts

(39.8% of total assets) that contribute to a large share of their liquid liabilities in LC. Indeed,

liquid market funding that accounts for 30.7% of total assets drives liquid liabilities that

account for 73% of total assets (see table 7). In fact, the difference in average IL_TA is

significant between US and European banks but it is not so large (at respectively, 30.4% and

27% of total assets), the liquid liabilities of US banks including mostly deposits instead the

liquid liabilities of European banks that include both deposits and market debts. Besides,

regarding the NSFD, the difference of this indicator from LC is that the majority of deposits

that are qualified as liquid in LC are considered as stable in NSFD. Thus, although US banks

are largely funded by deposits, the average share of their unstable deposits (12.9% of total

9 Liquid liabilities correspond to all liquid liabilities and to the semi liquid assets that are liquid. It is the weighted sum of all liquid liabilities scaled by total assets. The weights are defined in reference to the liquidity creation theory of Berger and Bouwman (2009). We assign a weight of 1 to all liquid liabilities and a weight of 0.5 to all semi liquid liabilities. Illiquid liabilities correspond to the semi liquid assets that are illiquid and to all illiquid liabilities. It is the weighted sum of all illiquid liabilities scaled by total assets. We assign a weight of 0.5 to all semi liquid liabilities and a weight of 1 to all illiquid liabilities. For further details about the breakdown of liabilities by liquidity categories, see table 3.10 Stable liabilities correspond to the amount of liabilities that are likely to stay within the bank following a shock. It is the sum of all liabilities weighted by their corresponding stable funding factor. Unstable liabilities correspond to the amount of liabilities that are likely to be suddenly claimed on demand following a shock. It is the sum of all liabilities weighted by their unstable funding factor. For further details about the breakdown of liabilities according to the importance of their stability, see table 4.

13

assets) is weakly higher than this of European banks (10% of total assets). However, a large

share of market debts contributes to increase unstable funding for European banks. Unstable

market funding that accounts for 21.6% of total assets drives unstable funding that accounts

for 31.6% of total assets. However, for US banks, unstable market funding accounts for only

6.2% of total assets, total unstable funding accounting for 19.1% of total assets (see table 8).

Consequently, European banks hold higher average share of unstable funding driven by

market debts compared with US banks. In fact, European banks hold weakly higher share of

liquid liabilities in LC than US banks. However, they hold much more unstable funding in

NSFD than US banks. Thus, US banks benefit from the stability of their large deposit base

and face a highly negative average NSFD. European banks are more funded by volatile

market funding and face a weakly negative average NSFD.

Besides, depending on the size of the bank, the ability to access external funding may

differ as large banks have a larger access to financial markets compared with small banks.

Thus, our findings concerning the impact of market funding on the liquidity profile of banks

are likely to differ according to the size of banks. Based on Berger and Bouwman (2009) and

on IBCA criterion, a bank is considered as large if total assets are greater than one billion

USD. We do the statistical analysis only for US banks (our sample including 233 large banks

and 341 small banks in the US) as our sample of European banks mainly includes large banks

(170 large banks in 207 European banks). In table 9, we present descriptive statistics of LC

and NSFD and Pearson’s coefficients of correlation for separately large and small US banks.

[Insert Table 9]

We notice that the average LC and the average NSFD of large banks (at respectively, 32.1%

and -9% of total assets) are significantly higher than those of small banks (at respectively,

30.8% and -12.1% of total assets, see table 9). However, the difference in average NSFDs is

higher (i.e., the mean test statistic being at -9.21) than the difference in average LCs (i.e., the

mean test statistic being at -3.26). Besides, Pearson’s coefficients of correlation outline the

strong linear and positive relationship between LC and NSFD whatever the size of banks.

Figures 4 and 5 illustrate our findings.

14

Figure 4: Scatter of LC and NFSD (in percent of total assets), for large US commercial

banks from 2000 to 2008

Figure 5: Scatter of LC and NFSD (in percent of total assets), for small US commercial

banks from 2000 to 2008

Like above, to deeper understand the higher difference in average NSFD than in average LC

between large and small banks, we do an average comparison of the components of LC and

NSFD. Statistics and mean tests according to the size of banks are shown in table 10.

[Insert Table 10]

The average differences between large and small banks are significant whatever the ratio

considered (see table 10). However, the higher average difference is for ASF_TA, the mean

test statistic being the greatest at 24.55. Thus, small banks hold on average significantly more

available stable funding than large banks. Consequently, this gap enables us to understand

why large banks have on average significantly higher NSFD than small banks. To further

investigate the components that drive this difference in ASF_TA and to understand why the

15

difference in IL_TA is not as important as the difference in ASF_TA between large and small

banks, we do an average comparison of the liquid versus illiquid liabilities in LC and of the

stable versus unstable funding in NSFD. Statistics and mean tests according to the size of

banks are shown in table 11 and 12.

[Insert Tables 11 and 12]

We can do similar comments for large banks (respectively, small banks) as those ever done

for Europeans banks (respectively, US banks). Finally, large banks hold weakly higher share

of liquid liabilities in LC than small banks (see table 11). However, they hold much more

unstable funding in NSFD than small banks (see table 12). Thus, small banks benefit from the

stability of their large deposit base and face a highly negative average NSFD. Large banks are

more funded by volatile market funding and face a weakly negative average NSFD.

3.4. The transformation risk neutral level of liquidity creation

After emphasizing the strong linear and positive relationship between LC and NSFD

whatever the location and the size of banks, we consider the following relationship (equation

(1), subscripts i and t denoting bank and period respectively): t,it,it,i NSFD*LC ε+β+α= .

After testing for cross section and time fixed versus random effects, we introduce cross

section fixed effects in our regressions. We run regressions for all banks in our sample, for US

and European banks separately and for large versus small US banks. From this equation, we

can calculate a level of liquidity creation for a given level of net stable funding difference.

Thus, we can calculate the level of liquidity creation for which a bank is continuously able to

meet its liquidity requirements with its own liquid assets, i.e. when the net stable funding

difference is null. This level of liquidity creation is the “transformation risk neutral level of

liquidity creation” (TRNLC). It corresponds to average cross section fixed effects11.

Consequently, a given level of liquidity creation could be reached but the bank is able to meet

its liquidity requirements without borrowing money or fire selling its assets (i.e., the value of

assets that cannot be monetised equals the amount of available stable funding). Regression

results and estimations of TRNLC are shown in table 13.

11 Average cross section fixed effects are calculated as follows: ∑=

α+αN

1ii N/)( .

16

[Insert Table 13]

Our results show that the TRNLC of all banks in our sample is 37.5% of total assets. More

precisely, TRNLC of European banks is 32.6% and 40.3% for US banks. In addition among

US banks, TRNLC of small banks is 40.9% and 39.2% for large banks. Consequently, the

level of LC banks can perform by being continuously able to face transformation risk is lower

for European banks (respectively, large US banks) than for US banks (respectively, small US

banks). These findings can be explained in light of the conclusions of our statistical analysis.

For weakly different levels of LC, NSFD of European banks (respectively, large US banks) is

much higher than of US banks (respectively, small US banks). In other words, for weakly

different levels of LC, European banks (respectively, large US banks) face much higher levels

of transformation risk driven by the importance of their unstable market debts compared with

US banks (respectively, small US banks) that benefit from the stability of their large deposit

base. Consequently, European banks (respectively, large US banks) can create less liquidity

than US banks (respectively, small US banks) to be continuously able to face transformation

risk. This result confirms the necessity as ever pointed out by the Basel Committee (2009),

especially for European banks, to strengthen the stability of their funding to mitigate their

transformation risk.

4. The determinants of transformation risk and regression framework

According to our empirical issue, we consider indicators of concentration on loans that

are potentially securitisable and on short term, potentially unstable market funding in the

determination of transformation risk. In addition, based on previous studies (Deep and

Schaefer, 2004; Rauch et al., 2008; Berger and Bouwman, 2009; Pana et al., 2009; Choi et al.,

2009; Chen et al., 2010; Fungacova et al., 2010), we consider a set of other explanatory

variables. Finally, we discuss the regression framework.

4.1. Indicators of concentration on loans that are potentially securitisable and on

short term, potentially unstable market funding

Liquidity creation is an essential role of banks. However, through this function, banks

face transformation risk and may become fragile. Nevertheless, financial innovation provides

new asset - liability management (ALM) framework for banks to manage their liquidity and

17

mitigate liquidity pressures. Following financial globalisation and deregulation, banks have

largely enhanced their market activities by increasing their market funding and by securitizing

their loans (Shleifer and Vishny, 2009). The use of market funding reduces bank reliance on

deposits (Mishkin, 2004). In addition, the securitisation of loans is a source of cash as it

allows banks to convert some of their loans into liquid funds (Loutskina, 2011). In this

perspective, we study how transformation risk is impacted by the concentrations on loans that

are potentially securitisable and on short term, potentially unstable market funding.

By holding totally illiquid assets, banks may experience acute liquidity problems.

Nevertheless, although some assets are not totally liquid as they are not directly saleable on

financial markets (i.e., in opposition to cash, near cash items and trading securities), they can

be sold through OTC transactions such as the loans that are securitised. Based on this fact, we

question whether the concentration on loans that are potentially securitisable rather than on

totally illiquid assets is likely to mitigate liquidity pressures on banks and may decrease

transformation risk. As a proxy of the loans that are potentially securitisable, we consider the

consumer loans (such as loans to consumers, credit card loans, residential mortgage loans and

instalment loans). Indeed, consumer loans are securitisable through the issuance of residential

mortgage backed securities (RMBS). Commercial loans and other loans (such as loans to

commercial and industrial entities, commercial real estate loans, construction loans, loans to

agriculture and loans to money market funds) are not securitisable or only securitisable

through the issuance of commercial mortgage backed securities (CMBS). However, central

banks and prime brokers charge a higher discount on CMBS than on RMBS (IMF, 2008).

Consequently, the securitisation of consumer loans provides larger amounts of cash than the

securitisation of commercial loans and other loans. Thus, consumer loans are more liquid than

commercial ones. Therefore all else equal, a bank increases its exposure to transformation risk

by investing one dollar of liquid liabilities into one dollar of commercial loans rather than into

one dollar of consumer loans. Thus, we can expect a negative relationship between the

concentration on consumer loans that are potentially securitisable and transformation risk.

Another important issue in bank liquidity analysis is the stability of funding. Short

term debts are less stable than long term ones. Besides, short term deposits may be considered

as more stable than short term market debts (BIS, 2009). Consequently, the more banks hold

short term market debts, the greater is the potential instability of their funding. Thus, we can

expect that the concentration on short term, potentially unstable market funding rather than on

short term, stable deposits may increase liquidity pressures on banks and transformation risk.

However, banks may consider possible liquidity shortages on funding markets (i.e., some

18

market debts may be rolled-off at short notice) to limit their liquidity creation. Thus, we can

expect that the concentration on short term, potentially unstable market funding is likely to

discourage banks for increasing their liquidity creation that leads to lower exposure to

transformation risk. The impact on transformation risk of the concentration on short term,

potentially unstable market funding is ambiguous.

To measure such concentrations, we compute normalised Herfindalh Hirschman

indexes (Stiroh, 2002; Acharya et al., 2002). We consider two proxies of the concentration on

loans that are potentially securitisable. First, we consider the concentration on loans that are

potentially securitisable rather than on loans that cannot be securitised. We test if the potential

liquidity of the loan portfolio is likely to mitigate transformation risk. Thus, we compute a

normalised Herfindalh Hirschman index to proxy the level of concentration on loans that are

potentially securitisable versus on loans that cannot be securitised (HHI_LOAN12). Second, we

consider the concentration on loans that are potentially securitisable rather than on loans that

cannot be securitised and on other illiquid assets. We test if the potential liquidity of the

illiquid assets portfolio is likely to mitigate transformation risk. Consequently, we compute a

normalised Herfindalh Hirschman index to proxy the level of concentration on loans that are

potentially securitisable versus on totally illiquid assets (i.e., including all loans that cannot be

securitised, others assets and fixed assets, HHI_ILASSET). In addition, we calculate a

normalised Herfindalh Hirschman index to proxy the level of concentration on short term

deposits versus on short term market debts (HHI_STFUND). Normalised Herfindalh

Hirschman index varies between 0 and 1. The more the index is closed to 1, the higher is

concentration. Besides, the higher is the ratio of loans that are potentially securitisable

(respectively, the share of total short term market debts) to total loans or to total loans and

other illiquid assets (respectively, to total short term debts), the higher is bank concentration

on loans that are potentially securitisable (respectively, on short term market debts). To

capture the concentration on loans that are potentially securitisable, we interact HHI_LOAN

with the ratio of loans that are potentially securitisable to total loans (SECLO_TLO).

12 We split bank loan portfolio into loans that are potentially securitisable and loan that cannot be securitised. Herfindalh Hirschman index (HHI_L) is then computed as follows:

22 )()(L_HHI loans total / blesecuritisa not are that loansloans total / blesecuritisay potentiall are that loans +=

We calculate normalised LOAN_HHI as follows:

211

21L_HHI

LOAN_HHI−

−= )

We calculate the other indicators of concentration by considering the same methodology.

19

Similarly, we interact HHI_ILASSET with the ratio of loans that are potentially securitisable

to total loans and other illiquid assets (SECLO_IA). In addition, to capture the concentration

on short term market debts, we interact HHI_STFUND with the ratio of short term market

debts to total short term debts (SMDBT_STDBT). As we conjecture a positive relationship

between transformation risk and the concentration on loans that cannot be securitised or on

illiquid assets, we can expect a positive sign for the coefficients of HHI_LOAN and

HHI_ILASSET in the determination of transformation risk. Then, as we conjecture a negative

relationship between transformation risk and the concentration on loans that are potentially

securitisable, we can expect that the sign of the sum of coefficients of HHI_LOAN

(respectively, HHI_ILASSET) and of its interaction with SECLO_TLO (respectively,

SECLO_IA) tends to be more and more negative as SECLO_TLO (respectively, SECLO_IA)

tends to increase. However, the expected signs for the coefficient of HHI_STFUND and for

the sum of coefficients of HHI_STFUND and of its interaction with STMDBT_TDBT are

ambiguous.

4.2. Other variables impacting transformation risk

Following the existing literature, we consider a large set of microeconomic and

macroeconomic indicators that are likely to impact transformation risk.

Based on Berger and Bouwman (2009), we consider the influence of bank capital in

the determination of transformation risk. The authors point out two hypotheses that largely

matter in the current debate of the relationship between bank capital and liquidity creation.

The “risk absorption hypothesis” predicts a positive relationship between bank capital and

liquidity creation. Liquidity creation increases the bank’s exposure to transformation risk as

its losses increase with the level of illiquid assets to meet the liquidity demands of customers

(Allen and Gale, 2004), while capital allows the bank to absorb risk (Repullo, 2004). Thus,

higher capital ratio may allow banks to increase their liquidity creation and their exposure to

transformation risk. By contrast, the “financial fragility hypothesis” (Diamond and Rajan,

2000, 2001) and the “deposit crowding-out hypothesis” (Gorton and Winton, 2000) predict a

negative relationship between capital and liquidity creation. In their model, Diamond and

Rajan (2000, 2001) suggest that bank capital may impede liquidity creation by making the

bank’s capital structure less fragile. They model a relationship bank that raises funds from

depositors and lends then to borrowers. By monitoring borrowers, the bank obtains private

information that gives it an advantage in assessing the profitability of its borrowers. However,

20

this informational advantage may create an agency problem. Indeed, as the bank maximises

its profitability, it may extort rents from its depositors by demanding a greater share of the

loan income. As depositors know that the bank may abuse their trust, the bank has to win their

confidence by adopting a fragile financial structure with a large share of liquid deposits.

Nevertheless, a contract with depositors mitigates the bank’s hold-up problem because

depositors can run on the bank if they have doubts about bank efforts for monitoring

borrowers and the fair reallocation of loan income. Consequently, financial fragility favours

liquidity creation since it allows the bank to collect more deposits and grant more loans. By

contrast, higher capital tends to mitigate the financial fragility and enhances the bargaining

power of the bank that leads to hamper the credibility of its commitment to depositors.

Consequently, higher capital tends to decrease liquidity creation and exposure to

transformation risk. Besides, Gorton and Winton (2000) show that a higher capital ratio may

reduce liquidity creation through the crowding-out of deposits. They argue that deposits are

more effective liquidity hedges for investors than investments in bank equity capital. Indeed,

deposits are totally or partially insured and withdrawable at par value. However, bank capital

is not exigible and with a stochastic value that depends on the state of bank fundamentals and

on the liquidity of the stock exchange. Thus, higher capital ratios shift investors’ funds from

relatively liquid bank deposits to relatively illiquid bank capital. Consequently, the higher is

bank capital ratio, the lower is liquidity creation and bank exposure to transformation risk. In

our study, we consider the ratio of Tier 1 and 2 capital to total assets (T12_TA). We consider a

broad definition of capital in line with some of the theoretical studies. For example, Diamond

and Rajan (2001) indicate that capital in their analysis may be interpreted as equity and long

term debts, the sources of funds that cannot run on the bank. Under the “financial fragility

hypothesis” and the “deposit crowding-out hypothesis”, we can expect a negative sign for the

coefficient of bank capital ratio in the determination of transformation risk. However under

the “risk absorption hypothesis”, we can expect a positive sign. The expected sign for the

coefficient of this variable is ambiguous.

We consider bank profitability to account for the impact of better financial soundness

on bank risk bearing capacity and on their ability to create liquidity (Rauch et al., 2008; Chen

et al., 2010). By assuming that better financial soundness can enhance bank ability to take

risk, we can expect a positive relationship between bank profitability and transformation risk.

Nonetheless, it can also account for the “too big to fail” status of large banks and the problem

of “gamble for resurrection”. A bank can create liquidity and take more risk even if it is

currently in trouble in order to boost its profitability as it knows that it will be rescued in case

21

of failure; or in order to obtain higher expected profits and improve its financial statements.

Thus, we can expect a negative relationship between bank profitability and transformation

risk. As a proxy of bank profitability, we consider the return on assets that corresponds to the

ratio of net income to total assets (ROA). The expected sign for the coefficient of this variable

is ambiguous.

We consider the impact of credit risk in the determination of transformation risk (Deep

and Schaefer, 2004; Rauch et al., 2008; Berger and Bouwman, 2009; Fungacova et al., 2010).

The lower is credit risk, the more a bank can enhance its credit activities by continuously

meeting the capital at-risk standards. Consequently, better quality of loans may improve the

ability of banks to create liquidity that leads to increase their exposure to transformation risk.

Consequently, we can expect a negative relationship between credit risk and transformation

risk. As a proxy of the quality of bank loans, we consider the ratio of total provisions for loan

losses to total loans (PLL_TLO). We can expect a negative sign for the coefficient of this

variable in the determination of transformation risk.

Berger and Bouwman (2009) shed light on the impact of bank market power in the

determination of bank liquidity creation and transformation risk as it may impact the

availability of funding (Petersen and Rajan, 1995) and the split of loan portfolio (Berger et al.

2005). Greater market power may enable banks to enhance their liquidity creation by making

more loans and by attracting more funds (i.e., deposits or market funding). Thus, we can

expect that the higher is bank market power, the higher may be liquidity creation and

exposure to transformation risk. As a proxy of bank market power, we consider the ratio of

total assets of bank i located in country j to total assets of the banking system in country j

(MKT_POW). Thus, we can expect a positive sign for the coefficient of this variable in the

determination of transformation risk.

We consider bank size to control for possible data distortions due to size heterogeneity

since small banks are likely to be more focused on traditional intermediation activities (Rauch

et al., 2008; Choi et al., 2009; Berger and Bouwman, 2009; Fungacova et al., 2010). Thus, we

can expect a negative relationship between bank size and transformation risk. However, bank

size accounts for possible “too big to fail” status of large banks that could lead to moral

hazard behaviour and excessive risk exposure. We can expect a positive relationship between

bank size and transformation risk. As a proxy of bank size, we consider the log of total assets

(LN_TA). The expected sign for the coefficient of this variable is ambiguous.

The existing empirical literature about liquidity creation outlines the relevance of

macroeconomic indicators concomitantly to microeconomic indicators (Rauch et al., 2008;

22

Pana et al., 2009; Chen et al., 2010). Indeed, macroeconomic context is likely to impact bank

activities and investment decisions. For example, the demand for differentiated financial

products is higher during economic boom and may improve bank ability to expand its loan

and securities portfolios at a higher rate. Similarly, economic downturns are exacerbated by

the reduction in bank credit supply. Based on these arguments, we can expect banks to

increase their liquidity creation and their exposure to transformation risk during economic

booms. To highlight the impact of macroeconomic context on transformation risk of banks,

we consider the annual growth rate of real GDP (GDP_GWT). Thus, we can expect a positive

sign for the coefficient of this variable in the determination of transformation risk.

We consider the impact of monetary policy on bank liquidity creation and

transformation risk (Rauch et al., 2008). The literature provides two opposite views on the

link between monetary policy and bank liquidity creation (Mishkin, 1996). First, when central

bank policy rate is relatively low, credit supply increases as banks’ liquidity creation. We can

expect a negative relationship between central bank policy rate and transformation risk.

Second, when interest rates are relatively high and because of adverse selection problems,

demands for risky investment projects with higher expected returns may supplant safe

investment projects that generate low profitability. Thus, banks are likely to face higher losses

through the lower quality and the higher illiquidity of their assets. We can expect a positive

relationship between central bank policy rate and transformation risk. In our study, we

consider the policy rate of the central banks that are located in the countries considered (CB).

Under the first view, we can expect a negative sign for the coefficient of this variable in the

determination of transformation risk. Under the second view, we can expect a positive sign.

Thus, the expected sign for the coefficient of this variable is ambiguous.

As liquidity shortages are likely to disturb the management of bank liquidity and may

lead to acute liquidity problems, we consider the impact of liquidity pressures on the

interbank market. As a proxy of liquidity pressures on the interbank market, we consider the

spread of the one month interbank rate and the policy rate of the central bank (IBK1M_CB).

The higher is the spread, the higher is the one month interbank rate compared with the policy

rate of the central bank. Thus, the interbank market is under liquidity pressures. In addition,

this variable is an indicator of banks’ mistrust towards their peers (Estrella and Mishkin,

1998; Estrella, 2005). Banks charge a high risk premium for lending to the other banks at

short notice compared with the policy rate of the lender on last resort. Consequently, the

higher cost of interbank funding may prevent banks to access to these sources of liquidity.

Thus, banks may face losses from having to sell some assets at fire sale prices to repay the

23

liabilities claimed on demand. Consequently, we can expect a positive sign for the coefficient

of this variable in the determination of transformation risk.

Finally, we consider supervisory regime (Laeven and Levine, 2008; Shehzad et al.,

2010) as it is likely to impact bank risk taking behaviour. We compute the index measuring

supervisory control from the World Bank’s 2007 Regulation and Supervisory Database (Barth

et al., 2007)13. As banking regulation is likely to vary across countries, this indicator enables

us to control for possible country effects. We can expect that under strong supervisory

oversight, banks are encouraged to control their risk exposure and manage their

transformation risk. Thus, we can expect a negative sign for the coefficient of this variable in

the determination of transformation risk.

In table 14, we present some descriptive statistics of our explanatory variables for US

and European publicly traded commercial banks over the 2000-2008 period.

[Insert Table 14]

4.3. Econometric model

To study the determinants of transformation risk, we consider as dependant variable

the net stable funding difference (NSFD). It is an indicator of the liquidity profile “at-risk” of

banks from their liquidity creation activities. In addition of including the liquidity unbalances

of both sides of bank balance sheet, it accounts for the impact of the liquidity of the financial

markets, on the valuation of assets and the availability of funding, to assess bank exposure to

transformation risk. A positive NSFD implies that banks are likely to face too many losses

13 To compute our proxy of supervisory regime (CONTROL), we combine two indicators. The first indicator refers to supervisory agency control and is the total number of affirmative answers to the following questions: (i) Is the minimum capital adequacy requirement greater than 8%? (ii) Can the supervisory authority ask banks to increase minimum required capital in the face of higher credit risk? (iii) Can the supervisory authority ask banks to increase minimum required capital in the face of higher market risk? (iv) Can the supervisory authority ask banks to increase minimum required capital in the face of higher operational risk? (v) Is an external audit compulsory obligation for banks? (vi) Can the supervisory authority force a bank to change its internal organization structure? (vii) Can the supervisory authority legally declare that a bank is insolvent? (viii) Can the supervisory authority intervene and suspend some or all ownership rights of a problem bank? (ix) Can the supervisory authority supersede shareholders rights? (x) Can the supervisory authority remove and replace managers? (xi) Can the supervisory authority remove and replace directors? The second indicator of the supervisory regime measures deposit insurance agency control and is the total number of affirmative answers to the following questions: (i) Can the deposit insurance agency legally declare that a bank is insolvent? (ii) Can the deposit insurance agency intervene and suspend some or all ownership rights of a problem bank? (iii) Can the deposit insurance agency remove and replace managers? (iv) Can the deposit insurance agency remove and replace directors? (v) Can the deposit insurance agency supersede shareholders rights?

24

from having to sell some assets at fire sale prices to repay some liabilities that may be claimed

at short notice. These losses may prevent banks to repay this amount of debts as the cash

value of assets may be too weak.

Our empirical methodology is closed to Berger and Bouwman (2009). Based on the

fact that portfolio changes take time to occur and likely reflect decisions made on the basis of

historical experience, we consider the one year lagged value of all explanatory variables

(except for LN_TA and CONTROL following Rauch et al. (2008) and Fungacova et al.

(2010)). Like Berger and Bouwman (2009), we suppose that the future cannot cause the past.

In a risk management perspective, the purpose is to outline how previous factors influence

bank decisions to determine their current profile of liquidity creation and transformation risk.

Besides, as we consider two proxies of concentration on loans that are potentially

securitisable (i.e., HHI_LOAN and HHI_ILASSET) and because they are highly correlated, we

introduce them alternatively in our regressions (equations 2.a and 2.b). Our model is specified

as follows (subscripts i and t denoting bank and period respectively):

t,i

13

12pt,ip

11

5p1t,ip

1t,i1t,i41t,i3

1t,i1t,i21t,i1itit

TDTD

STDBT_STMDBT*STFUND_HHISTFUND_HHI

TLO_SECLO*LOAN_HHILOAN_HHINSFD

ε+β+β+

β+β+

β+β+α=

∑∑==

−

−−−

−−−

(2.a)

t,i

13

12pt,ip

11

5p1t,ip

1t,i1t,i41t,i3

1t,i1t,i21t,i1itit

TDTD

STDBT_STMDBT*STFUND_HHISTFUND_HHI

IA_SECLO*ILASSET_HHIILASSET_HHINSFD

ε+β+β+

β+β+

β+β+α=

∑∑==

−

−−−

−−−

(2.b)

Where HHI_LOAN and HHI_ILASSET are respectively normalised Herfindalh

Hirschman indexes that proxy the level of concentration on loans that are potentially

securitisable versus on loans that cannot be securitised or alternatively on illiquid assets (i.e.,

including all loans that cannot be securitised and other illiquid assets). HHI_STFUND is a

normalised Herfindalh Hirschman index that proxy the concentration on short term deposits

25

versus on short term market debts. SECLO_TLO corresponds to the ratio of loans that are

potentially securitisable to total loans. SECLO_IA is the ratio of loans that are potentially

securitisable to total loans and other illiquid assets. STMDBT_STDBT corresponds to the ratio

of short term market debts to total short term debts. TD corresponds to the determinants of

transformation risk from previous literature. After testing the presence of cross section and /

or time fixed versus random effects and possible heteroskedasticity of error, we introduce

cross section and time fixed effects in our regressions. To avoid colinearity problems, we

orthogonalise the correlated variables if their introduction disturbs the results of our

regressions14. To deal with heteroskedasticity problem, we use the Huber-White robust

covariance method.

5. Regression results

We test for the contribution of the concentrations on loans that are potentially

securitisable and on short term, potentially unstable market funding to explain transformation

risk beyond the factors documented in the literature. Equations 2.a and 2.b correspond to the

estimation of equation (2) by considering alternatively two proxies of the concentration on

loans that are potentially securitisable. Table 15 presents our regression results.

[Insert Table 15]

All proxies of concentration come out significant in the baseline of our estimations.

The coefficients of HHI_LOAN and HHI_ILASSET are significantly positive. The coefficient

of HHI_STFUND is significantly negative. Consequently, the higher is the concentration on

loans that cannot be securitised, the higher is transformation risk. In addition, the higher is the

concentration on short term, stable deposits, the lower is transformation risk. Besides, the

sums of coefficients of HHI_LOAN (respectively, HHI_ILASSET) and of its interaction with

SECLO_TLO (respectively, SECLO_IA) tends to become more and more negative as

SECLO_TLO (respectively, SECLO_IA) tends to increase. Furthermore, the sum of

coefficients of HHI_STFUND and of its interaction with STMDBT_TDBT tends to become

more and more positive as STMDBT_TDBT tends to increase. Thus, the more banks are

concentrated on loans that are potentially securitisable, the lower is transformation risk.

Besides, the higher is the concentration on short term, potentially unstable market funding, the

14 We orthogonalise LN_TA with MKT_POW.

26

higher is transformation risk. These findings highlight the benefits of the concentration on

loans that are potentially securitisable and on short term deposits to mitigate transformation

risk. These results outline the advantages for banks to develop the securitisation of loans in

order to benefit from the liquidity of securitisation markets to manage their transformation

risk. However, the advantages provided by the securitisation of loans depend on the liquidity

of securitisation markets that is likely to be impugned following a market collapse (i.e., like

during the Subprime crisis). Thus, holding such loans is likely to be inefficient to manage

transformation risk when the liquidity of securitisation markets is tightening. Besides, our

findings emphasize the benefit, as ever pointed out by the Basel Committee (2009), of the

stability of funding to mitigate transformation risk. These results outline the negative impact

of the liquidity shortages on funding markets as the higher instability of short term market

funding is likely to increase transformation risk.

Concerning the other determinants of transformation risk documented in previous

literature, most variables are significant except bank profitability (ROA). The coefficient of

the ratio of Tier 1 and 2 capital to total assets (T12_TA) is significantly negative.

Consequently, it is the “financial fragility hypothesis” and “the debt crowding hypothesis”

that seem to prevail. Thus, higher capital ratio may hamper liquidity creation and mitigate

transformation risk, banks benefiting from the stability of their liabilities. The coefficient of

the ratio of total provisions for loan losses to total loans (PLL_TLO) is significantly negative.

Consequently, under reduced credit risk, the ability of banks to create liquidity may be

improved, their exposure to transformation risk tending to be higher. Moreover, the

coefficients of the log of total assets (LN_TA) and the coefficient of our proxy of bank market

power (MKT_POW) are significantly negative. Thus, small banks tend to face higher

transformation risk. This result may be explained as they are likely to be focused on

traditional intermediation activities because of their restricted access to financial markets

compared with large banks. We can conjecture similar conclusions for banks with weak

market power as they are likely to be the smaller ones. Besides, the coefficient of our index of

supervisory regime (CONTROL) is significantly negative. Thus, the stronger is supervisory

oversight, the more banks are encouraged to control their risk exposure and face lower

exposure to transformation risk. Furthermore, our findings point out the importance to

consider macroeconomic indicators in the analysis of transformation risk. The coefficient of

the annual growth rate of real GDP (GDP_GWT) is significantly positive. Consequently,

during economic booms, banks may expand their loan and securities portfolios and increase

their transformation risk. In addition, the coefficient of the central bank policy rate (CB) is

27

significantly positive. This result implies that when interest rates are relatively high and

because of adverse selection problems, banks are likely to face higher risk exposure through

the lower quality and the higher illiquidity of their assets. Thus, higher interest rates

contribute to increase transformation risk. Finally, the coefficient of our proxy of the liquidity

pressures on the interbank market (IBK1M_CB) is significantly positive. The higher are

liquidity pressures on the interbank market, the higher is the cost of interbank funding. This

may prevent banks to access to these sources of liquidity and increase their transformation

risk. This finding highlights the importance of considering the state of the interbank market

for the analysis of transformation risk. Furthermore, it emphasizes the importance of the

effective transmission of monetary policy to the interbank market.

6. Robustness checks

We examine the robustness of our findings considering the impact of bank location on

the relationship between transformation risk and the concentrations on loans that are

potentially securitisable and on short term, potentially unstable market funding. We estimate

equation (2) separately for US and European banks (see table A2.1) in order to examine

whether the results are driven by US banks alone as they account for a large share of our

sample. The conclusions for all indicators of concentration are consistent with those

previously obtained whatever the location of banks.

We now check the stability of our results considering bank size. Depending on the size

of the bank, the ability to access external funding may differ. Large banks may benefit from

their “too big to fail status” and from their larger access to financial markets. By contrast,

small banks have a more restricted access to financial markets. This is likely to impact the

link between transformation risk and the concentrations on loans that are potentially

securitisable and on short term, potentially unstable market funding. Based on Berger and

Bouwman (2009) and on IBCA criterion, a bank is considered as large if total assets are

greater than one billion USD. Besides, as our sample of European banks mainly includes large

banks (170 large banks in 207 European banks), we also consider the location of banks.

Consequently, we consider bank size but we also check the stability of our results for banks

located in Europe versus in the US. Thus, we consider a dummy variable that takes the value

of 1 for US banks and 0 otherwise (DUM_LOC). We estimate equation (2) separately for

large and small banks and we introduce all the interactions of each explanatory variable with

this dummy variable and the dummy variable alone (see table A2.2). The conclusions for all

28

indicators of concentration are consistent with those previously obtained whatever the size

and the location of banks, except in one case. For small banks whatever their location, the

concentration on short term, potentially unstable market funding becomes not significant in

the determination of transformation risk. Consequently, transformation risk of small banks is

not impacted by the potential instability of their short term funding. However, small banks are

largely funded by deposits15. Our results emphasize the benefit of the stability of their large

deposit base, and especially from the stability of their short term deposits, to mitigate their

transformation risk.

7. Concluding remarks

Liquidity creation is an essential role of banks but a major source of their vulnerability

to shocks. Thus, it is not liquidity creation that may be damaging for banks but excessive

liquidity creation. In this perspective, we consider a measure of bank liquidity profile “at-risk”

as defined in the Basel III accords (called the “net stable funding difference”) that enables us

to assess a level of liquidity creation beyond which a bank may not be able to meet its

liquidity requirements without borrowing money or fire selling its assets (called the

“transformation risk neutral of liquidity creation”). Besides, although through their function

of liquidity creation, banks face transformation risk and may become fragile, financial

innovation provides new ways for banks to manage their liquidity and mitigate liquidity

pressures. Consequently, we study how transformation risk is decreased by the concentration

on loans that are potentially securitisable and impacted by the concentration on short term,

potentially unstable market funding.

Our findings confirm the similarity of the liquidity creation indicator of Berger and

Bouwman (2009) and the net stable funding difference and exhibit a strong linear and positive

relationship between these two indicators. We find that on average, the transformation risk

neutral of liquidity creation of US and European banks is 37.5% of total assets over the

2000-2008 period. More precisely, the average transformation risk neutral of liquidity

creation of European banks (respectively, large US banks) is lower than this of US banks

(respectively, small US banks). US banks (respectively, small US banks) benefit from the

stability of their large deposit base, European banks (respectively, large US banks) being

more funded by volatile market debts. Consequently, European banks (respectively, large US

15 The average share of total deposits in total assets is 80% for small US banks and of 68.7% for small European banks.

29

banks) can create less liquidity than US banks (respectively, small US banks) to be

continuously able to face transformation risk. Furthermore, our results show that the more

banks are concentrated on loans that are potentially securitisable and on short term deposits,

the lower is their transformation risk. These results outline the advantages for banks to

develop the securitisation of loans in order to benefit from the liquidity of securitisation

markets to manage their transformation risk. However, the benefits provided by loan

securitisation depend on the liquidity of securitisation markets that is likely to be impugned

following a market collapse. Consequently, holding such loans is likely to be inefficient to

manage transformation risk when securitisation markets become more illiquid. Besides, our

findings emphasize the benefit, as ever pointed out by the Basel Committee (2009), of the

stability of funding to mitigate transformation risk. These results shed light on the negative

impact of the liquidity shortages on funding markets that increase the instability of bank

funding and transformation risk.

Finally, the transformation risk neutral of liquidity creation may be useful to add to the

debate on liquidity assessment in banking. In a prudential approach, this level of liquidity

creation could be considered to appreciate the ability of banks to face transformation risk

when they create liquidity. In addition, by better understanding what factors significantly

impact bank exposure to transformation risk, it can help banks to improve their risk

management framework.

30

ReferencesAcharya V., Hasan I., Saunders A., 2002, “Should banks be diversified? Evidence

from individual bank loan portfolios”, Bank of International Settlements, Working Paper.

Allen F., Gale. D., 2004, “Financial Intermediaries and Markets”, Econometrica, No

72, pp. 1023-1061.

Barth J. R., Caprio G., Levine R., 2007, “The regulation and supervision of banks

around the world - a new database”, The World Bank, Policy Research Working Paper.

Berger A. N., Miller N. H., Petersen M. A., Rajan R. G., Stein J. C., 2005, “Does

function follow organizational form? Evidence from the lending practices of large and small

banks”, Journal of Financial Economics, No 76, pp. 237-269.

Berger A. N., Bouwman C. H. S., 2009, “Bank liquidity creation”, Review of

Financial Studies, Vol 22, No 9, pp. 3779-3837.

Berger A. N., Bouwman C. H. S., 2009, “Bank liquidity creation, monetary policy and

financial crises”, Working Paper.

Berger A. N., Bouwman C. H. S., 2010, “Bank liquidity creation and risk taking

during distress”, Working Paper.

Blake J., Poorman F., 2005, “Measuring and modelling liquidity risk: New ideas and

metrics,” Financial Managers Society Inc., Woking Paper.

Bryant J., 1980, “A model of reserves, bank runs, and deposit insurance”, Journal of

Banking and Finance, Vol 4, pp. 335-344.

Chen Y. K., Shen C. H., Kao L. F., Yeh C. Y., 2010, “Bank liquidity risk and

performance”, Working Paper.

Choi P. P., Park J., Ho C., 2009, “Insurer liquidity creation: The evidence from U.S.

property and liability insurance industry”, Working Paper.

Cornett M. M., Saunders A., 2006, “Financial institutions management: a risk

management approach”, McGraw-Hill Boston, Working Paper.

Demirguc-Kunt A., 1990, “Deposit institutions failures: A review of empirical

literature”, Federal Reserve of Cleveland, Working Paper.

Diamond D.W., Dybvig P. H., 1983, “Bank runs, deposit insurance, and liquidity”,

Journal of Political Economy, No 91, pp 401-419.

Diamond D. W., Rajan R. G., 2000, “A theory of bank capital”, Journal of Finance,

No 55, pp. 2431-2465.

31

Diamond D. W., Rajan R. G., 2001, “Liquidity risk, liquidity creation, and financial

fragility: a theory of banking”, Journal of Political Economy, Vol 109, No 2, pp. 287-327.

Diamond D. W., Rajan R. G., 2001, “Banks and liquidity”, American Economic

Review, Vol 91, No 2, pp. 422-425.

Deep A., Schaefer G., 2004, “Are banks liquidity transformers?”, Harvard University

Faculty Research, Working Paper.

Estrella A., Mishkin F. S., 1998, “Predicting US recessions: Financial variables as

leading indicators”, Review of Economics and Statistics, No 80, pp. 45-61.

Estrella A., 2005, “The yield curve as a leading indicator: Frequently asked question”,

Federal Reserve of New York, Working Paper.

Fungacova F., Weill L., Mingming Z., 2010, “Bank capital, liquidity creation and

deposit insurance”, Bank of Finland, Working Paper.

“Global financial stability report”, 2008 (October), International Monetary Fund.

Gorton G., Winton A., 2000, “Liquidity provision, bank capital, and the

macroeconomy”, University of Minnesota, Working Paper.

Harvey J., Spong K., 2001, “The decline in core deposits: What can do banks?”,

federal Reserve Bank of Kansas City, Working Paper.

Holmstrom B., Tirole J., 1998, “Private and public supply of liquidity”, Journal of

Political Economy, Vol 106, No 1, pp.1-40.

Hull J., Rothman J. L., 2010, “Credit ratings and the securitisation of loans”, Federal

Reserve Bank of Atlanta, Working Paper.

“International framework for liquidity risk, measurement and monitoring”, 2009, Bank

of International Settlements, Consultative Document.

Kashyap A. K., Rajan R. G., Stein J. K., 2002, “Banks as liquidity providers: an

explanation for the coexistence of lending and deposit taking”, Journal of Finance, Vol 58,

No 1, pp. 33-73.

Laeven L. A., Levine R., 2008, “Bank governance, regulation and risk taking”, NBER,

Working Paper.

Loutskina E., 2011, “The role of securitisation in bank liquidity and funding

management”, Journal of Financial Economics, Vol 100, No 3, pp. 663-684.

Mishkin F.S., 1996, « Les canaux de transmission de la politique monétaire: Leçons

pour la politique monétaire », Bulletin de la Banque de France, No 27, Banque de France.

Mishkin F., 2004, Money, bank and financial market, Pearson Education, 7th Ed.

32

Pana E., Park J., Query T., 2009, “The impact of mergers on liquidity creation”,

Working Paper.

Petersen M. A., Rajan R. G., 1995, “The effect of credit market competition on

lending relationships”, Quarterly Journal of Economics, No 110, pp. 407-443.

Rauch C., Steffen S., Hackethal A., Tyrell M., 2008, “Determinants of bank liquidity

creation - evidence from savings banks”, Working Paper.

Repullo R., 2004, “Capital requirements, market power, and risk taking in banking”,

Journal of Financial Intermediation, No 13, pp.156-182.

Shehzad C. T., De Haan J., Scholtens B., 2010, “The impact of bank ownership

concentration on impaired loans and capital adequacy”, Journal of Banking and Finance, No

34, pp. 399-408.

Shleifer A., Vishny R. W., 2009, “Unstable banking”, Journal of Financial

Economics, Vol 91, No 2, pp. 77-95.

Stiroh K., 2002. “Diversification in banking: Is non interest income the answer?”,

Journal of Money, Credit and Banking, Blackwell Publishing, Vol. 36, No 5, pp. 853-882.

33

Table 1: Distribution of US and European commercial banks

Banks available in Bloomberg

Banks included in

our final sample

Total assets of banks in final

sample / total assets of the banking

systemUnited States 645 574 66.4Europe 225 207 60.4Austria 8 8 57.3Belgium 4 3 80.3Cyprus 4 4 69.7Denmark 44 38 60.6Finland 2 2 71.2France 22 22 62.1Germany 15 14 40.1Greece 12 12 80.6Iceland 2 2 66.3Ireland 3 3 31.3Italy 24 22 59.6Liechtenstein 2 2 50.1Malta 4 4 32.5Netherlands 2 2 47.6Norway 23 20 70.3Portugal 6 6 55.3Spain 15 15 64.4Sweden 4 4 72.6Switzerland 22 18 74.8United Kingdom 7 6 61.5

Source: Bloomberg, European Central Bank, Bank of England, National Bank of Switzerland, Sveriges Riskbank, Danmarks Nationalbank, Central Bank of Iceland, FDIC and Finance Norway (2000-2008). To deal with the issue of sample representativeness, we compare aggregate total assets of banks included in our final sample (i.e., US and European publicly traded commercial banks) to aggregate total assets of the whole banking system.

Table 2: General statistical description of our data set of US and European commercial banks, on average from 2000 to 2008

Total assets in billion

USD

Total loans / total assets

Total deposits /

total assets

Provisions for loan losses /

total loans

Tier 1 capital /

total assets

Tier 1 & 2 capital /

RWAROA

Total interest income /

total income

Mean 48.9 66.4 70.2 0.5 8.2 13.2 0.9 72.3 Median 1.1 68.3 75.4 0.3 7.7 12.5 0.9 75.9 Max 3768.2 95.1 96.0 7.2 35.2 34.0 6.9 100.0 Min 0.02 3.7 4.1 -1.2 0.1 4.5 -13.3 4.7 Std. Dev. 222.5 14.2 17.0 0.6 3.4 3.3 0.9 15.6

Source: Bloomberg (2000-2008). All variables are expressed in percentage (except Total assets). Total assets in billion USD; Total loans / total assets: (commercial loans + consumer loans + other loans) / total assets; Total deposits / total assets: (demand deposits + saving deposits + time deposits + other time deposits) / total assets; Total provisions for loan losses / total loans: total provisions for loan losses / (commercial loans + consumer loans + other loans); Tier 1 capital / total assets: Tier 1 capital / total assets; Tier 1 & 2 capital / RWA: (tier 1 capital + tier 2 capital) / total risk weighted assets; ROA: net income / total assets; Total interest income / total income: (interest income from loans + resale agreements + interbank investments + other interest income or losses) / total income.

34

Table 3: Balance sheet weighting used to calculate the liquidity creation

Assets Liquidity level Weights

Cash and near cash items Liquid -0.5Interbank assets Semi Liquid 0Short term marketable assets Liquid -0.5Commercial loans Illiquid 0.5Consumer loans Semi Liquid 0Other loans Semi Liquid 0Long term marketable assets Semi Liquid 0Net fixed assets Illiquid 0.5Other assets Illiquid 0.5Custumer acceptances Semi Liquid 0

LiabilitiesDemand deposits Liquid 0.5Saving deposits Liquid 0.5Time deposits Semi Liquid 0Other term deposits Semi Liquid 0Short term borrowings Liquid 0.5Other short term liabilities Liquid 0.5Long term borrowings Semi Liquid 0Other long term liabilities Semi Liquid 0Subordinated debentures Illiquid -0.5Prefered equity Illiquid -0.5Minority interests Illiquid -0.5Shareholder common capital Illiquid -0.5Retained earnings Illiquid -0.5

35

Table 4: Balance sheet weighting used to calculate the net stable funding difference

Assets Corresponding definition from the BIS Weights

Cash and near cash items Cash 0

Interbank assets Non renewable loans to financials with remaining maturity < 1 yr 0

Marketable securities and other short term investments

Short term unsecured actively traded instruments (with remaining maturity < 1 yr)

0

Commercial loans All other assets 1

Consumer loans Loans to retail clients (with remaining maturity < 1 yr) 0.85

Other loans All other assets 1

Long term investment

Unemcumbered listed equity or non financial senior unsecured corporate bonds rated at least A- (with remaining maturity < 1 yr)

0.5

Net fixed assets All other assets 1

Other assets All other assets 1

Custumer acceptances

Unemcumbered listed equity or non financial senior unsecured corporate bonds rated at least A- (with remaining maturity < 1 yr)

0.5

Liabilities Corresponding definition from the BIS Weights

Demand deposits 0.7

Saving deposits 0.7

Time deposits Other liabilities with an effective maturity > 1 yr 1

Other term deposits Other liabilities with an effective maturity > 1 yr 1

Short term borrowings All other liabilities or equity not included above 0

Other short term liabilities All other liabilities or equity not included above 0

Long term borrowings Other liabilities with an effective maturity > 1 yr 1

Other long term liabilities Other liabilities with an effective maturity > 1 yr 1

Subordinated debentures 1

Prefered equity 1

Minority interests 1

Shareholder common capital 1

Retained earnings 1

70% of deposits of retail and small business customers (non-maturity or residual maturity < 1yr)

Tier 1 & 2 capital instruments, other preferred shares and capital instruments in excess of Tier 2 allowable amount having an effective maturity > 1 yr

Required amount of stable funding

Available amount of stable funding

Note: The net stable funding difference (NSFD) is the difference of the required amount of stable funding and the available amount of stable funding, scaled by total assets. It is based on the net stable funding ratio as defined in the Basel III accords (BIS, 2009). For further details about the weighting of bank balance sheet as suggested by the BIS (2009) to compute this ratio, see appendix 1.

36

Table 5: Statistical analysis of the liquidity creation (LC) and the net stable funding difference (NSFD), for US and European banks from 2000 to 2008

Mean Std Dev Mean Std Dev

All banks 31.6 12.7 -7.9 14.0 0.72 ***(82.81)

US banks 31.3 13.2 -10.8 11.3 0.82 ***(98.45)

European banks 32.4 11.5 -0.2 17.1 0.69 ***(39.12)

Test statistic & %level

3.08 ***(0.00)

1.31 ***(0.00)

28.77 ***(0.00)

2.28 ***(0.00) -

LC NSFD Pearson coefficient of correlation

Note: All variables are expressed in percentage. LC: liquidity creation / total assets, for further details about the computation of LC, see table 3; NSFD: (required amount of stable funding - available amount of stable funding) / total assets, for further details about the computation of NSFD, see table 4. T-statistics test for null hypothesis of identical means or null Pearson’s coefficient of correlation; *, **, *** indicate statistical significance respectively at the 10%, 5%, and 1% level, for bilateral test.

Table 6: Average comparison of the components of the liquidity creation (LC) and the net stable funding difference (NSFD), for US and European banks from 2000 to 2008

IA_TA IL_TA RSF_TA ASF_TAUS banks 61.7 30.4 70.1 80.9

European banks 59.5 27.0 68.2 68.4

Test statistic & %level

6.87 ***(0.00)

-16.41 ***(0.00)

-5.49 ***(0.00)

-52.25 ***(0.00)

LC NSFD

Note: All variables are expressed in percentage. IA_TA corresponds to all illiquid assets, i.e., to totally illiquid assets and to the semi liquid assets that are illiquid. To calculate illiquid assets, we assign a weight of 1 to all illiquid assets and a weight of 0.5 to all semi liquid assets; IA_TA: (0*liquid assets + 0.5*semi liquid assets + 1*illiquid assets) / total assets. IL_TA corresponds to all illiquid liabilities, i.e., to totally illiquid liabilities and to the semi liquid liabilities that are illiquid. To calculate illiquid liabilities, we assign a weight of 1 to all illiquid liabilities and a weight of 0.5 to all semi liquid liabilities; IL_TA: (0*liquid liabilities + 0.5*semi liquid liabilities + 1*illiquid liabilities) / total assets. For further details about the breakdown of assets and liabilities by liquidity categories, see table 3. RSF corresponds to the required amount of stable funding, i.e., the amount of assets that cannot be readily monetised. It is the sum of all assets weighted by their corresponding required stable funding factor; RSF: (0*cash and near cash items + 0*interbank assets + 0*marketable assets and other short term investments + 1*commercial loans + 0.85*consumer loans + 1*other loans + 0.5*long term investments + 1*net fixed assets + 1*other assets + 0.5*customer acceptances) / total assets. ASF corresponds to the available amount of stable funding, i.e., the amount of liabilities that are likely to stay within the bank following a shock. It is the sum of all liabilities weighted by their corresponding stable funding factor; ASF: (0.7*demand and saving deposits + 1*time and other time deposits + 0*short term borrowings and other short term liabilities + 1*long term borrowings and other long term liabilities + 1*subordinated debts + 1*equity) / total assets. For further details about the breakdown of assets and liabilities according to the importance of their stability, see table 4. T-statistics test for null hypothesis of identical means; *, **, *** indicate statistical significance respectively at the 10%, 5%, and 1% level, for bilateral test.

37

Table 7: Average comparison of liquid versus illiquid liabilities in the liquidity creation (LC), for US and European banks from 2000 to 2008

DEPO_L DEPO_IL STMD_L LTMD_L LTMD_IL K_ILUS banks 60.2 17.1 6.2 3.3 3.3 10.0

European banks 42.3 9.0 21.6 9.1 9.1 9.0

Test statistic & %level

-55.06 ***(0.00)

-44.53 ***(0.00)

60.05 ***(0.00)

47.75 ***(0.00)

47.75 ***(0.00)

-10.31 ***(0.00)

Note: All variables are expressed in percentage. Liquid liabilities correspond to all liquid liabilities and to the semi liquid assets that are liquid. To calculate liquid liabilities, we assign a weight of 1 to all liquid liabilities and a weight of 0.5 to all semi liquid liabilities. Illiquid liabilities correspond to the semi liquid assets that are illiquid and to all illiquid liabilities. To calculate illiquid liabilities, we assign a weight of 0.5 to all semi liquid liabilities and a weight of 1 to all illiquid liabilities. For further details about the breakdown of liabilities by liquidity categories, see table 3. DEPO_L: (1*demand and saving deposits + 0.5*time and other time deposits) / total assets; DEPO_IL: 0.5*time and other time deposits / total assets; STMD_L: 1*short term borrowings and other short term liabilities / total assets; LTMD_L: 0.5*long term borrowings and other long term liabilities / total assets; LTMD_IL: 0.5*long term borrowings and other long term liabilities / total assets; K_IL: 1*equity / total assets. T-statistics test for null hypothesis of identical means; *, **, *** indicate statistical significance respectively at the 10%, 5%, and 1% level, for bilateral test.

Table 8: Average comparison of stable versus unstable liabilities in the net stable funding difference (NSFD), for US and European banks from 2000 to 2008

DEPO_STB DEPO_USTB STMD_USTB LTMD_STB K_STBUS banks 64.4 12.9 6.2 6.5 10.0

European banks 41.3 10.0 21.6 18.2 9.0

Test statistic & %level

-77 ***(0.00)

-24.51 ***(0.00)

60.05 ***(0.00)

47.75 ***(0.00)

-10.31 ***(0.00)

Note: All variables are expressed in percentage. Stable liabilities correspond to the amount of liabilities that are likely to stay within the bank following a shock. It is the sum of all liabilities weighted by their corresponding stable funding factor. Unstable liabilities correspond to the amount of liabilities that are likely to be suddenly claimed on demand following a shock. It is the sum of all liabilities weighted by their unstable funding factor. For further details about the breakdown of liabilities according to the importance of their stability, see table 4. DEPO_STB: (0.7*demand and saving deposits + 1*time and other time deposits) / total assets; DEPO_USTB: 0.3* demand and saving deposits / total assets; STMD_USTB: 1*short term borrowings and other short term liabilities / total assets; LTMD_STB: 1*long term borrowings and other long term liabilities / total assets; K_STB: 1*equity / total assets. T-statistics test for null hypothesis of identical means; *, **, *** indicate statistical significance respectively at the 10%, 5%, and 1% level, for bilateral test.

Table 9: Statistical analysis of the liquidity creation (LC) and the net stable funding difference (NSFD), for US banks by size from 2000 to 2008

Mean Std Dev Mean Std Dev

Large banks 32.1 12.2 -9.0 11.5 0.81 ***(60.40)

Small banks 30.8 13.8 -12.1 11.0 0.84 ***(80.37)

Test statistic & %level

-3.26 ***(0.00)

1.29 ***(0.00)

-9.21 ***(0.00)

1.08 ***(0.04) -

LC NSFD Pearson coefficient of correlation

Note: All variables are expressed in percentage. A bank is considered as large if its total assets is greater than one billion USD. LC: liquidity creation / total assets, for further details about the computation of LC, see table 3; NSFD: (required amount of stable funding - available amount of stable funding) / total assets, for further details about the computation of NSFD, see table 4. T-statistics test for null hypothesis of identical means or null Pearson’s coefficient of correlation; *, **, *** indicate statistical significance respectively at the 10%, 5%, and 1% level, for bilateral test.

38

Table 10: Average comparison of the components of the liquidity creation (LC) and the net stable funding difference (NSFD), for US banks by size from 2000 to 2008

IA_TA IL_TA RSF_TA ASF_TALarge banks 61.4 29.3 69.4 78.4Small banks 62.0 31.1 70.7 82.7Test statistic & %level

1.62(0.11)

9.02 ***(0.00)

3.89 ***(0.00)

24.55 ***(0.00)

LC NSFD

Note: All variables are expressed in percentage. A bank is considered as large if its total assets is greater than one billion USD. IA_TA corresponds to all illiquid assets, i.e., to totally illiquid assets and to the semi liquid assets that are illiquid. To calculate illiquid assets, we assign a weight of 1 to all illiquid assets and a weight of 0.5 to all semi liquid assets; IA_TA: (0*liquid assets + 0.5*semi liquid assets + 1*illiquid assets) / total assets. IL_TA corresponds to all illiquid liabilities, i.e., to totally illiquid liabilities and to the semi liquid liabilities that are illiquid. To calculate illiquid liabilities, we assign a weight of 1 to all illiquid liabilities and a weight of 0.5 to all semi liquid liabilities; IL_TA: (0*liquid liabilities + 0.5*semi liquid liabilities + 1*illiquid liabilities) / total assets. For further details about the breakdown of assets and liabilities by liquidity categories, see table 3. RSF corresponds to the required amount of stable funding, i.e., the amount of assets that cannot be readily monetised. It is the sum of all assets weighted by their corresponding required stable funding factor; RSF: (0*cash and near cash items + 0*interbank assets + 0*marketable assets and other short term investments + 1*commercial loans + 0.85*consumer loans + 1*other loans + 0.5*long term investments + 1*net fixed assets + 1*other assets + 0.5*customer acceptances) / total assets. ASF corresponds to the available amount of stable funding, i.e., the amount of liabilities that are likely to stay within the bank following a shock. It is the sum of all liabilities weighted by their corresponding stable funding factor; ASF: (0.7*demand and saving deposits + 1*time and other time deposits + 0*short term borrowings and other short term liabilities + 1*long term borrowings and other long term liabilities + 1*subordinated debts + 1*equity) / total assets. For further details about the breakdown of assets and liabilities according to the importance of their stability, see table 4. T-statistics test for null hypothesis of identical means; *, **, *** indicate statistical significance respectively at the 10%, 5%, and 1% level, for bilateral test.

Table 11: Average comparison of liquid versus illiquid liabilities in the liquidity creation (LC), for US banks by size from 2000 to 2008

DEPO_L DEPO_IL STMD_L LTMD_L LTMD_IL K_ILLarge banks 58.0 15.6 8.9 3.8 3.8 9.9Small banks 61.8 18.2 4.2 2.9 2.9 10.1

Test statistic & %level

13.40 ***(0.00)

14.32 ***(0.00)

-28.17 ***(0.00)

-10.93 ***(0.00)

-10.93 ***(0.00)

2.05 **(0.04)

Note: All variables are expressed in percentage. A bank is considered as large if its total assets is greater than one billion USD. Liquid liabilities correspond to all liquid liabilities and to the semi liquid assets that are liquid. To calculate liquid liabilities, we assign a weight of 1 to all liquid liabilities and a weight of 0.5 to all semi liquid liabilities. Illiquid liabilities correspond to the semi liquid assets that are illiquid and to all illiquid liabilities. To calculate illiquid liabilities, we assign a weight of 0.5 to all semi liquid liabilities and a weight of 1 to all illiquid liabilities. For further details about the breakdown of liabilities by liquidity categories, see table 3. DEPO_L: (1*demand and saving deposits + 0.5*time and other time deposits) / total assets; DEPO_IL: 0.5*time and other time deposits / total assets; STMD_L: 1*short term borrowings and other short term liabilities / total assets; LTMD_L: 0.5*long term borrowings and other long term liabilities / total assets; LTMD_IL: 0.5*long term borrowings and other long term liabilities / total assets; K_IL: 1*equity / total assets. T-statistics test for null hypothesis of identical means; *, **, *** indicate statistical significance respectively at the 10%, 5%, and 1% level, for bilateral test.

Table 12: Average comparison of stable versus unstable liabilities in the net stable funding difference (NSFD), for US and European banks from 2000 to 2008

DEPO_STB DEPO_USTB STMD_USTB LTMD_STB K_STBLarge banks 61.0 12.7 8.9 7.6 9.9Small banks 66.9 13.1 4.2 5.7 10.1

Test statistic & %level

21.16 ***(0.00)

2.95 ***(0.00)

-28.18 ***(0.00)

-10.93 ***(0.00)

2.04 ***(0.04)

Note: All variables are expressed in percentage. A bank is considered as large if its total assets is greater than one billion USD. Stable liabilities correspond to the amount of liabilities that are likely to stay within the bank following a shock. It is the sum of all liabilities weighted by their corresponding stable funding factor. Unstable liabilities correspond to the amount of liabilities that are likely to be suddenly claimed on demand following a shock. It is the sum of all liabilities weighted by their unstable funding factor. For further details about the breakdown of liabilities according to the importance of their stability, see table 4. DEPO_STB: (0.7*demand and saving deposits + 1*time and other time deposits) / total assets; DEPO_USTB: 0.3* demand and saving deposits / total assets; STMD_USTB: 1*short term borrowings and other short term liabilities / total assets; LTMD_STB: 1*long term borrowings and other long term liabilities / total assets; K_STB: 1*equity / total assets. T-statistics test for null hypothesis of identical means; *, **, *** indicate statistical significance respectively at the 10%, 5%, and 1% level, for bilateral test.

39

Table 13: The transformation risk neutral level of liquidity creation (TRNLC), for US and European banks over the 2000-2008 period

β α TRNLC

All banks 0.74 ***(75.11)

37.50 ***(422.39) 37.5

European banks 0.60 ***(46.51)

32.56 ***(377.97) 32.6

US banks 0.81 ***(60.72)

40.11 ***(270.22) 40.3

US banks - Large 0.78 ***(41.03)

39.13 ***(218.13) 39.2

US banks - Small 0.82 ***(43.26)

40.67 ***(174.77) 40.9

Note: All variables are expressed in percentage. A bank is considered as large if its total assets is greater than one billion USD. LC: liquidity creation / total assets, for further details about the computation of LC, see table 3; NSFD: (required amount of stable funding - available amount of stable funding) / total assets, for further details about the computation of NSFD, see table 4. Coefficients α and β are obtained by estimating equation (1): LCi,t= α +β *NSFDi,t+ ε i,t. After testing for cross section and time fixed versus random effects, we introduce cross section fixed effects in our regressions. *, **, *** indicate statistical significance respectively at the 10%, 5%, and 1% level. TRNLC is the transformation risk level of

liquidity creation. It corresponds to the average cross section fixed effects (i.e., ∑=

α+αN

1ii N/)( , subscript i denoting bank).

Table 14: Descriptive statistics of our main explanatory variables, for US and European commercial banks, on average from 2000 to 2008

Variables Mean Median Max Min Std. Dev. Obs.T12_TA 9.6 9.0 60.2 0.8 3.8 6414ROA 0.8 0.9 6.9 -15.1 1.0 6440PLL_TLO 0.5 0.3 7.2 -1.2 0.6 6289MKT_POW 1.7 0.0 74.5 0.0 6.3 6414GDP_GWT 2.3 2.5 9.5 -3.5 1.3 7029CB 3.0 2.3 22.0 0.3 1.9 7029IBK1M_CB 0.2 0.1 3.5 -0.4 0.2 7029LN_TA 7.6 7.0 15.1 2.8 2.1 6414CONTROL 10.5 11.0 12.0 4.0 1.3 7029HHI_LOAN 0.2 0.1 1.0 0.0 0.2 6414HHI_ILASSET 0.2 0.1 1.0 0.0 0.2 6414HHI_STFUND 0.5 0.5 1.0 0.0 0.3 6414SECLO_TLO 42.2 41.0 99.4 0.0 20.9 6414SECLO_IA 37.0 35.6 97.7 0.0 19.0 6414STMDBT_STDBT 20.1 13.9 100.0 0.0 19.8 6414

Source: Bloomberg (2000-2008). All variables are expressed in percentage (except LN_TA, all HHI and CONTROL). T12_TA: (tier 1 capital + tier 2 capital) / total assets; ROA: net income / total assets; PLL_TLO: total provisions for loan losses / total loans; MKT_POW: total assets of bank i in country j / total assets of the banking system in country j; GDP_GWT: annual growth rate of real GDP; CB: central bank policy rate; IBK1M_CB: spread of one month interbank rate and central bank policy rate; LN_TA: log of total assets; CONTROL: index of supervisory regime; HHI_LOAN: normalised Herfindalh Hirschman index for concentration on loans that are potentially securitisable versus on loans that cannot be securitised; HHI_ILASSET: : normalised Herfindalh Hirschman index for concentration on loans that are potentially securitisable versus on illiquid assets (i.e., including the loans that cannot be securitised and the other illiquid assets); HHI_STFUND: normalised Herfindalh Hirschman index for concentration on short term deposits versus on short term market debts; SECLO_TLO: consumer loans / total loans; SECLO_IA: consumer loans / (total loans + long term investments + customer acceptances + fixed assets + other assets); STMDBT_STDBT: short term market debts / (demand and saving deposits + short term market debts).

40

Table 15: The determinants of transformation riskThis table shows the result of estimating equation (2) for US and European publicly traded commercial banks, over the 2000-2008 period. The dependent variable is NSFD. Equations 2.a and 2.b correspond to the estimation of equation (2) by considering alternatively two proxies of the concentration on loans that are potentially securitisable. All explanatory variables are one year lagged (except LN_TA and CONTROL). See table 14 for the definition of the explanatory variables. To deal with colinearity problems in all regressions, we orthogonalise LN_TA with MKT_POW. We include cross section and time fixed effects and we use the Huber-White robust covariance method. *, **, *** indicate statistical significance respectively at the 10%, 5%, and 1% level.

2. a 2. b

HHI_LOAN 0.07 ***(16.66) -

HHI_LOAN *SECLO_TLO

-0.15 ***(-18.33) -

HHI_ILASSET - 0.08 ***(20.31)

HHI_ILASSET * SECLO_IA - -0.17 ***

(-13.15)

HHI_STFUND -0.03 ***(-8.78)

-0.03 ***(-8.77)

HHI_STFUND *STMDBT_STDBT

0.27 ***(13.87)

0.27 ***(13.90)

T12_TA -0.05 **(-2.13)

-0.06 ***(-2.39)

ROA -0.07(-0.56)

-0.06(-0.45)

PLL_TLO -1.73 ***(-8.33)

-1.72 ***(-8.30)

MKT_POW -0.45 ***(-16.50)

-0.47 ***(-16.99)

GDP_GWT 1.16 ***(8.07)

1.15 ***(8.05)

CB 1.57 ***(11.37)

1.54 ***(11.23)

IBK1M_CB 2.01 ***(3.77)

2.03 ***(3.86)

LN_TA -0.01 ***(-19.88)

-0.01 ***(-20.24)

CONTROL -0.03 ***(-39.88)

-0.03 ***(-39.77)

C 0.20 ***(18.47)

0.20 ***(18.05)

R² 0.82 0.82Fisher Stat 1174 1176P-Value F 0.00 0.00Total Obs. 5567 5567

41

APPENDIX 1Table A.1: Summary of the balance sheet weighting used to calculate “net stable funding ratio” as defined in the Basel III accords

Available funding source Availability factor

Tier 1 & 2 Capital Instruments

Other preferred shares and capital instruments in excess of Tier 2 allowable amount having an effective maturity of one year or greater

Other liabilities with an effective maturity of 1 year or greater

Less stable deposits of retail and small business customers (non-maturity or residual maturity < 1yr)

85%

Less stable deposits of retail and small business customers that are not covered by effective deposit insurance, high-value deposits, internet deposits and foreign currency deposits (non-maturity or residual maturity < 1yr)

70%

Wholesale funding provided by nonfinancial corporate customers (non-maturity or residual maturity < 1yr)

50%

All other liabilities and equity not included above 0%

Required funding source Required factor

CashShort-term unsecured actively traded instruments (< 1 yr)

Securities with exactly offsetting reverse repo

Securities with remaining maturity < 1 yrNon-renewable loans to financials with remaining maturity < 1 yr

Debt issued or guaranteed by sovereigns, central banks, BIS, IMF, EC, non-central government, multilateral development banks

5%

Unencumbered non-financial senior unsecured corporate bonds (or covered bonds) rated at least AA, maturity ≥ 1 yr

20%

Unencumbered listed equity securities or non-financial senior unsecured corporate bonds (or covered bonds) rated at least A-, maturity ≥ 1 yr

GoldLoans to non-financial corporate clients having a maturity < 1 yr

Loans to retail clients having a maturity < 1 yr 85%

All other assets 100%

100%

0%

50%

Source: “International framework for liquidity risk, measurement and monitoring”, 2009, Basel Committee of Banking Regulation and Supervision, Consultative Document.

42

APPENDIX 2

Table A2.1: The determinants of transformation risk, for US and European banks separatelyThis table shows the result of estimating equation (2) separately for US and European publicly traded commercial banks, over the 2000-2008 period. The dependent variable is NSFD. Equations 2.a and 2.b correspond to the estimation of equation (2) by considering alternatively two proxies of the concentration on loans that are potentially securitisable. All explanatory variables are one year lagged (except LN_TA and CONTROL). See table 14 for the definition of the explanatory variables. To deal with colinearity problems in all regressions, we orthogonalise LN_TA with MKT_POW. We include cross section and time fixed effects and we use the Huber-White robust covariance method. *, **, *** indicate statistical significance respectively at the 10%, 5%, and 1% level.

2. a 2. b 2. a 2. b

HHI_LOAN 0.11 ***(9.89) - -0.004

(-0.28) -

HHI_LOAN *SECLO_TLO

-0.22 ***(-10.61) - -0.07 ***

(-3.33) -

HHI_ILASSET - 0.12 ***(11.13) - 0.03 ***

(2.34)

HHI_ILASSET * SECLO_IA - -0.22 ***

(-6.75) - -0.14 ***(-5.71)

HHI_STFUND -0.07 *(-1.79)

-0.07 *(-1.77)

-0.11 ***(-11.31)

-0.11 ***(-11.24)

HHI_STFUND *STMDBT_STDBT

0.11 **(2.18)

0.11 **(2.20)

0.40 ***(15.97)

0.40 ***(16.07)

T12_TA -0.13 ***(-3.03)

-0.15 ***(-3.41)

0.04(0.61)

0.05(0.80)

ROA -0.28(-1.48)

-0.26(-1.39)

-0.94 ***(-2.70)

-0.93 ***(-2.67)

PLL_TLO -2.52 ***(-8.65)

-2.52 ***(-8.65)

-2.88 ***(-8.19)

-2.84 ***(-8.11)

MKT_POW 0.03(0.09)

0.04(0.11)

-0.22 ***(-7.79)

-0.22 ***(-7.91)

GDP_GWT 0.52 ***(3.91)

0.52 ***(3.90)

-0.26 *(-1.76)

-0.25 *(-1.74)

CB 0.19 ***(3.25)

0.19 ***(3.12)

1.46 ***(8.25)

1.44 ***(8.06)

IBK1M_CB 7.24 ***(6.33)

7.26 ***(6.35)

0.01(0.02)

0.11(0.20)

LN_TA 0.03 ***(8.50)

0.03 ***(8.14)

0.01 ***(6.46)

0.01 ***(7.22)

CONTROL - - 0.0004(0.36)

0.001(1.27)

C -0.12 ***(-13.57)

-0.13 ***(-14.05)

-0.02(-1.18)

-0.03 **(-2.11)

R² 0.78 0.78 0.80 0.80Fisher Stat 21 21 426 426P-Value F 0.00 0.00 0.00 0.00Total Obs. 4036 4036 1531 1531

US banks European banks

43

Table A2.2: The determinants of transformation risk, for large and small banks separatelyThis table shows the result of estimating equation (2) separately for large and small banks, over the 2000-2008 period. A bank is considered as large if its total assets is greater than one billion USD. The dependent variable is NSFD. Equations 2.a and 2.b correspond to the estimation of equation (2) by considering alternatively two proxies of the concentration on loans that are potentially securitisable. All explanatory variables are one year lagged (except LN_TA and CONTROL). See table 14 for the definition of the explanatory variables. Besides, US banks account for a large share of our sample (574 US banks against 207 European banks) and small US banks account for a large share of the sample of small banks (170 US banks against 37 European banks). Consequently, we consider bank location to check the stability of our results according to bank size. Thus, we add in equation (2), all the interactions of our explanatory variables with a dummy variable that takes the value of 1 for US banks and 0 otherwise (DUM_LOC). To deal with colinearity problems in all regressions, we orthogonalise LN_TA with MKT_POW. We include cross section and time fixed effects and we use the Huber-White robust covariance method. *, **, *** indicate statistical significance respectively at the 10%, 5%, and 1% level.

Coefficient Large banks Small banks Large banks Small banks

C(1) HHI_LOAN 0.04 ***(2.63)

0.20 **(2.15) HHI_ILASSET 0.07 ***

(5.82)0.18 ***(2.62)

C(2) HHI_LOAN * DUM_LOC 0.03 *(1.63)

-0.07(-0.80) HHI_ILASSET * DUM_LOC 0.02

(1.40)-0.05

(-0.71)

C(3) HHI_LOAN *SECLO_TLO

-0.14 ***(-6.35)

-0.30 ***(-3.13) HHI_ILASSET * SECLO_IA -0.18 ***

(-6.46)-0.28 ***(-3.72)

C(4) HHI_LOAN *SECLO_TLO * DUM_LOC

-0.05 *(-1.78)

0.08(0.79)

HHI_ILASSET * SECLO_IA * DUM_LOC

-0.005(-0.13)

0.03(0.37)

C(5) HHI_STFUND -0.05 ***(-3.84)

-0.04 ***(-2.73) HHI_STFUND -0.05 ***

(-3.73)-0.04 ***(-2.84)

C(6) HHI_STFUND * DUM_LOC 0.03 ***(2.45)

0.02(0.90) HHI_STFUND * DUM_LOC 0.03 **

(2.30)0.02

(1.05)

C(7) HHI_STFUND *STMDBT_STDBT

0.33 ***(10.86)

0.03(0.64)

HHI_STFUND *STMDBT_STDBT

0.33 ***(10.97)

0.04(0.74)

C(8) HHI_STFUND *STMDBT_STDBT * DUM_LOC

-0.11 *(-1.75)

0.03(0.35)

HHI_STFUND *STMDBT_STDBT * DUM_LOC

-0.11 *(-1.70)

0.04(0.43)

C(9) T12_TA -0.03(-0.48)

0.24 **(1.97) T12_TA -0.03

(-0.46)0.23 **(1.95)

C(10) T12_TA * DUM_LOC -0.07(-0.87)

-0.26 **(-2.03) T12_TA * DUM_LOC -0.07

(-0.88)-0.26 **(-2.13)

C(11) ROA -0.91 **(-2.08)

-0.96 *(-1.67) ROA -0.87 **

(-2.00)-0.89

(-1.57)

C(12) ROA * DUM_LOC 0.59(1.21)

0.65(1.07) ROA * DUM_LOC 0.51

(1.06)0.64

(1.05)

C(13) PLL_TLO -1.17 ***(-2.81)

-2.44 ***(-3.75) PLL_TLO -1.15 ***

(-2.74)-2.39 ***(-3.62)

C(14) PLL_TLO * DUM_LOC -1.18 **(-1.99)

0.62(0.82) PLL_TLO * DUM_LOC -1.23 **

(-2.06)0.58

(0.75)

C(15) MKT_POW -0.25 ***(-9.08)

-3.25(-1.35) MKT_POW -0.25 ***

(-9.30)-2.08

(-0.84)

C(16) MKT_POW * DUM_LOC -0.02(-0.14)

-353.46 ***(-3.77) MKT_POW * DUM_LOC -0.01

(-0.06)-356.98 ***

(-3.81)

C(17) GDP_GWT -0.36 **(-2.13)

0.74 ***(2.37) GDP_GWT -0.38 **

(-2.28)0.76 ***(2.43)

C(18) GDP_GWT * DUM_LOC 0.54 **(2.21)

-0.22(-0.63) GDP_GWT * DUM_LOC 0.56 **

(2.29)-0.24

(-0.69)

C(19) CB 1.33 ***(6.35)

0.77 ***(2.82) CB 1.30 ***

(6.17)0.77 ***(2.70)

C(20) CB * DUM_LOC -0.83 ***(-3.69)

-0.70 ***(-2.45) CB * DUM_LOC -0.81 ***

(-3.58)-0.70 ***(-2.37)

C(21) IBK1M_CB -0.08(-0.17)

2.48(0.74) IBK1M_CB 0.03

(0.07)2.77

(0.83)

C(22) IBK1M_CB * DUM_LOC 6.17 ***(3.74)

3.06(0.83) IBK1M_CB * DUM_LOC 5.85 ***

(3.55)2.84

(0.78)

C(23) LN_TA 0.002(1.08)

-0.001(-0.12) LN_TA 0.004 **

(2.03)-0.0001(-0.01)

C(24) LN_TA * DUM_LOC 0.005 **(2.15)

-0.001(-0.12) LN_TA * DUM_LOC 0.004 **

(2.03)-0.0001(-0.01)

C(25) CONTROL 0.001(1.02)

0.01 **(1.92) CONTROL 0.002 *

(1.84)0.01 *(1.75)

C(26) CONTROL * DUM_LOC -0.001(1.04)

-0.01 **(1.93) CONTROL * DUM_LOC -0.01 *

(1.87) -0.01 *(1.78)

C(27) C -0.01(-0.78)

-0.19 ***(-3.10) C -0.03 *

(-1.81)-0.18 ***(-3.08)

R² 0.82 0.79 R² 0.82 0.79Fisher Stat 520 356 Fisher Stat 520 355P-Value F 0.00 0.00 P-Value F 0.00 0.00Total Obs. 3013 2554 Total Obs. 3013 2554

Wald stat c(1)+c(2)=0 0.07 *** 0.12 *** c(1)+c(2)=0 0.09 *** 0.13 ***and % level c(1)+c(2)+c(3)+c(4)=0 -0.12 *** -0.10 *** c(1)+c(2)+c(3)+c(4)=0 -0.09 *** -0.13 *** to reject: c(5)+c(6)=0 -0.01 *** -0.02 *** c(5)+c(6)=0 -0.02 *** -0.02 ***

c(5)+c(6)+c(7)+c(8)=0 0.20 *** 0.04 c(5)+c(6)+c(7)+c(8)=0 0.20 *** 0.05

2. a 2. b

44