The Liquidity Regulation and Savings Banks’ Liquid Assets

For their short-term payment obligations, savings banks hold substantially more liquid assets than the liquidity regulation requires. This paper investigates whether sight deposits, an important funding source for savings banks, help in explaining liquid asset holdings in excess of regulatory requirements. We analyze whether savings banks transform sight deposits in illiquid assets less intensively than is permitted because (i) the liquidity regulation underestimates actual withdrawal rates (underestimation effect) and/or (ii) savings banks are subject to limits in their lending to non-banks that they do not offset by, for instance, mediumterm interbank lending or fixed asset holdings (lending effect). In our sample, we do not find the underestimation effect to be applicable as actual deposit withdrawal rates are in most cases lower than the regulatorily specified rate. However, we find the lending effect to be at work: Savings banks with low shares of loans to non-banks do not transform sight deposits into illiquid assets as intensively as savings banks with high shares of non-bank loans. Our analysis does not only show that liquid assets positively depend on sight deposits, but also shines a light on how bank size and the individual bank’s position in the interbank market affect liquid assets.


I. Introduction
In Germany, the Regulation on the Liquidity of Institutions requires banks to have a liquidity ratio which is at least equal to one.This ratio is calculated by dividing regulatorily specified liquid assets that are available within one month by short-term payment obligations that are callable within this period.Savings banks have, unlike big banks and state banks, almost three times as much liquid assets for their payment obligations than the regulator requires.This paper investigates whether sight deposits, which are not only factored into the payment obligations with 10% but which are also an important funding source for savings banks but less so for big banks (Memmel/Schertler (2010)), contribute to the high amounts of savings banks' liquid assets.
We investigate two effects that may induce savings banks to hold more liquid assets for each unit of sight deposits than regulatorily required so that they would not transform sight deposits into illiquid assets as intensively as the regulator permits.The first effect, which we call the underestimation effect, occurs if the regulator underestimates the likelihood of deposit withdrawals and when liquidity shortages are expensive.The second effect, which we call the lending effect, occurs when savings banks have limits in non-bank lending which they do not offset through investments in other illiquid assets, such as medium-term interbank lending and/or investments in fixed assets.Disentangling these two effects is important because both effects can induce savings banks to use sight deposits less intensively to finance illiquid assets than allowed but they require different responses by the regulator.If the first effect is at work, the regulator may want to re-specify regulatory weights to capture potential deposit withdrawals adequately, while if the latter effect is at work, there is currently no necessity to change the weighting of sight deposits used by the liquidity regulation.
Looking at bank reports of German savings banks and taking into account that banks determine their assets and payment obligations simultaneously, we find evidence that a higher amount of sight deposits is associated with much more liquid assets holdings than the regulator requires, while a higher amount of other short-term payment obligations is associated with an amount of liquid assets holdings closer to the one required by the regulator.To investigate whether the regulation underestimates deposit withdrawals, we look at historical changes in sight deposits.Our findings suggest the liquidity regulation captures actual deposit withdrawals quite adequately in our sample.Thus, the withdrawal rate of sight deposits specified in the liquidity regulation can be considered as conservative and does, therefore, not explain why savings banks hold much more liquid assets than the regulator requires.To investigate whether the lending effect forces savings banks to hold more sight deposits in liquid assets than required, we test whether savings banks with high shares of loans to non-banks use sight deposits more intensively to finance illiquid assets than savings banks with low shares of loans to non-banks.Using an interaction term within a dynamic panel data approach and controlling for non-bank lending and the deposit-liability ratio, we find that liquid assets depend negatively on the interaction between non-bank lending and sight deposits.This finding suggests that it is not profitable for savings banks with low lending to non-banks to transform sight deposits into other illiquid assets such as medium-term interbank lending or securities stated as fixed financial assets.It may imply that it is more profitable for savings banks with low non-bank lending to hold liquid assets than to grant medium-term interbank loans or to hold securities to maturity to use the advantages of financial reporting (i.e., gemilderte Niederstwertvorschrift).
While our analysis is primarily focused on how liquid assets depend upon sight deposits, it also provides information on how these assets vary according to bank size and the individual bank's position in the interbank market, both of which we find to be inversely related to liquid assets.Overall, our results indicate that there is no single factor determining the amount of liquid assets.Instead, liquid assets are determined by a multiplicity of factors, some of which should be the subject of further research.
Our paper expands the recent literature on banks' liquidity, which has, broadly speaking, focused on reserve requirements (Bartolini et al. (2001), Jallath-Coria et al. (2002)), securities holdings and cash balances (Aspachs et al. (2005), Freedman/Click (2006)), and the creation of li-quidity (Berger/Bouwman (2009)), i. e., transforming short-term liabilities into illiquid assets.The recent literature dealing with the German liquidity regulation is primarily descriptive: Moch/Schöning (2008) provide some evidence that savings banks use Principle II (in addition to other methods) to monitor their liquidity positions.In our paper, we consider insights gained in the recent literature, such as the role of the interbank market: if banks' liquidity shocks are imperfectly correlated, banks can protect themselves against liquidity shortages by being active in the interbank market (Rochet/Tirole (1996)).Therefore, we control for savings banks' activity in the interbank market.Other insights, such as the role of central banks as a lender of last resort for banks' liquidity (e.  2006)), are not particularly relevant for our paper, since we focus on savings banks that are organized in a network.
The remainder of the paper is organized as follows.In Section II, we give an overview of the current and past prudential regulatory framework for liquidity.In Section III, we present predictions on the relationship between liquid assets and sight deposits.In Section IV, we describe the dataset and Section V presents our estimation methodology and findings.Section VI summarizes our main findings and suggests topics for future research.

II. Prudential Liquidity Regulation in Germany
In recent years, the German liquidity regulation for banks has undergone several revisions.Banks' liquidity requirements are specified in Section 11 of the Banking Act, which states that banks "must invest their funds in such a way as to ensure that adequate liquidity for payment purposes is guaranteed at all times" (FBSO (1998a)).Since 2007, Section 11 of the Banking Act has been made concrete by the Regulation on the Liquidity of Institutions (Liquiditätsverordnung). Between 2000 and 2006 Section 11 of the Banking Act was made concrete by Principle II (Grundsatz II), while before 2000 it was made concrete by the original Principle II and Principle III (Grundsatz II and III).
The Regulation on the Liquidity of Institutions encompasses Principle II by requiring banks to calculate and to report liquidity ratios (Standardized Approach, stated in Sections 2-7), but expands on Principle II by allowing banks to use their own liquidity models (Section 10) that have to be approved by the regulator.Thus, for those banks not opting to use their own liquidity model, the requirements of the liquidity regulation -apart from some smaller changes -did not change in 2007.The regulatory liquidity ratio (RLR) is the quotient of assets available within the next month and payment obligations callable within the next month (FBSO (1998a)).The liquidity of a bank is deemed as adequate if the RLR is at least one.
Principle II was intended to adapt the German regulatory structure to international standards by taking into account not only the style of EU liquidity schemes but also recent developments in credit institutions' business environment (FBSO (1998a)).Therefore, Principle II differed from the original Principle II and Principle III in several respects.First, it was built on the proposition that a solvent and profitable bank should face no obstacles in ensuring medium-and long-term refunding (Deutsche Bundesbank (1999)).In the short run, however, solvent and profitable institutions may face the risk of liquidity shortages (FBSO (1998a)).It focuses, therefore, on withdrawal risks of liabilities and refunding risks in the short run (Schöning (2004a), Spörk/Auge-Dickhut (1999)).The original Principle II and Principle III, in contrast, focused on the middle and long-term liquidity needs of banks and put emphasis on refunding risks resulting from banks' maturity transformation (Hartmann-Wendels/Wendels (1999), Spörk/Auge-Dickhut (1999)).In so doing, they expanded on (i) the golden banking rule by specifying that longterm (medium-term) assets were to be financed by long-term (mediumterm) liabilities, (ii) the deposit base theory by assuming that callable deposits were not withdrawn at once, but were available to the banks for a longer period, and (iii) the shiftability theory by specifying that particular asset types did not need to be funded by liabilities with the same maturity (Schöning (2004a)).A third source of liquidity risk, the time risk, was not captured by Principle II and the original Principle II and III (Grelck/Rode (1999)).Second, Principle II was based on residual maturities, while the original Principle II and III were founded on original maturities.Third, Principle II allowed market values to be taken into consideration for particular liquid assets, while the original Principles II and III were based on book values only.Thus, Principle II combined a maturity-mismatch approach (since residual maturities of liquid assets and payment obligations are used when calculating the RLR) with a stock-market approach (since securities traded on a regular stock market are classified as highly liquid assets) (Schöning (2004b)).
The liquidity regulation requires several on-balance sheet as well as off-balance sheet positions to enter the RLR.Table 1  relevant liquid assets and payment obligations as a percentage of the numerator and denominator of the RLR.The most relevant liquid assets are debt securities (33%), followed by securities listed in a regular market (16%), shares in money market and securities funds (16%), and loans maturing within the next month to customers (16%) and credit institutions (11%).Off-balance sheet positions, such as irrevocable lending commitments received by credit institutions, are not very important for liquid assets.The most relevant payment obligations are customer liabilities maturing within the next month (37%), savings deposits (28%), customer liabilities that are due on demand (15%), and liabilities to credit institutions that will mature within the next month (5%).Thus, the de-nominator of the RLR is dominated by customer liabilities -which is, as we are looking at savings banks, not much of a surprise.Off-balance sheet positions, such as placement and underwriting commitments or undrawn irrevocable credit facilities, are not very important for short-term payment obligations.

III. Predictions
Predictions are gained from the literature focusing on reserve requirements since these requirements are, from an economic point of view, comparable to regulatory liquidity requirements (for models dealing with reserve requirements, see Freixas/Rochet (1997), chapter 8, Baltensperger/Milde (1987), chapter 2).As in the case of reserve requirements, we can think of prudential regulatory liquidity requirements as an additional constraint in a bank's profit maximization problem (in addition to the internal liquidity constraint).The liquidity regulation requires banks to fulfill the following constraint: where LA T denotes liquid assets, LB S denotes regulatory sight deposits (i.e., 10% of the sight deposits at the bank's disposal), and LB O denotes the amount of other short-term payment obligations (including, e. g., 10% of its savings deposits and 100% of its customer liabilities maturing within the next month).
When the liquidity regulation introduces a binding constraint, i. e., RLR ã 1, we expect banks will hold one unit of liquid assets for each unit of payment obligations.This unit of liquid assets can be in the form of either securities holdings or repayments from loans maturing within the next month.If the repayments from loans maturing within the next month are sufficiently high to meet the regulatory liquidity constraint, the regulator permits banks to use as much as 100% of the sight deposits at their disposal to grant illiquid loans to non-banks or credit institutions and/or to invest in securities stated as financial fixed assets (both of which are not factored in the numerator of the RLR).
The regulator intervenes when the bank fails to meet the requirements specified by the liquidity regulation. 2 This is not in the interest of the bank managers, since they either loose control of their banking operations or are restricted in conducting them.The bank can reduce the probability of a regulatory intervention by having liquid assets for each unit of short-term payment obligations in excess of regulatory requirements.Thus, the danger of a regulatory intervention may induce the bank to build up and to keep a liquidity buffer.Then equation (1) changes to: with a 1 > 1: È2ê We bring forward two effects, both of which can induce savings banks to hold an amount of liquid assets per unit of sight deposits that exceeds the one per unit of other short-term payment obligations.The first effect we call the underestimation effect.The bank will hold excess liquid assets for each unit of sight deposits when the liquidity regulation underestimates deposit withdrawals and when liquidity shortages are expensive.The regulation underestimates deposit withdrawals, when the actual monthly deposit withdrawal rate, D m D W , exceeds the monthly sight deposit withdrawal rate of 10% specified in the liquidity regulation.Such underestimated deposit withdrawals induce savings banks to hold more sight deposits in securities and cash balances than regulatorily required, if the price per unit of the liquidity shortage exceeds the price for sight deposits.
Banks that mainly store liquidity (Saunders/Cornett (2006)) are not only interested in monthly deposit withdrawal rates, they also care about the correlation of deposit withdrawals over time.To sketch the implications on liquid assets of a bank being hit by deposit withdrawals in several subsequent months, we assume the bank, which has only sight deposits and equity at its disposal, initially has an RLR of 2 and is hit, ceteris paribus, in all subsequent months by deposit withdrawals amounting to either 2% or 10%.When the bank is only hit by a 2% deposit withdrawal, the bank meets the regulatory requirement in all up to the sixth month.However, when the bank is hit by a 10% deposit withdrawal in each period, it meets the regulatory requirement in the first but not in the second month.Thus, when the bank stores liquidity and expects a deposit withdrawal rate in subsequent months, D s D W , that exceed the regulatory withdrawal rate, it holds, on average, more liquid assets than a bank that expects deposit withdrawals not to be correlated over time.In this case, the regulatory constraint will force banks to hold excess liquidity even if the observed average RLR does not indicate that it is binding.This brings us to our first prediction.
Prediction 1: How intensively sight deposits are used to finance illiquid assets depends upon banks' deposit withdrawal rates.Banks with higher deposit withdrawal rates in a single month or within subsequent months than assumed by the regulator, i. e., D m D W > 10 % or D s D W > 10 % hold more liquid assets for each unit of sight deposits than their respective counterparts.
The second effect, which we call the lending effect, exists when savings banks with low non-bank lending hold more liquid assets than savings banks with high lending to non-banks.At first sight, one might argue that a bank with low non-bank lending has to opt for more liquid asset holdings.But a bank with low non-bank lending can realize the same degree of transforming sight deposits into illiquid assets as a bank with high amounts of non-bank lending when it invests more into securities stated as fixed financial assets and/or when it grants loans to other credit institutions with an initial maturity of more than 1 month, which are not regulatorily specified as liquid assets.
To identify whether such an effect is at work, we measure the relationship between liquid assets and sight deposits for those banks with high lending to non-banks and those banks with low non-bank lending separately.For all banks, we expect the amount of liquid assets held for each unit of regulatorily specified sight deposits, a 1 S , will be strictly larger than one, while for those banks with high non-bank lending, we expect that they will use sight deposits more intensively to fund illiquid loans, so that they have fewer liquid assets for each unit of sight deposits, i. e., a 2 S < 0. Our second prediction summarizes the effect of lending to nonbanks on liquid assets.
Prediction 2: How intensively sight deposits are used to finance illiquid assets depends upon banks' lending to non-banks, L. Banks with high non-bank lending, D L = 1, hold fewer liquid assets for each unit of sight deposits than banks with low non-bank lending, D L = 0: O 0 and a L < 0: È3ê

IV. The Data
We analyze regulatory reporting data on savings banks' liquidity for the period 2000-2006 (reports on Principle II).We use the following three measures for liquid assets: (i) total liquid assets, LA T , (ii) debt and equity securities holdings, LA S , and (iii) cash balances, LA C .Since all these measures are size-sensitive, we scale them as a percentage of total assets.Table 2 shows that, on average, liquid assets account for about 36% of total assets, debt and equity securities for 24% and cash balances for about 2.3%.
For payment obligations, we use the following measures: (i) total regulatory short-term payment obligations, LB T , (ii) regulatory sight deposits of non-banks, LB S , (iii) and other regulatory short-term payment obligations, LB O .All the measures we use for payment obligations are the regulatory amounts relative to total assets (in percent), i. e., regulatory sight deposits, LB S , are calculated as (0.1Ásight deposits)/(total assets).Total payment obligations account for more than 13% of savings banks' total assets, regulatory sight deposits account for 1.9% (i.e., sight deposits account for 19% of total assets), and other payment obligations account for more than 11%.We take into account that liquid assets likely depend on several bankspecific characteristics, such as the ratio of loans to non-banks relative to total assets, L, and the annual growth in loans to non-banks, DL.On average, as Table 2 shows, savings banks use almost 60% of their assets to grant loans to non-banks.The stock of loans to non-banks grows on average by 1% per year.Moreover, we control for the interest margin, IM, which measures the banks' opportunity costs of holding liquid assets in terms of forgone higher returns from loans (Aspachs et al. ( 2005)).Therefore, we expect the interest margin will affect liquid assets negatively.
We control for bank size, SIZE, measured by the bank's total assets, because we expect it to be correlated with using sophisticated liquidity management techniques.In particular, large banks are more likely to use sophisticated techniques of managing liquidity risk than small banks because the costs of implementing such a technique might be independent of bank size, while the benefits certainly do.Banks that use sophisticated liquidity techniques likely hold smaller volumes of liquid assets.In addition, bank size may be positively correlated with using purchasing liquidity techniques, since using such a technique may have fixed-cost character.
Finding a more precise measure than size for capturing whether savings banks employ purchased liquidity techniques is difficult, since savings banks are part of liquidity networks that they use to manage their liquidity when monetary policy conditions change (Ehrmann/Worms (2004)).Thus, per se, all savings banks have access to purchased liquidity.However, we expect the positions of savings banks within the liquidity network in terms of price and quantity conditions to differ and therefore we count the number of connections the bank has as a borrower in the interbank market.For many savings banks, the number of these interbank connections is greater than one because the formerly strong single relationship between savings banks and their head institutions has become much weaker in the past few years.Head institutions in northern Germany also offer liquidity to savings banks in southern Germany.We expect the number of interbank connections relative to total assets, IB, to be negatively related to liquid assets.Such a negative relationship arises when interbank connections are negatively related to the conditions of purchasing liquidity.However, a negative relationship can also arise when some banks faced a strong loan demand in the past, which forced them to shift liquid assets into illiquid loans first (as much as the regulator permits) and afterwards to raise additional funding.In either case, we expect the number of interbank connections will be negatively correlated with liquid assets.
Table 2 also gives summary statistics for regulatory bank capital, loan write-offs, and a Herfindahl index of loan portfolio concentration.We use these variables in extension models in which we check the robustness of our estimation results.

V. Underestimated Deposit Withdrawals or Limits
in Non-Bank Lending?

Underestimated Deposit Withdrawals
According to our first prediction, savings banks may hold more liquid assets than regulatory required because the liquidity regulation underestimates the likelihood of sight deposit withdrawals.To identify whether this underestimation effect is at work requires identifying those banks with higher withdrawal rates than regulatorily specified.Figure 1 plots the changes in sight deposits calculated from monthly, quarterly, semiannual, and annual stocks of sight deposits since savings banks may not only care about deposit withdrawals within one month but within subsequent months.The left upper plot in Figure 1 suggests a relatively low likelihood of experiencing a monthly change in sight deposits above the regulatory value of 10%.In addition, the other three plots for quarterly, semi-annual, and annual changes in sight deposits do not indicate negative changes that would add up to more than 10% of initial sight deposits in subsequent months.These changes differ from the one which the regulator specifies, since the changes presented in Figure 1 are not controlled for growing sight deposit bases.However, even if we control for growing sight deposits bases, the distributions of sight deposit withdrawals do not change substantially.
As Figure 1 suggests, the historical changes in sight deposits seldom exceed the weight for deposit withdrawals specified by the regulator during our sample period.Thus, the regulatory value for expected deposit withdrawals of 10% can be regarded as a conservative value.The deposit withdrawal rates depicted in Figure 1 imply that we cannot classify a group of savings banks that experience higher than regulatorily specified deposit withdrawals.Therefore, we tentatively conclude that underestimating actual withdrawal rates cannot explain why savings banks hold more liquid assets for their payment obligations than required.However, we have two cautionary notes on this interpretation of our findings.First, savings banks' managers may be risk-averse and they may not decide on the basis of historical withdrawal rates but rather may take into account that all sight deposits can be withdrawn at once.Second, the stock of sight deposits reported at the end of each month (which we used in Figure 1) can be substantially higher than if it was averaged over each month.If wages and salaries are mainly paid at the end of the month, savings banks will store part of the sight deposits in liquid assets that they can easy transform into cash to be prepared for deposit withdrawals.Of course, a deposit withdrawal from a customer's perspective does not have to coincide with a deposit withdrawal from the bank's perspective since money often changes from one account to another and since withdrawn sight deposits are compensated for by other sight or savings deposit inflows.Nevertheless, we do not know how intensively sight deposits fluctuate between the two points in time at which banks report their liquidity to the regulator.

Limits in Non-Bank Lending
To test whether savings banks hold more liquid assets than regulatory required because it is unprofitable for them to offset limits in their lending to non-banks, we use an econometric analysis built on equation 3. Central in this analysis is a dummy variable, D L , which is equal to one if the savings bank has a high ratio of loans to non-banks relative to total assets and zero otherwise, that we interact with the variables capturing payment obligations.The interaction terms are potentially endogenous because savings banks simultaneously decide on their liquid assets and payment obligations.We, therefore, instrument them by using past values.To yield appropriate instruments for these potentially endogenous interaction variables, we employ a time-invariant dummy variable.

a) The Econometric Models
The baseline econometric models that we use to test our second prediction takes into account dynamic changes in liquidity by including a lag of the dependent variable in the list of the RHS variables.The baseline models have the following form: with y 2 fLA T ; LA S ; LA C g È4ê where y i;t denotes the dependent variable of interest for bank i at time t.
Apart from bank-specific characteristics, we control for the macroeconomic environment by including the change in the short-term interest rate, D i, and the real GDP growth rate, DGDP.Additionally, we include year dummies to further control for time-fixed effects.We assume e i; t ã h i þ g i; t , where h i is a bank-specific fixed effect and g i; t is a disturbance term.
Our baseline models include only a subset of bank-specific variables that we consider to be relevant for liquid assets.The reason for this is that some bank-specific variables are highly correlated.We also observe a high correlation between interbank connections and bank size.However, the effects of interbank connections and bank size on liquid assets do not depend on whether or not these variables are included jointly in the models.We introduce further bank-specific characteristics, such as bank capital, in extensions of these baseline models discussed below.
To estimate the baseline and extension models, we take into account that some RHS variables are endogenous; some variables have even been used as dependent variables in other empirical studies.For example, the growth in loan stocks is analyzed by Kashyap/Stein (2000), Ashcraft (2006), Kishan/Opiela (2000), and Merkl/Stolz (2009), among many others.Bank capital and its interdependency with risk-weighted assets is analyzed by Shrieves/Dahl (1992), Jacques/Nigro (1997), Aggarwal/Jacques (2001), Rime (2001), and Heid et al. (2004).To minimize endogeneity problems, we use lagged variables whenever the variable under focus is related to one point in time only, i. e., we use the lagged ratio of loans to non-banks relative to total assets, interbank connections, and capital, while we instrument those variables that are calculated from two points in time, i. e., loan growth.
We estimate all following models by using the dynamic panel data estimator (which is a generalized method of moments estimator) proposed by Blundell/Bond (1998) and a finite sample correction proposed by Windmeijer (2005).Estimation results will be consistent if we use appropriate instruments for the lag of the dependent and RHS variables, and if there is no higher-order autocorrelation.We use a test for overidentifying restrictions (Arellano/Bond (1991), Blundell/Bond (1998)) to select the models presented in Table 3.Since each variable has both first-round and second-round effects, for example, an increase in the ratio of loans to non-banks relative to total assets in t impacts on liquid assets in t, which in turn impacts on liquid assets in t þ 1, we also report the longrun coefficients.

b) Results of the Baseline Models
We present estimation results of the baseline models for the three measures of liquid assets in Table 3.In all baseline models, the test on overidentifying restrictions indicates that the hypothesis that the instruments are valid cannot be rejected and that there is no higher-order autocorrelation.
The interaction terms between the dummy variable equal to one for those savings banks with high lending to non-banks and the two types of short-term payment obligations shed light on the relevance of our second prediction.We estimate each interaction term in a separate equation be- cause the two terms are correlated.The interaction term of sight deposits impacts significantly negatively on total liquid assets (column 1 in Panel a), suggesting that savings banks with high non-bank lending use a greater amount of sight deposits to finance illiquid assets than banks with low non-bank lending.In the long-run, in which second-round effects are taken into account, liquid assets held for each unit of sight deposits are 1.3 units lower for savings banks with high non-bank lending than for banks with low non-bank lending (column 1 in Panel b).Since total liquid assets comprise repayments from loans and advances maturing within the next month, the results for total liquid assets do not provide information on whether savings banks actually store more sight deposits in liquid assets than regulatory required.Therefore, we report results for securities holdings and cash balances.For each unit of regulatory sight deposits, savings banks with high non-bank lending store 1.8 units less in securities than savings banks with low non-bank lending (column 2 in Panel b), while the interaction term for cash balances is insignificant (column 3 in Panel b).The interaction terms of other regulatory short-term payment obligations are insignificant throughout (columns 4-6).These findings support our second prediction: savings banks do not transform sight deposits into illiquid assets to the degree permitted by the regulator because they face limits in their lending to non-banks which they do not offset by investing in other illiquid assets.Our findings also indicate that savings banks with low and high nonbank lending hold significantly more liquid assets per unit of sight deposits than required by the regulator.For savings banks with low nonbank lending, one unit of regulatory sight deposits is associated with 5.69 units of liquid assets in the long-run.For savings banks with high non-bank lending, one unit of regulatory sight deposits is associated with 4.38 (ã 5:69 À 1:31) units of liquid assets in the long-run.Thus, even savings banks with high non-bank lending do not transform sight deposits into illiquid assets to the extent permitted by the liquidity regulation.This effect might be caused by a lower availability of sight deposits within one month.
Noteworthy is also the long-run coefficient of other regulatory payment obligations, since it indicates each unit of these obligations is associated with more liquid assets than required: For each unit of these obligations, savings banks have 1.7 times the amount of total liquid assets required by the liquidity regulation.This finding indicates that savings banks hold a liquidity buffer.However, the liquidity buffer we estimated seems to be rather large.One reason for such a large liquidity buffer might be that savings banks manage their liquidity in such a way that they can meet regulatory liquidity requirements even if they are seeking to expand their loans to non-banks.Savings banks aiming at expanding their loans to non-banks may reserve the repayments from loans and advances (which are factored into the numerator of the RLR) for the funding of new illiquid loans (which are not factored into the denominator of the RLR).In line with this reasoning is the less pronounced effect of other short-term payment obligations on debt and equity securities and cash balances: one unit of other regulatory obligations is associated with 0.66 units of securities and 0.05 units of cash balances.Thus, with respect to securities and cash balances, savings banks do not, per se, hold liquidity buffers.This implies savings banks need some of the repayments from loans maturing within the next month to meet the requirements in the liquidity regulation.
Several of our control variables related to bank-specific characteristics and the macroeconomic environment help in explaining liquid assets.Total liquid assets, securities, and cash balances are lower when savings banks have higher shares of loans to non-banks relative to total assets.Additionally, savings banks reduce their securities holdings when they increase loans to non-banks, while they do not reduce their cash balances significantly irrespective of whether or not cash balances contain lending commitments received by other institutions.This might be be-cause cash balances are kept at a minimum and that this minimum is necessary to meet reserve requirements.
Bank size, i. e., the logarithm of total assets, and savings banks' interbank connections relative to total assets impact significantly negatively on liquid assets.Thus, larger savings banks and savings banks with more connections in the interbank market in the previous period have a smaller volume of liquid assets in the current period.As argued in the last section, savings banks with multiple connections may hold a smaller amount of liquid assets either because they do not have to prepare for liquidity shortages as much as their counterparts do or because they used their liquid assets in the past to grant illiquid loans before they started to raise additional funds in the interbank market which increased their number of connections.
As to the macroeconomic environment, coefficients have the expected signs: an increase in GDP growth or in the short-term interest rate reduces total liquid assets.However, the types of liquid assets are differently affected by changes in the macroeconomic environment.An increase in the short-term interest rate, which increases the opportunity costs of holding cash, results in higher securities holdings, while it leads to lower cash balances.An increase in GDP growth, which may be associated with increasing loan demands, results in smaller securities holdings, whereas it results in larger cash balances.Thus, when the economy expands, savings banks change their composition of liquid assets towards those assets that we classified as being more liquid.This might be because savings banks need liquidity to be prepared to provide additional loans to non-banks.
During the sample period, the German banking sector underwent a substantial wave of mergers and acquisitions (M&A).The estimation results presented so far are based on a sample in which the two pre-M&A banks are separated from the post-M&A bank and the resulting three banks are handled separately.This procedure minimizes the loss of information but it produces an unbalanced panel dataset.As a robustness test, we employ a dataset from which the two pre-M&A banks and the post-M&A bank are dropped.This procedure can result in a substantial information loss but it produces a balanced dataset, i. e. the number of observations over time is identical for those banks that remain in the dataset.Employing this sample confirms our estimation results: the coefficients of sight deposits and other short-term payment obligations do not change at all.Since actual withdrawal rates do not induce savings banks to hold more liquid assets than required, the question arises why savings banks do not reduce the amount of sight deposits when they cannot use it to finance illiquid assets as intensively as permitted by the regulator.We have three explanations for this behavior.First, the costs per unit of sight deposits may be less than the returns per unit of liquid assets, giving savings banks an incentive to collect as many sight deposits as possible and hold them in liquid assets.This strategy is, however, not riskless, as it is subject to market risks.Second, reducing the amount of sight deposits might not only lower the number of depositors but also the number of other customers (cross-selling).Finally, reducing the amount of sight deposits implies banks' size may shrink which may not mesh with the interests of managers, who may equate bigger with better (empire building).

c) Extensions
We use extensions of the baseline models to gain insights on whether bank capital and risks in the loan portfolio (summary statistics are given in Table 2), impact on liquid assets significantly.None of the model extensions alters the insights we gained with respect to the short-term payment obligations and, more specifically, with respect to sight deposits.The results of these extensions are not reported but are available upon request.
As for bank capital, the recent literature hypothesizes that it absorbs risks.When banks mainly employ purchased liquidity techniques, banks' capital is expected to affect securities holdings negatively.A well-capitalized bank may raise funds at a lower cost, as capital absorbs risks and expands a bank's risk-bearing capacity (the risk-absorption hypothesis, see Bhattacharya/Thakor (1993), Repullo (2004)).As a consequence, a well-capitalized bank is expected to have only the amount of liquid assets required by the regulator and to use the remaining funds for lending.Since savings banks purchase only a relatively small amount of their liquidity (see Table 1), it is, however, unlikely that bank capital has a tremendous risk-absorbing effect for them.Alternatively, Principle I, which implemented the Basel I Accord in Germany and was in force during our sample period, may cause a negative relationship between bank capital and liquid assets.A bank very close to the regulatory threshold of the prudential capital rules may not increase loans, since it would then fail to meet regulatory capital requirements (except if it can increase regulatory capital).Thus, the bank would hold each additional unit of debt liabilities in those liquid assets that enter the liquid assets specified under the liquidity regulation but not the risk-weighted assets specified under Principle I.By contrast, a well-capitalized bank can decide whether or not to increase loans or securities.In our model extensions, we do not employ the Basel I capital ratio, but rather loans, which determine the denominator of the capital ratio, and regulatory bank capital, which is the numerator of the capital ratio.
We run several regressions to test the impact of bank capital on liquid assets.However, the results are inconclusive.When bank capital is additionally included in our baseline models, it turns out to be significantly positively related to liquid assets, contrasting with the view of risk absorbing capacity.When the loan-asset ratio is removed from the baseline models, bank capital has no significant impact on total liquid assets and securities, while it impacts on cash balances significantly negatively.
As for risks in banks' loan portfolio, we use new loan loss provisions and loan write-offs divided by loans to non-banks and a Herfindahl-Hirschman index based on exposures to 23 sectors.New loan loss provisions and loan write-offs are used to approximate credit risk (e. g., Merkl/Stolz (2009)) that determines the certainty of repayments from loans maturing within the next month.The Herfindahl-Hirschman index is used to approximate concentration risk that may impact on liquid assets since the degree of diversification of a loan portfolio determines banks' resilience against sectoral shocks.Banks with lower concentration risk due to a well-diversified loan portfolio may hold fewer liquid assets since they are less exposed to sectoral shocks than banks with a specialized loan portfolio.However, these risk measures do not impact significantly on liquid assets.

VI. Conclusions
We analyzed the relationship between liquid assets and sight deposits, a relationship that gives insights into whether savings banks transform sight deposits into illiquid assets as intensively as permitted by the regulator.For each unit of sight deposits, banks have to show 0.1 units of liquid assets that contain, e. g., securities holdings, cash balances and repayments from loans maturing within the next month.Thus, if banks receive sufficient repayments from loans maturing within the next month, they can use all sight deposits at their disposal to grant illiquid loans to non-banks and credit institutions or to invest into securities stated as fixed financial assets.We formulated two predictions of why savings banks hold an amount of liquid assets for each unit of sight deposits that is higher than required by the regulator: (i) the actual withdrawal rate for sight deposits is higher than the regulator assumes (underestimation effect), and/or (ii) savings banks have limits in their non-bank lending that they do not offset by medium-term interbank lending or investments in other illiquid assets (lending effect).
Our analysis showed that savings banks actually hold an amount of liquid assets for each unit of sight deposits that is much higher than required by the regulator.As to the underestimation effect, we documented that in our sample the deposit withdrawal rate assumed by the regulator can be regarded as conservative.Thus, the underestimation effect is not present in our sample.As to the lending effect, we investigated whether non-bank lending impacts on how much sight deposits banks hold in liquid assets.Our findings suggest that savings banks with high lending to non-banks relative to total assets do not only have fewer liquid assets but do also hold a smaller volume of sight deposits in liquid assets than banks with low lending to non-banks.These findings indicate that it is more profitable for savings banks to hold liquid assets than to invest in illiquid assets, such as medium-term interbank lending to other credit institutions.However, even savings banks with high shares of loans to non-banks hold more liquid assets per unit of sight deposits than regulatorily required.We discussed several alternative explanations for why even savings banks with high shares of loans to total assets hold more liquid assets per unit of sight deposits than regulatory required.One explanation was that banks report the amount of sight deposits available at the end of the month, monthly averages of sight deposits might be lower.
Our findings suggest two areas for further research.First, while the impact of prudential capital regulation on bank behavior has been well analyzed for banks located in several countries, such as the United States and Germany, little is known about the impact of prudential liquidity regulation on bank behavior.Of particular interest is whether prudential liquidity regulation puts banks under pressure to increase their liquid assets or to decrease their short-term payment obligations as they converge to the regulatory threshold of the liquidity ratio.Second, the potential interaction between prudential capital and liquidity regulation is a relatively unexplored research area.Our findings show no clear relationship between savings banks' regulatory bank capital and liquid assets.However, we might only identify how prudential liquidity and capital regulations interact when regulatory pressure caused by capital or/ and liquidity regulation is modeled jointly.In our paper, we did not focus on these questions but leave them for future research.

VII. Data Definitions and Sources
Dependent variables (Source: Deutsche Bundesbank, Principle II)

LA j
Liquid assets relative to total assets with j ã A for total regulatorily specified liquid assets, j ã S for regulatorily specified securities based on market and book values, and j ã C for cash balances.

RHS variables LB k
Payment obligations relative to total assets with k ã T for total regulatory short-term payment obligations, k ã S for regulatory sight deposits of non-banks, and k ã O for other regulatory short-term payment obligations (Source: Deutsche Bundesbank, Principle II)

L
Loans to non-banks relative to total assets (Source: Deutsche Bundesbank).

D L
A dummy variable equal to 1 if L of the bank under focus is larger or equal than the 70 percentile of L, and 0 otherwise.

DL
Growth in loans to non-banks (Source: Deutsche Bundesbank).

IB
Number of interbank connections a bank has as a borrower relative to total assets (Source: Deutsche Bundesbank, Credit Register). 5

IM
Interest margin calculated as interests received divided by total outstanding loans less the costs of funding (Source: Deutsche Bundesbank).

CAP
Regulatory capital relative to total assets (Source: Deutsche Bundesbank, Principle I).

LL
Loan loss defined as new loan loss provisions and loan writeoffs divided by loans to non-banks (Source: Deutsche Bundesbank).

HHI
Herfindahl-Hirschman index of the loan portfolio over various sectors (Source: Deutsche Bundesbank, Borrowers Statistics).6 Di Change in the short-term interest rate (EURIBOR 1-month) (Source: Thomson Financial Datastream).

Disclaimer and Acknowledgements
This paper represents the authors' personal opinions and does not necessarily reflect the views of the Deutsche Bundesbank or its staff.The authors would like to thank Andreas Pfingsten, Peter Nippel, Ben Craig, Marco Wilkens, Ralf Elsas, Klaus Düllmann, Thilo Liebig and participants at the Bundesbank seminar, the ZEW seminar in Mannheim, the banking workshop in Münster, the Kleistvilla workshop in Potsdam and an anonymous referee for most helpful comments.The paper was written while Andrea Schertler was visiting the Deutsche Bundesbank.The hospitality and financial support of the Deutsche Bundesbank are gratefully acknowledged.All remaining errors, omissions and inaccuracies are the sole responsibility of the authors.

Summary
The Liquidity Regulation and Savings Banks' Liquid Assets For their short-term payment obligations, savings banks hold substantially more liquid assets than the liquidity regulation requires.This paper investigates whether sight deposits, an important funding source for savings banks, help in explaining liquid asset holdings in excess of regulatory requirements.We analyze whether savings banks transform sight deposits in illiquid assets less intensively than is permitted because (i) the liquidity regulation underestimates actual withdrawal rates (underestimation effect) and/or (ii) savings banks are subject to limits in their lending to non-banks that they do not offset by, for instance, mediumterm interbank lending or fixed asset holdings (lending effect).In our sample, we do not find the underestimation effect to be applicable as actual deposit withdrawal rates are in most cases lower than the regulatorily specified rate.However, we find the lending effect to be at work: Savings banks with low shares of loans to non-banks do not transform sight deposits into illiquid assets as intensively as savings banks with high shares of non-bank loans.Our analysis does not only show that liquid assets positively depend on sight deposits, but also shines a light on how bank size and the individual bank's position in the interbank market affect liquid assets.(JEL G21)

Table 1
Main Positions in Savings Banks' Regulatory Liquidity Ratio 1