1
How does capital influence banks’ liquidity risk?
Supervisor: S.R.APRING
Name: LI JUAN
Student Number: 10824278
Word account: 10194
MSc Business Economics Finance
2
Statement of Originality
This document is written by Student LI JUAN who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3
Abstract
Nowadays capital becomes an extremely important part of banks’ balance sheet, as capital can act as buffer when liquidity shortfall comes, however, at the same time capital also causes several problems related to liquidity risk for banks. Among the current theories, one argues that capital reduces liquidity risk while the other holds the opinion that increasing capital in fact increases the probability of liquidity risk. This thesis will imply the OLS model to explore the relationship between banks’ capital ratio and liquidity risk, measured by liquidity creation ratio. In addition, I will decompose liquidity creation ratio into two parts to see whether capital’s influence on assets is the same as its influence on liabilities, as well as introduce a lag variable to address the endogenous problem embedded in independent variable. Since my data is the US banks’ quarterly balance sheet positions from 2004 to 2013, covering the period of subprime crisis, it’s also necessary to use a sub sample of banks with higher pre-crisis capital ratio to solve the question whether these banks encountered more liquidity risk during crisis.
Keyword: capital ratio, liquidity risk, liquidity creation ratio, asset side originated liquidity problem, liability side originated liquidity problem, subprime crisis, US banks.
4
Table of content
1 Introduction
52 Literature review
92.1 Capital buffer theory 10
2.2 Financial fragility structure theory 11
2.3 Deposit crowding-out effect theory 12
2.4 Effort-aversion moral hazard theory 12
3 Method and sample
133.1 OLS model implied 13
3.2 Control variables 16
3.3 Basic hypotheses 19
4 Results
204.1 Regression result of LC on capital ratio 24
4.2 Regression result of asset side LC on capital ratio 28
4.3 Regression result of liability side LC on capital ratio 29
4.4 Regression result of banks with top 5% pre-crisis capital ratio 31
5 Conclusion
345
How does capital influence banks’ liquidity risk?
1 Introduction
Liquidity risk for banks consists of two parts: market liquidity risk and funding liquidity risk. The first one refers that assets can’t be sold in the market because of lack of liquidity while banks need to sell assets in exchange of money, and the second one means borrowers, in this situation refers to banks, are unable to meet liabilities when they come due, meaning they have difficulty in funding the required payment for people who put money under their custody. In recent years, banks gradually transfer from traditional “originate and hold” business model to “originate to distribute” model. Banks usually fund through issuing structured products and using short-term repo agreements, in consequence banks are exposed to more market liquidity risk and fund liquidity risk (Giordana and Schumacher, 2013). That is to say, in the past, when loans were granted, they will be remained on the banks’ balance sheet as their assets, mostly illiquid assets. But with the development of new financial instruments, banks found that it may be more profitable to package the loans and sell them to others, by doing so they can obtain more funds as well as liberate themselves from duties of monitoring the loans, this laying off of assets in operation leads to less market risk for banks.
When loans are treated as assets of banks, the problem in loans, such as increase in non-performing percentage, could lead to depression in the banks’ share price or even regulatory penalty. In some countries, when the non-performing ratio rises above certain benchmark, central bank will force the subject to reduce their loan-to-deposit ratio, harming the subject’s ability to generate profitability. But when banks could sell these loans, things change. After generating loans, they transfer them into structured financial products, to name a few, collateralized debt obligation, mortgage backed securities and asset backed securities, and sell them to investment banks. During this process, the seed of liquidity risk occurs. Firstly, since originators retain no interest in these multi-tranched securities, their motivation to exercise care and prudence is weakened. What’s more, as the loans are removed from the balance
6
sheet, they won’t post influence on banks’ risk control, consequently, in order to sell more CDOs and MBSs to get more cash to grant new loans, banks could even lax the lending standards to attract some unqualified borrowers, this moral hazard induces banks to take more risk at the same time. Secondly, it’s not hard to find that the long parties usually don’t have sufficient reserves, making them quite vulnerable in default, which would result in liquidity risk for the entire market.
For example, in August29, 2006, Citibank bought 4499 mortgage loans from the New century financial corporation at the price of 979,000,000 dollars, then Citibank issued MBS against these mortgage to investors around the world. These MBSs had different tranches to satisfy the need of diversified investors. The percentage of equity tranches in CMLTI 2006-NC2 was less than 2%, if the loss of that mortgage loan packages exceed 2%, the capital defense of equity tranches will be exhausted, since the sub tranches couldn’t absorb losses any more, the mezzanine tranches would suffer loss, and most of the mezzanine tranches were sold to CDOs. In fact, the two main institutional investors of mezzanine tranches of these MBS stepped into dilemma in 2007. One of them is Cheyne Finance Limited, a special purpose entity bought 7 million dollars, went bankruptcy in the summer of 2007. And the other one is Parvest ABS Euribor, who spent 20 million dollars in these MBSs. At that time, most mezzanine tranches of CMLTI 2006-NC2 were sold to CDOs, among them there were 12 million dollars’ M9 tranches. These M9 securities were used as nominal underlying assets in the synthetic CDOs of Auriga, Volans and Neptune CDO IV. To buyers and sellers of the synthetic CDOs, whether the M9 tranches default or not is quite crucial to their profit. Supposing that other nominal underlying assets remain the same, if the M9 didn’t default, sellers of CDS would constantly paid CDOs swaps. But when default occurred, the buyers of CDS need to pay to CDOs according to the sequence of absorbing loss. Under the CDO contracts where two parties both have collaterals, even though the M9 tranches hadn’t default yet, the long must increase collateral in the case where the M9 or the CDS buyer’s credit rating was lowered. We could see that the nominal principal of CMLTI 2006-NC2’s M9 tranches in synthetic
7
CDS exceeded 50 million dollars in the beginning of 2007, even though the mezzanine tranches only had 12 million in fact. In consequence, the default of M9 involved other 50 million at that time. Since the long parties expected M9 won’t default and the short parties would have to fulfill the payment obligation within a few years, they didn’t retain enough cash in hand. However, in July, Standard & Poor’s and Moody began to depress the credit rating of subprime loan securities, the last three mezzanine tranches of CMLTI 2006-NC2, M9, M11 and M12 were included in this storm. At that time, 12% of CMLTI 2006-NC2’s respective mortgage loans were repaid before due time and 11% were incurred overdue more than 90 days. Following this, the M4 to M8 tranches were lowered in October. As a result the long parties stepped into liquidity trouble because of lack of cash, and this liquidity risk quickly spread over the whole market.
As we can also learn that before the subprime mortgage crisis, US banks often sold the loans to third parties in OTD market, rather than remain them in balance sheet, under this circumstance banks’ incentive to prudently review and monitor is reduced, many loans of excessively poor quality came into being at that time, which amplified the following bust as a result.
When explore the reason behind this, Amiyatosh Purnanandam found that observed quality of borrowers, cost of capital or geographic location of houses can’t explain this this result comprehensively, his study supports that the true reason is that the banks didn’t execute enough efforts to screen the borrowers. In a word, nowadays, with the development of the financial market, banks now face more liquidity risk than in the past.
Banks, one of the most vital financial intermediaries in today’s market, face random withdrawal from their depositors, when failed to meet depositors’ demand in time, they will incur serious penalty. Then in order to prevent these costly liquidity shortfalls, banks need to hold capital, especially regulatory capital, which consists of common stocks, retained earnings, capital surplus and disclosed capital reserves, as one of their liquidity supply. This capital would certainly post some influence on
8
banks’ liquidity risk. However, among current theories, there are different opinions about this influence, this thesis will try to find out the relationship between capital ratio and liquidity risk of US banks, which is measured by liquidity creation ratio, by regressing LC ratio on capital ratio and several necessary control variables and a dummy variable that tells normal period from crisis period as my sample will cover the US subprime crisis, based on the US banks’ quarterly data from 2004 to 2013. As is known to us all, in the period from 2007 to 2009, US economy was depressed by the serious financial crisis, so there is probability that the relationship between capital and liquidity risk during crisis may be not as same as normal time relationship, as to take this possibility in to account, I will include a dummy variable of time D in the OLS model, if the data is from 2007 to 2009, D equals 1, otherwise D will be 0. In the meantime, it’s necessary to include other control variables because liquidity risk is not only influenced by capital ratio, macro- economic factor, treasury rate, S&P 500 index and gross total assets of the banks are all important control variables for the regression.
After finishing the basic regression in the first step, whether the banks that had high capital ratio before the financial crisis took place would come across more liquidity risk during the bust is an issue worth discussion as well.
At last, although this article uses liquidity creation ratio to measure liquidity risk, this ratio has some shortcomings. In the calculation of LC ratio, the numerator is weighted sum of different balance sheet positions, including both assets and liabilities, and the denominator is total assets. It’s not proper to mix assets and liabilities together in this ratio, since capital may affect liabilities and assets in different ways and to different extent. From a general point of view, liquidity risk could come from two sources: the asset source and the liability source. For instance, if the creditors, under these circumstances are depositors, claim their money but the bank is unable to meet the obligations within limited time period, liquidity problem comes into being, and this liquidity risk rises from the liability side. Or sometimes the market value of assets fall below certain level and make it unable to conduct its daily
9
operation efficiently, here asset side liquidity risk comes. So measuring liquidity risk by using this LC ratio will make the relationship between capital ratio and liquidity risk ambiguous to some degree, in order to avoid the mutual interfere of asset side and liability side, we should explore the effects separately by splitting this ratio into asset side LC ratio and liability side LC ratio, then regress them separately to get two coefficients of capital ratio, and by comparing these two coefficient, we could conclude whether capital ratio has same impact on asset side originated liquidity risk and liability side originated liquidity risk, and if the two coefficients are not equal, comparing their value will give an insight which one is influenced more profound by capital.
This article is structured as following: the related literatures section describes major current theories about banks’ capital’s effects, the method and data section describes the sample I use, the regression model and hypotheses and the results section presents the regression results and tables, the last is the conclusion section. From the regression results we could find out that the capital ratio has negative impact on banks’ liquidity risk, the higher the capital ratio, the higher the liquidity creation ratio, and the influence on asset side is different from that on liability side. What’s more, the banks with higher capital ratio before the subprime mortgage crisis actually faced less liquidity shortfall during bust, after we put lag value of capital ratio into the OLS regression. In sum we could get the conclusion that only increasing capital ratio is not enough to reduce liquidity risk for banks, especially when banks already hold enormous liquid assets on their balance sheet.
2 Literature review
As referring to current literatures, there are two main viewpoints about capital’s impact on liquidity risk. The first theory holds the opinion that capital can act as buffer to absorb loss when banks are faced with liquidity shock and more capital means less liabilities such as short-term debts. According to the empirical test conducted by Berger and Bouwman on US banks, they measure banks’ performance, whose indicators are banks’ market capitalization and probability of survival in three
10
situations: the normal time, the bank crisis which derives from the banking sector and the market crisis period. Their research focuses on the data base of US banks from the first quarter of 1984 to the last quarter of 2010. Assuming that the effects of capital may depend on banks’ size, they divided the sample into three subgroups after dropping those are “to large to fail”: banks with gross total assets less than 1 billion dollars belong to the small group, banks whose total assets were between 1 to 3 billion dollars belong to the medium group and banks with gross total assets more than 3 billion dollars are the so-called large ones. Finishing the collection of data, they began to construct the regression model. In the first step, Berger and Bouwman implied the logit regression method to explore the contribution of capital to banks’ survival probability, they use the logic odd ratio of survival probability to be dependent variable and capital ratio as independent variable. Based on their realization that other factors also influence banks’ survival, necessary control variables such as size, safety net protection, location, competition, risk proxy and profitability were included as control variables.
Secondly, in order to see the impact on market share, another model with similar control variables was built, the independent variable is average pre-crisis capital ratio and the dependent variable is percentage change in market share. Besides, in order to see the interaction between crisis and market share change, the authors constructed three dummy variables, bank risk dummy, market risk dummy and normal time dummy. Through the regression result and robustness checks they found that higher capital ratio contributes to small banks’ market capitalization and survival in both usual and crisis period. While as to large and medium size ones, capital also improved their performance in bank crisis times, a good example is the US credit crunch in early 1990s. This positive effect works on three mechanisms: increasing non-core funding, loans reflected on balance sheet and off balance sheet guarantee. What’s more, they also have other supporting findings: First of all, even for banks that failed in crisis, their past capital ratio also affects their resolution methods, for instance, most banks with higher pre-crisis capital ratio exit through
11
unassisted merger and acquisition rather than government supported merger and acquisition （Berger and Bouwman, 2013）.
The reason why capital reduces liquidity risk is that capital increases banks’ skin in the game, that is to say, they have something to lose. Without necessary capital, banks have incentive to seek excessive risk, especially under the situation of debt overhang, because they could benefit from the success of risky projects. And most of their liabilities, deposits are guaranteed by deposit insurance, so when the risky projects fail, the cost is bear by the government and taxpayers. In other words, without necessary capital, banks only enjoy the upside potential of risk taking but don’t pay for the failure, this mismatch in risk and return profile leads to over leverage and excessive liquidity risk. But after the introduction of capital requirement, banks are force to hold certain quantity of capital in their balance sheet, now they would lose as well if the excessive risk leads to collapse of operation, then they will exercise reasonable efforts to control liquidity risk. But In terms of accounting, as far as I am concerned, if banks hold more capital, there will be less debt on their balance sheet, especially short-term debts, this higher capital is likely to generate more risk as banks have more illiquid capital holding in hand.
At the same time, the second viewpoint argues that capital impedes liquidity in two ways: the “financial fragility structure” (Diamond and Rajan, 2000 and Diamond and Rajan, 2001) and “crowding-out of deposit” effect (Gorton and Winton, 2000). Financial fragility structure argues that, the fragility is a desired character for banks. As financial intermediary providing various services to the market, banks needs to do efforts to reduce agency problem and improve services to attract more depositors, As indicated by Diamond and Rajan, banks creates liquidity by lending the money collected from depositors to borrowers, because the depositors can withdraw money with short notice and the assets side of bank’s balance sheet is illiquid, making it difficult to sell or collateralize at full value when bank has liquidity need, this mismatch of liquidity between assets and liabilities is where liquidity risk comes from . So banks need to issue fresh demand deposits as long as the depositors are
12
confident that a bank has the ability to repay, which means that the bank’s loan collection ability persists. When depositors lose confidence in a bank, they would claim back their money and new depositors are reluctant to give their money to the bank, result in inability for banks to grant loans to generate profits and liquidity problem of satisfying depositors’ claim. Worse still, when a large number of depositors come to the bank to call their funds at the same time, a bank is incapable of obtaining enough money as most of its assets have longer maturity than that of liabilities, then bank run arises. What’s more, after one bank falls down, people will run on other banks, banks have to sell their assets at discount price in exchange of funds, this fire sale would depress other banks’ assets, causing downward spiral to the whole bank sector. Then liquidity risk would spread to the whole financial sector. However, when banks have sufficient capital, their bargaining power against depositors improves, so they have less motivation of risk management and agency problem solving. From this point of view, capital actually increases liquidity risk. On the other side, according to deposit crowding-out theory, although rising capital ratio has some positive effects, it also reduces the amount that bank can raise, since capital holders would negotiate as they don’t face immediate collective problem as depositors. Deposit’s value is stable, but the value of capital, such as common stocks, would depend on the bank’s fundamentals and exchange liquidity in most cases. When banks hold more capital, they reduce the percentage of stable deposits and face more fluctuation in value, increasing the probability of liquidity shortfall. For example, when crisis comes, liquidity evaporates in the market, some loan borrowers are incapable of making timely payment of loan interest and principal payment or even default on their loan, the value of loan package products, for example, mortgage backed securities, suffers negative shock, this illiquidity can eat up capital quickly.
Additionally, Besanko and Kanatas also do some research on whether bank capital regulation really achieves the risk mitigation objectives, they emphases on effort-aversion moral hazard and argues that the net effect of capital requirement
13
would be more risk taking because even though this requirement solves one agency problem, it causes another problem at the same time. In their model, they set up the premise that bank insiders were capable of taking some actions that maximize their own benefits while these actions might harm the welfare of outsiders, and these actions are not detected by outsiders. When required to hold more capital, the insiders’ portion on surplus that depends on successful repayment of loans was reduced, so they may exercise less efforts to make sure the loan repayment. In addition, as the announcement of capital requirement leads to decline in bank’s share price, the decline is related to the portion of insider holdings, the higher the percentage of management’s holding, the less the slump in share price. Their conclusion that enforcement of capital requirement increases risk in fact is, because of the dilution effect on insiders’ holding. As their shares of benefit decreases, insiders have less incentive to protect safety of banks (Besanko and Kanatas, 1994). Generally speaking, there are three ways for banks to improve their liquidity position: holding liquid assets, interbank market coinsurance and increasing capital ratio. As interbank borrowing and lending is costly, and sometimes it also has numerous barriers for entry, such as market capitalization and assets under management, banks may prefer the other two methods. That is to say, banks with higher capital ratio are less eager to take part in the interbank market (Fabio, Gyongyi and Loriana, 2014), so it’s hard to measure precisely the total influence of capital on liquidity.
As it’s known to us that financial intermediaries have incentive to gamble because they are not risking their own money, it’s unclear that whether capital is enough to induce them to avoid inappropriate risk taking. Now capital plays an ambiguous role in banks’ liquidity risk management, its true effects worth research, I will imply OLS model to explore the relationship between liquidity risk and capital ratio.
3 Method and sample
To find out the impact of capital, the first thing I need to do is to define an appropriate measurement for banks’ liquidity risk, which is the dependent variable in the OLS regression. In this thesis I will use the liquidity creation ratio, invented by
14
Berger and Bouwman. High LC ratio means high liquidity risk since banks transfer more liquid liabilities into illiquid assets, and this mismatch in maturity of them is where liquidity risk originates. In calculation of LC ratio, we need to assign different weights to balance sheet positions according to their ability to convert into liquidity within certain time period and without incurring significant loss in value. Here is the table of balance sheet positions weight:
Table 1: balance sheet position weight by Berger and Bouwman
Assets Liquidity level Weights
Cash and near cash items Liquid −0.5
Interbank assets Semiliquid 0
Short-term marketable assets Liquid −0.5
Commercial loans Illiquid 0.5
Consumer loans Semiliquid 0
Other loans Semiliquid 0
Long-term marketable assets Semiliquid 0
Fixed assets Illiquid 0.5
Other assets Illiquid 0.5
Customer acceptances Semiliquid 0
Liabilities Liquidity level Weights
Demand deposits Liquid 0.5
Saving deposits Liquid 0.5
Time deposits Semiliquid 0
Other term deposits Semiliquid 0
Short-term borrowings Liquid 0.5
15
Assets Liquidity level Weights
Long-term borrowings Semiliquid 0
Other long-term liabilities Semiliquid 0
Subordinated debentures Illiquid −0.5
Preferred equity Illiquid −0.5
Minority interests Illiquid −0.5
Shareholder common capital Illiquid −0.5
Retained earnings Illiquid −0.5
And LC =
0.5∗𝑖𝑙𝑙𝑖𝑞𝑢𝑖𝑑 𝑎𝑠𝑠𝑒𝑡+0∗𝑠𝑒𝑚𝑖𝑙𝑖𝑞𝑢𝑖𝑑 𝑎𝑠𝑠𝑒𝑡−𝑜.5∗𝑙𝑖𝑞𝑢𝑖𝑑 𝑎𝑠𝑠𝑡𝑒𝑠+0.5∗𝑙𝑖𝑞𝑢𝑖𝑑 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦+0∗𝑠𝑒𝑚𝑖𝑙𝑖𝑞𝑢𝑖𝑑 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦−0.5∗𝑖𝑙𝑙𝑖𝑞𝑢𝑖𝑑 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡
Here is one thing we need to note, in the calculation of LC, this ratio simply adds everything together after multiplying by their weight. However, intuitively speaking, it’s obvious that capital’s impact on assets is not the same as its impact on liabilities. Besides, liability positions and asset positions are put together in the numerator while the denominator is total assets. From a basic point of view, liquidity risk has two resources, it may arise either from the asset side or liability side of balance sheet, different resources may lead to different characteristics and therefore need respective measurement. For instance, on one hand, if the assets of a bank, like short term marketable securities which would be recorded at fair value on the balance sheet, surfs sharp decline in value, or its borrowers default on loans, the bank is likely to encounter liquidity trouble, here the liquidity risk originates from the asset side. While on the other hand, if the bank has problem to repay its short-term borrowing, in most cases it means the bank has insufficient funds to satisfy depositors withdrawal, this illiquidity is derived from liability side. From this point of view, the calculation of liquidity creation ratio is kind of problematic, its underlying assumption is that more liquid assets and illiquid liabilities mean less liquidity risk, while it’s more
16
reasonable to study the asset originated liquidity risk and liability liquidity risk separately to find out whether capital posts same impact on them. So after finishing the initial regression of LC ratio on capital ratio, necessary control variables and dummy variable, we should take it apart into asset side LC ratio and liability side LC ratio to see the respective reaction to capital ratio by regressing asset side of LC and liability side of LC separately.
𝐿𝐶𝑎𝑠𝑠𝑒𝑡 = 0.5 ∗ 𝑖𝑙𝑙𝑖𝑞𝑢𝑖𝑑 𝑎𝑠𝑠𝑒𝑡 + 0 ∗ 𝑠𝑒𝑚𝑖𝑙𝑖𝑞𝑢𝑖𝑑 𝑎𝑠𝑠𝑒𝑡 − 𝑜. 5 ∗ 𝑙𝑖𝑞𝑢𝑖𝑑 𝑎𝑠𝑠𝑒𝑡
𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡 𝐿𝐶𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦
=0.5 ∗ 𝑙𝑖𝑞𝑢𝑖𝑑 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦 + 0 ∗ 𝑠𝑒𝑚𝑖𝑙𝑖𝑞𝑢𝑖𝑑 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦 − 0.5 ∗ 𝑖𝑙𝑙𝑖𝑞𝑢𝑖𝑑 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡
Next step is to define the necessary control variables as capital ratio is not the only factor influencing liquidity risk of banks, we must include control variables to avoid omitted variable bias. There are several factors that affect liquidity:
1 Percentage change in GDP
Macroeconomic environment and business cycle affect market liquidity remarkably. Business cycle and credit cycle interact with one another, credit cycle amplifies business cycle in most cases, that is to say, in booms, the probability of default is low for borrowers, this virtuous environment improves banks’ confidence and increases their capital base. To seek more profitability, financial institutions are very optimistic about the economy, they would relax the standards of lending and grant more funds to the market, enable the entrepreneurs to extend production further. While in the bust, anticipating decline in asset price and high default probability, banks will choose to impose harsh standards of lending and reduce supply of credit to the market, from historical evidence, more business failures follow a change in lending standards. In the mean time, there is another opinion, when macroeconomic environment starts to deteriorate, people expect the probability of loss to be higher, and they are reluctant to invest, they would like to deposit their money in the banks, where they perceive to be safer than investing in the market. With these abundant
17
deposit inflows, banks are likely to widen the loan standards and grant more loans. The reason behind the phenomena is that in most cases loan officers are compensated based on their loan volume, in order to get more bonus, they may try to grant more loans without prudence and care. This trend stimulates asset bubbles and excessive lending, which sows the seeds of crisis (Gatev and Strahan, 2006). In the OLS model, I will use the percentage change in gross domestic productivity as proxy of macroeconomic environment.
2 Return on S&P 500 composite index and treasury rates.
Other factors that influence banks’ liquidity creation ability are the return in both stock and bond market. Applying bank loans is not the only approach for corporations to raise capital, they could also get access to funds through issuing debt or initial public offering. As the requirement of qualification is often relatively high to get loans from banks and the procedure is complex, compared to stock market and bond market issuing, companies sometimes may turn to stock market and bond market to raise money, especially when the return in these markets are low, result in lower capital costs for the issuers. In my OLS model, the quarterly S&P 500 index return and three month T-bill rate will reflect the condition of these markets. By the way, as for floating rate bonds, the expectation of interest rate in the future also plays an important role in investors’ preference for these issues, the long-term interest rate should be taken into consideration, so I also include the 10 year treasury bond rate as a control variable.
3 Size of the bank.
According to Berger and Bouwman’s study, when exploring banks’ capital ratio’s relationship with their survival probability and market capitalization, they divided the sample into three groups based on gross total assets, and found that size does make sense. The positive influence of capital works for small size banks all the times, while for medium and large banks, capital only contributes to their success of survival during bank crisis, which is the crisis that coming from the bank sector. Additionally, small banks often suffer more mismatch between the maturity of assets and
18
maturity of liabilities when short term interest rate is driven high by monetary policy. They are likely to lend less compared to large banks since they can’t compensate the reduction in core deposit by finding other alternative funding sources as those large banks (Angeloni, Kashyap,and Mojon 2003; Chatelain 2003; Ehrmann 2002; Kashyap and Stein 1995; Kishan and Opiela 2000; Peek and Rosengren 1995).
Moreover, since my sample is the 10 year quarterly data of US banks from 2004 to 2013, it covers the period the subprime crisis of 2007 to 2009, and there is a probability that the crisis may distort the relationship between capital ratio and liquidity risk, the relationship in normal times may be different from that during crisis period, and the extent of capital’s effects may depend on the economic condition. I think it’s better to create of dummy variable of time to identify the different time periods, in order to observe the influence of crisis. If the data is from 2007 to 2009, the dummy variable equals 1 and otherwise it will be 0.
Now we could finish the construction of the OLS model:
𝑋𝑐𝑎𝑝𝑖𝑡𝑎𝑙 𝑟𝑎𝑡𝑖𝑜 = 𝑒𝑞𝑢𝑖𝑡𝑦 𝑎𝑠𝑠𝑒𝑡 LC= 0.5∗𝑖𝑙𝑙𝑖𝑞𝑢𝑖𝑑 𝑎𝑠𝑠𝑒𝑡+0∗𝑠𝑒𝑚𝑖𝑙𝑖𝑞𝑢𝑖𝑑 𝑎𝑠𝑠𝑒𝑡−𝑜.5∗𝑙𝑖𝑞𝑢𝑖𝑑 𝑎𝑠𝑠𝑡𝑒𝑠+0.5∗𝑙𝑖𝑞𝑢𝑖𝑑 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦+0∗𝑠𝑒𝑚𝑖𝑙𝑖𝑞𝑢𝑖𝑑 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦−0.5∗𝑖𝑙𝑙𝑖𝑞𝑢𝑖𝑑 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡 𝑌𝐿𝐶 = 𝛽0+ 𝛽1𝑥𝑐𝑎𝑝𝑖𝑡𝑎𝑙 𝑟𝑎𝑡𝑖𝑜+ 𝛽2𝑋∆𝐺𝐷𝑃+ 𝛽3𝑋𝑆&𝑃500+ 𝛽4𝑋3 𝑚𝑜𝑛𝑡ℎ 𝑇−𝑏𝑖𝑙𝑙 + 𝛽5𝑋10 𝑦𝑒𝑎𝑟 𝑏𝑜𝑛𝑑+ 𝛽6𝑋𝐺𝑇𝐴+ 𝛽7𝐷𝑐𝑟𝑖𝑠𝑖𝑠+ 𝜀
My sample is the 10-year US banks’ data from 2004 to 2013, I collect the US banks’ balance sheet statements from Wharton research data service’s database, after dropping the banks whose balance sheet don’t cover the whole ten year period and whose balance sheet positions are not completed (some of the positions are blank or equity equals 0), there are 731 banks left, generate 29240 observations for the OLS regression.
19
If 𝛽1 < 0, then increasing capital ratio is an effective method to reduce banks’
liquidity risk or vice versa.
Hypothesis 2: banks that have higher capital ratio before the crisis tend to have less liquidity risk during the crisis.
As we now have the balance sheet positions of banks, we could calculate capital ratio of them at the last quarter of 2006, right before the break up of subprime crisis, using the quarter capital ratio can remove the end-of-year bias on December to some degree since sometimes banks are likely to make their financial statements more attractive at the last month of a year. After calculating the last quarter capital ratio, pick up the banks with top 5 percent capital ratio, which means the 37 banks with highest capital ratio from the 731 banks in sample, and put them into the OLS regression model to get 𝛽1∗, if 𝛽1∗ < 𝛽1, then we could conclude that banks with
higher pre-crisis capital ratio met less liquidity risk in bust, otherwise we could say higher capital ratio didn’t guarantee less liquidity risk during subprime crisis.
Here we need to pay attention to the interaction of capital ratio and the subprime crisis, as capital ratio affects the liquidity risk during crisis, at the same time liquidity risk affects capital ratio as well. For example, in bust, banks face liquidity problem, they would convert their capital holding, such as Cocos (contingent convertible notes that could be converted into equity under certain conditions, to name a few, if the bank’s regulatory capital ratio falls below 4%), into funds to solve liquidity problem, and their capital will be burnt up quickly. So as to solve this problem, we need to replace the independent variable in the model with lag capital ratio, which is capital ratio at time t+1, and drop the dummy variable of crisis because of collinearity.
𝑌𝐿𝐶 = 𝛽0+ 𝛽1𝑋𝑙𝑎𝑔 𝑐𝑎𝑝𝑖𝑡𝑎𝑙 𝑟𝑎𝑡𝑖𝑜+ 𝛽2𝑋∆𝐺𝐷𝑃+ 𝛽3𝑋𝑆&𝑃500+ 𝛽4𝑋3 𝑚𝑜𝑛𝑡ℎ 𝑇−𝑏𝑖𝑙𝑙𝑠
+ 𝛽5𝑋10 𝑦𝑒𝑎𝑟 𝑇−𝑏𝑜𝑛𝑑𝑠+ 𝛽6𝑋𝐺𝑇𝐴+ 𝜀
Here is the scatter plot diagram of the capital ratio of US banks at the last quarter of 2006.
20
Hypothesis 3: capital has different influence on asset originated liquidity risk and liability originated liquidity risk.
By replace the LC ratio with the asset part of LC ratio and liability part of LC ratio on the left side of the model, and then compare the two results of regression, we could see whether the impact is different or not.
4 Results
After consolidating the methodology, I collected data from Wharton research database, including monthly return on S&P 500 composite index, US quarterly GDP, banks’ quarterly balance sheet positions as well as quarterly yield on treasury securities, where we use return of 90-day T-bills to represent short-term interest and return of 10-year T-bonds to represent long-term interest rate, the sample consists of data from 2004 Q1 to 2013 Q4, since some banks’ balance sheet contain missing positions and some banks don’t cover the whole time interval, after dropping those are not qualified, we get an sample consisting of 731 US banks, this panel data has 29240 observations. As we need quarterly return of S&P 500 index, I translate the monthly return by calculating their time-weighted quarterly return using the formula
21
of geometric return In which 𝑟1, 𝑟2, 𝑟3 represents three monthly return in the
quarter:
𝑟𝑞𝑢𝑎𝑟𝑡𝑒𝑟 = (1 + 𝑟1)(1 + 𝑟2)(1 + 𝑟3) − 1
Here is the data description of the 29240 observations: Table3 data description for the observations
bank: 1, 2, ... ,731 n = 731 date: 21Mar4, 20Jun4, … ,31Dec13 T = 40 Delta (date) = 1 day
Span (date) = 3563 periods
(bank*date uniquely identifies each observation)
Distribution of T_i : min 5% 25% 50% 75% 95% max 40 40 40 40 40 40 40 Freq. Percent Cum. 𝑃𝑎𝑡𝑡𝑒𝑟𝑛∗
731 100.00 100.00 1111111111111111111111 1111111111111111111
731 100.00 XXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXX
*Each column represents 87 periods
In order to calculate the liquidity creation ratio, we need to assign different weight to every balance sheet position based on its liquidity converting ability, based on Berger and Bouwman’s liquidity level table (This table contains several positions that are missed out in Berger and Bouwman’s ):
Table4 balance sheet position weight based on US banks’ balance sheet
Assets Liquidity level weights
Cash Liquid -0.5
Market value securities Liquid -0.5
Hold to maturity securities Illiquid 0.5
22
Federal funds sold Liquid -0.5
Real estate loans Semiliquid 0
Agricultural loans Semiliquid 0
Commercial and industrial loans Illiquid 0.5
Loans to individuals Liquid -0.5
Lease financing receivables Liquid -0.5
Loan loss allowances Illiquid 0.5
Transfer risk reserves Illiquid 0.5
Intangible assets Illiquid 0.5
Premises and Fixed Assets Including Capitalized Lease
Illiquid 0.5
Liabilities Liquidity level weights
Demand deposits Liquid 0.5
Transactions deposits Semiliquid 0
Fed funds purchased Liquid 0.5
Mortgage debt semiliquid 0
Banks’ liability on acceptances Semiliquid 0
Other liabilities semiliquid 0
Subordinate debt Illiquid -0.5
Limited life preferred stock Illiquid -0.5
Noncumulative Perpetual
Preferred Stock and related surplus
Illiquid -0.5
Mandatory Convertible Debt, net
of common or perpetual
preferred stock dedicated to redeem the debt
Semiliquid 0
23
Preferred stocks Illiquid -0.5
Surplus, not related to preferred stocks
Illiquid -0.5
Undivided profits & capital reserves
Illiquid -0.5
Cumulative foreign currency
adjustment
Illiquid -0.5
After determining the appropriate weight to every balance sheet position, we can now calculate the asset side liquidity creation ratio, liability side liquidity creation ratio and liquidity creation ratio on every quarter of the 731 banks across the 10-year time interval, then regress them step by step on the respective capital ratio, taking quarterly percentage change in gross domestic product, quarterly return on Standard & Poor’s Index, quarterly return on 90-day US treasury bills, quarterly return on 10-year US treasury bonds and the bank’s gross total assets as control variables. Besides, it’s crucial to generate a dummy variable to represent the crisis from 2007 to 2009. Using fixed effects regression on this panel data defined by bank and date, we could get the results as following:
4.1 Whether higher capital ratio leads to lower liquidity risk, reflected by lower liquidity creation ratio.
Table5 regression result of LC on capital ratio
Fixed-effects (within) regression Numbers of obs =29240 Group variable: bank Numbers of groups=731 R-sq: within=0.0476 obs per group: min=39 between=0.0749 avg=40.0 overall=0.0642 max=40 F(7,28345) = 202.47 Corr(u_i,Xb) = -0.0276 (assumed) Prob > F = 0.0000
24
LC Coef. Std.Err t p>︳t︳ 95% confidence interval
Capital ratio Change in GDP Return on S&P 500 Return on T-bills Return on T-bonds Gross total assets Crisis dummy constant .9074 .0264 34.36 0.000 .8556 .9591 -.0007 .0002 -4.62 0.000 -.0010 -.0004 -.0652 .0122 -5.35 0.000 -.0891 -.0413 1.5215 .1669 9.12 0.000 1.1944 1.8487 -.0879 .0240 -3.66 0.000 -.1349 -.0409 -7.88e-10 2.29e-10 -1.10 0.001 -1.24e-09 -3.39e-10 .0083 .0018 -1.10 0.270 .0048 .0118 -.2347 .0029 -81.05 0.000 -.2404 -.2291 sigma_u sigma_e rho .1603079 .13051233
.6013893 (fraction of variance due to u_i)
F test that all u_i = 0 F(726, 28345) = 60.20 Prob > F = 0.0000
From this table we could find that the p values for all variables and constant, except for crisis, are significant at 5% significance level, even at 1% significance level. Based on this point we could infer that the happen of subprime crisis didn’t distort the relationship between capital ratio and banks’ liquidity creation ratio, It will be checked on the separate regression of asset side LC ratio and liability side LC ratio regression later.
Since the coefficient of capital ratio is 0.9074 > 0, and it is statistically significant, then the higher the capital ratio, the higher the liquidity creation ratio, meaning holding more equity results in higher liquidity risk for banks, which is contradict to hypothesis 1. As come to the control variables, the coefficients for percentage change in GDP, return on S&P 500 index, 10 year treasury bond and banks’ gross total assets are all negative, reflecting their positive contribution to reduction in liquidity risk, put it in another way, the growth in GDP, increase in stock market return and
25
long-term interest rate expectation, and strength of banks’ assets base are beneficial for liquidity of banks.
This is intuitively easy to understand. Firstly, when GDP grows stably, creating a healthy macro-economic environment for the whole financial market, the market participants, including financial intermediaries, investors, entrepreneurs and government all have more confidence on the market. Banks are willing to lend as they perceived the probability of default is low when macro-economic environment is optimistic, as a result they generate more profits and strength their asset base, and the efficient operation of the whole market lower the probability of defaults of borrowers further, banks’ liquidity position will be improved at the same time. However, when there is negative growth in gross domestic productivity, banks perceive the probability of default increases, they become reluctant to lend, contracting their lending standard, which would worsen the evaporation of liquidity in the market, more enterprises whose cash flow chain is broken step into difficulty and are unable to make interest and principal payment to the banks, at last banks encounter liquidity problem as well, here the credit cycle amplifies the business cycle.
Secondly, when the return of stock market is high, the cost of raising capital through issuing stock for a company is high, enterprises choose to turn to banks for funds rather than focus on IPO issues. Meanwhile, as the return on stock market is high, it becomes more attractive for investors, investors are more likely to throw their money into stock market instead of deposit it in the bank, under this situation the banks’ assets increase and their liabilities decrease, making them face less liquidity pressure.
Thirdly, when the market considers the long-term interest rate curve to increase smoothly, their perspective about the economic is optimistic, banks are glad to lend and borrowers are less likely to default, the credit cycle and business cycle interact with each other healthily, banks face less liquidity problem as well.
26
At last, banks with large gross total assets have higher ability to bear liquidity shock and they can recover from depression more easily, in contrast, small banks usually fail first when crisis comes and spreads to the financial sector, and large banks tend to obtain government support more easily during hard times, when we look at the subprime crisis in America, it’s obvious that large financial institutions, are the first to save on the waiting list. As the US government decided the financial market rescue plan, after the fail of Lehman Brother, the US government spent more than 600 billion dollars to support America International Group, Bear Stearns Cos and Fannie Mae & Freddie Mac, then at October 8th, 2008, the Congress passed the proposal to contribute 700 billion dollars financial assistance to the market. What’s more, on
September 19th, the federal funds began to use the asset-backed commercial paper
money market mutual fund liquidity facility, it granted loans without recourse for commercial banks to induce them to purchase assets backed by commercial notes, in order to stop the shrink of the commercial notes market. Since there was no recourse, commercial banks were not responsible for the loss incurred from purchasing commercial notes. Within only one week after the announcement of AMLF, the loan granted under AMLF exceeded 152 billion dollars per day. All these financial instruments benefited the large financial institutions more than small ones, most of whom were not qualified to get the opportunities to participate the bailout or even went bankruptcy before the government began the rescue plan.
There is only one control variable that has positive relationship with the liquidity risk, return on 90 day treasury bills, which is the proxy of short term interest rate. The reason why high short term interest leads to high liquidity risk for banks is that, when short-term interest rate increases, investors are more glad to hold financial instrument with longer maturity, so do banks, they may tilt their asset structure to long term holdings, but the maturity of their liabilities doesn’t switch with assets, exacerbating the mismatch of maturity between assets and liabilities, which is the source for liquidity risk. At the same time there will be less funds on the market due to increase in short term interest rate, that is to say, funds turn out to be “more
27
expensive”, when banks lend and borrow in the interbank market, their cost of capital increases, but sometimes banks still needs to use interbank lending even though burdening the high cost of funds, which usually happens at the end of a month or quarter, result in financial strain for them.
4.2 Whether capital ratio has different influence on asset side originated liquidity risk and liability side originated liquidity risk.
From the OLS regression of LC ratio on capital ratio, we could draw a negative relationship between banks’ capital ratio and their liquidity risk, higher capital ratio doesn’t contribute to the mitigation of liquidity risk. After reaching this conclusion, the next question we need to answer is that whether capital’s influence on asset side originated liquidity risk is the same as that on liability side originated liquidity risk. To solve this issue, it’s crucial to find appropriate proxy for asset side originated liquidity risk and liability side originated risk. Looking into the calculation of LC ratio, the numerator is composed of liquid assets, semiliquid assets, illiquid assets, liquid liabilities, semiliquid liabilities and illiquid liabilities, weighting by their liquid converting ability, and the denominator is total assets. So as to explore the influence on asset side and liability side, we need to decompose the liquidity creation ratio into two parts: asset side LC ratio and liability side LC ratio, their sum is LC ratio.
𝐿𝐶𝑎𝑠𝑠𝑒𝑡 = 0.5 ∗ 𝑖𝑙𝑙𝑖𝑞𝑢𝑖𝑑 𝑎𝑠𝑠𝑒𝑡 + 0 ∗ 𝑠𝑒𝑚𝑖𝑙𝑖𝑞𝑢𝑖𝑑 𝑎𝑠𝑠𝑒𝑡 − 𝑜. 5 ∗ 𝑙𝑖𝑞𝑢𝑖𝑑 𝑎𝑠𝑠𝑒𝑡 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡 𝐿𝐶𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦 =0.5 ∗ 𝑙𝑖𝑞𝑢𝑖𝑑 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦 + 0 ∗ 𝑠𝑒𝑚𝑖𝑙𝑖𝑞𝑢𝑖𝑑 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦 − 0.5 ∗ 𝑖𝑙𝑙𝑖𝑞𝑢𝑖𝑑 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡 LC=𝐿𝐶𝑎𝑠𝑠𝑒𝑡+𝐿𝐶𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦
Then by regressing the asset side LC ratio and liability side LC ratio separately and compare the magnitude of 𝛽1 of the two regressions, the influence on asset side
and influence on liability side can be obtained. Table6 regression result of 𝐿𝐶𝑎𝑠𝑠𝑒𝑡 on capital ratio
28
Fixed-effects (within) regression Numbers of obs =29240 Group variable: bank Numbers of groups=731 R-sq: within=0.1200 obs per group: min=39 between=0.0552 avg=40.0 overall=0.0703 max=40 F (7,28345) = 552.4
Corr(u_i,Xb) =- 0.1474 Prob > F = 0.0000
Asset side LC Coef. Std.Err z p>︳z︳ 95% confidence interval
Capital ratio Change in GDP Return on S&P500 Return on T-bills Return on T-bonds Gross total assets Crisis dummy Constant .5327 .0095 56.13 0.000 .5141 .5513 -.0006 .0001 -9.87 0.000 -.0007 -.0004 -.0546 .0044 -12.47 0.000 -.0632 -.0460 .9632 .0600 16.06 0.000 .8456 1.0808 -.0878 .0086 -10.18 0.000 -.1047 -.0709 -2.00e-10 1.28e-10 -1.57 0.001 -4.51e-10 -5.05e-11 .0016 .0006 2.56 0.010 .0004 .0029 -.1826 .0010 -175.44 0.000 -.1847 -.1806 sigma_u sigma_e rho .07285888 .06490607
.70697857 (fraction of variance due to u_i)
F test that all u_i = 0 F(726,28345) = 92.76 Prob > F = 0.0000
Based on this table, every coefficient is significant, the impact of crisis on asset side LC ratio is significant as well, indicating that the break out of subprime crisis does affect the liquidity risk arising from the asset side of banks’ balance sheet, because the worsen of economic condition will post negative influence on the value of banks’
29
assets, and the dramatic fluctuation in value causes liquidity difficulty for banks, moreover, this difficulty will overspread from one bank to others.
The effect of other variables is not prominently different from the LC ratio regression, the positive increase in gross domestic productivity, high return on stock market, positive expectation on long-term interest rate and sufficient asset base all contribute to reduction of banks’ asset side originated liquidity risk, while the increase in short-term interest rate leads to generation of asset side liquidity risk. The stable increase in GDP creates beneficial outside environment for banks’ operation since the probability of borrower’s default is low and banks are capable of extending their asset base as more enterprise need funds for expansion under the steadily development economy. Almost every bank holds common stocks of other corporation as part of their assets, as the return to stock market, represented by S&P 500 index return, is favorable, the price of common stocks will suffer less fluctuation generally, so does long-term interest rate, when the market expects a mildly upward sloping interest rate curve in the long run, the value of banks’ asset tends to be stable as well, and the liquidity risk is less likely to rise from banks’ asset side of balance sheet. In addition, large asset base gives a bank higher ability to stand the fluctuation in asset value, so the coefficient of gross total asset is negative too. However, high short-term interest indicates high discount rate for banks’ assets such as marketable securities and available for sale securities, the rising of short term interest rate causes decline in value for these assets, so the relationship between return on 90 day treasury bills and 𝐿𝐶𝑎𝑠𝑠𝑒𝑡 is positive.
Table7 regression result of 𝐿𝐶𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦 on capital ratio
Fixed-effects (within) regression Numbers of obs =29240 Group variable: bank Numbers of groups=731 R-sq: within=0.0098 obs per group: min=39 between=0.0470 avg=40.0 overall=0.0299 max=40 F(7,28345) = 40.28
30
Corr(u_i,Xb) = -0.0276 (assumed) Prob > F = 0.0000
Liability side LC Coef. Std.Err t p>︳t︳ 95% confidence interval
Capital ratio Change in GDP Return on S&P 500 Return on T-bills Return on T-bonds Gross total assets Crisis dummy constant .3747 .0246 15.23 0.000 .3264 .4229 -.0002 .0001 -1.15 0.250 -.0005 -.0001 -.0106 .0114 -0.94 0.350 -.0329 .0116 .5583 .1555 3.59 0.000 .2535 .8632 -.0001 .0224 -0.00 0.996 -.0439 .0437 3.83e-10 3.31e-10 1.15 0.248 -2.26e-10 1.03e-09 .0067 .0017 4.01 0.000 .0034 .0100 -.0521 .0027 -19.31 0.000 -.0574 -.0468 sigma_u sigma_e rho .13118803 .12162052
.53779074 (fraction of variance due to u_i)
F test that all u_i = 0 F(726, 28345) = 46.40 Prob > F = 0.0000
Looking at this table we can find that the coefficient of crisis dummy variable is also significant, which means the subprime crisis from 2007 to 2009 post statistically significant influence on US banks’ liability side originated liquidity risk, as the bust weaken banks’ ability to meet their liabilities. The direction of impact to liability liquidity creation ratio of other variables is basically the same as their impact to asset liquidity creation ratio. Liquidity side originated liquidity risk, in most cases, comes into being because banks can’t meet their liabilities which are mainly made up of deposits. The situation will become worse if this kind of liquidity problem becomes contagion and causes bank run that can destroy the financial system quickly if no further action taken by the government. Generally speaking, steady increase in GDP, attractive return on stock market and optimistic market expectation of mid-term to
31
long-term interest rate as a whole, enhance investors’ confidence in the overall bank sector, as a result they will not claim their deposit together within short time period, reducing the possibility of bank’ liquidity problem coming from unable to meet depositors’ withdrawal demand in time. And mostly, banks’ main liabilities are the source of their assets, they obtain funds from depositors and translate them into loans, if a bank has abundant assets in hand means it has greater ability to grant more loans, sometimes it will generate excessive loans in order to seek more profits and fortune, at the same time more loans increases the mismatch between the maturity of its assets and liability, the bank bears more pressure from depositors’ claim against it, therefore more liability side originated liquidity risk when borrowers fail to repay these loans. However, the coefficients for these four variables are not significant at 5% significance level.
In this regression another factor that has statistically significant influence is return on 90-day T-bills, increase in return of 90 day T-bills leads to more liability side liquidity risk for banks, as mentioned before, the going up of short term interest rate often initiates financial strain for all market participates including depositors, as a result they may come to banks to withdraw their money at that time, however, banks also don’t have enough liquidity in hand, result in liability side originated liquidity risk.
Comparing the coefficients of capital ratio of these two regression, the absolute value of 𝛽1 in the 𝐿𝐶𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦 regression is smaller than that in the 𝐿𝐶𝑎𝑠𝑠𝑒𝑡
regression (0.3746 < 0.5327), and this difference between these two 𝛽1 turns out to
be statistically significant after conducting a F-test. In other words, capital ratio has large effects over asset side originated liquidity risk than over liability side originated liquidity risk, confirming Hypothesis2.
4.3 Whether banks with higher capital ratio before crisis perform better than others during the subprime crisis.
Since we already confirm that capital ratio has important influence over banks’ liquidity risk, both types included, it’s also necessary to research whether those
32
equipped with more pre-crisis capital would encounter more liquidity problem during the time period from 2007 to 2009. To address this question, what comes first is to identify a sub-sample of banks with higher pre-crisis capital ratio. In this step I use the capital ratio at the last quarter at 2006 to represent the pre-crisis capital, through comparing it of the 731 America banks, I draw out the top 5% percent of the 731 banks, which is 37 banks with highest 2006 q4 capital ratio, then put them into the OLS regression model (this time we need to use lag value of capital ratio as independent variable and drop the crisis dummy for the reason that the whole time interval is covered by the crisis period, inclusion of crisis dummy will cause collinearity) to get the result as following.
Table8: regression result of top 5% percent’s LC on lag capital ratio.
Fixed-effects (within) regression Numbers of obs =444 Group variable: bank Numbers of groups=37 R-sq: within=0.5365 obs per group: min=12 between=0.3018 avg=12.0 overall=0.3505 max=12 F (6,401) = 77.37
Corr(u_i,Xb) =- 0.3803 Prob > F = 0.0000
Asset side LC Coef. Std.Err z p>︳z︳ 95% confidence interval
Lag capital ratio Change in GDP Return on S&P500 Return on T-bills Return on T-bonds Gross total assets Constant .8566 .0433 19.80 0.000 .7716 .9416 .0010 .0007 1.43 0.155 -.0004 .0023 -.0302 .0362 -0.83 0.405 -.1013 .0410 -.2041 .6522 -0.31 0.754 -1.4864 1.0780 -.2014 .0905 -2.22 0.027 -.3793 -.0234 2.99e-09 2.21e-09 1.35 0.176 -1.35e-09 7.33e-09 -.2353 .0126 -18.61 0.000 -.2602 -.2105
33 sigma_u sigma_e rho .12340419 .06235061
.7966332 (fraction of variance due to u_i)
F test that all u_i = 0 F(36,401) =37.46 Prob > F = 0.0000
When the sample is narrowed into the 37 banks with highest pre-crisis capital ratio within the subprime crisis, we have only 444 observation left, and the influence of percentage change in GDP, return on S&P 500 index, return on 90-day treasury bills as well as gross total assets on liquidity creation ratio become statistically insignificant. Other two variables are significant at α=0.05. The coefficient of lag capital ratio is also positive, indicating that higher capital ratio leads to higher liquidity risk for banks, but the coefficient is smaller than that of the first regression of LC of all the 731 banks (0.8566 < 0.9074), which is consistent with hypothesis 2, indicating that after taking endogeneity into account and use lag value of capital ratio as independent variable, we can conclude that banks with higher pre-crisis capital ratio would experience less liquidity risk during subprime crisis, but the rule that higher capital ratio results in higher liquidity risk also applies to subprime crisis as 𝛽1 is still positive in this regression. Meanwhile, the coefficients of percentage
change in GDP, return on S&P 500 index, gross total assets and short-term government bills return are proved to be insignificant based on their p value. The reason behind this is that the effects of GDP, stock market and total assets work in the long run, time interval of this regression is only two years, not enough to detect significant influence of them. But long-term interest rate, whose proxy is return on 10-year government bond, still contributes to the reduction of liquidity risk for banks because at that time. In order to stimulate the recovery of economy, the Federal Funds lower its interest rate to nearly zero, although this may cause Keynes’s liquidity trap, making market participants insensitive to further change in short-term interest rate, the positive expectation of long-term interest rate would rebuild the confidence of market, which is beneficial for liquidity risk reduction. Besides, as the
34
𝑅2 increases to nearly 40 percent, the explanation power of our model improves,
meaning that capital ratio has more predominant impact on banks with higher pre-crisis capital ratio after taking endogeneity into consideration.
5 Conclusions
To sum up, based on the results, it’s clearly that only increasing capital ratio of banks is not enough to prevent liquidity shortfall. Since the increase in capital ratio is beneficial for reducing agency problem and prevent the management from taking excessive risk sometimes, why it does fail to reduce banks’ liquidity risk? To answer this question, we need to look into the composition of liquidity creation ratio and banks’ assets and liabilities structure. In our model, the LC ratio and its two parts are calculated as: LC= 0.5∗𝑖𝑙𝑙𝑖𝑞𝑢𝑖𝑑 𝑎𝑠𝑠𝑒𝑡+0∗𝑠𝑒𝑚𝑖𝑙𝑖𝑞𝑢𝑖𝑑 𝑎𝑠𝑠𝑒𝑡−𝑜.5∗𝑙𝑖𝑞𝑢𝑖𝑑 𝑎𝑠𝑠𝑡𝑒𝑠+0.5∗𝑙𝑖𝑞𝑢𝑖𝑑 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦+0∗𝑠𝑒𝑚𝑖𝑙𝑖𝑞𝑢𝑖𝑑 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦−0.5∗𝑖𝑙𝑙𝑖𝑞𝑢𝑖𝑑 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡 LC=𝐿𝐶𝑎𝑠𝑠𝑒𝑡+𝐿𝐶𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝐿𝐶𝑎𝑠𝑠𝑒𝑡 = 0.5 ∗ 𝑖𝑙𝑙𝑖𝑞𝑢𝑖𝑑 𝑎𝑠𝑠𝑒𝑡 + 0 ∗ 𝑠𝑒𝑚𝑖𝑙𝑖𝑞𝑢𝑖𝑑 𝑎𝑠𝑠𝑒𝑡 − 𝑜. 5 ∗ 𝑙𝑖𝑞𝑢𝑖𝑑 𝑎𝑠𝑠𝑒𝑡 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡 𝐿𝐶𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦 =0.5 ∗ 𝑙𝑖𝑞𝑢𝑖𝑑 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦 + 0 ∗ 𝑠𝑒𝑚𝑖𝑙𝑖𝑞𝑢𝑖𝑑 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦 − 0.5 ∗ 𝑖𝑙𝑙𝑖𝑞𝑢𝑖𝑑 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡
It’s obvious in the calculation of LC ratio, the most important part is liquid assets, illiquid assets, liquid liabilities and illiquid liabilities. Among the 28940 observations, some of them has negative asset side LC ratio but the percentage of negative liability side LC ratio is evidently smaller, so we can speculate that most banks hold abundant liquid assets in their balance sheet, which means cash or cash equivalents and short-term marketable assets account for large part of banks’ assets. Let’s take MINEOLA CMNTY BK SSB for an example, its three LC ratio from 2004 q1 to 2013 q4 is as following:
35
date Asset side LC Liability side LC LC
2004/3/31 0.248654397 0.063273772 0.31192817 2004/6/30 0.242679072 0.065191998 0.30787107 2004/9/30 -0.0680753 0.037912985 -0.030162313 2004/12/31 -0.06779536 0.040049746 -0.027745617 2005/3/31 -0.06536169 0.039692318 -0.025669371 2005/6/30 -0.06788106 0.040166765 -0.027714294 2005/9/30 -0.06392147 0.038883785 -0.025037686 2005/12/30 -0.05879721 0.04053783 -0.01825938 2006/3/31 -0.06425191 0.038500943 -0.025750963 2006/6/30 -0.06013218 0.041593831 -0.01853835 2006/9/29 -0.06279954 0.039606216 -0.023193325 2006/12/29 -0.06327462 0.042082384 -0.021192241 2007/3/30 -0.06848865 0.040868001 -0.027620648 2007/6/29 -0.07197286 0.037799598 -0.034173265 2007/9/28 -0.07530046 0.038565751 -0.036734705 2007/12/31 -0.07989045 0.043276625 -0.036613821 2008/3/31 -0.09495019 0.036304345 -0.05864585 2008/6/30 -0.08628999 0.035438851 -0.050851141 2008/9/30 -0.08323847 0.036406838 -0.046831629 2008/12/31 -0.08203635 0.038649025 -0.043387323 2009/3/31 -0.07778328 0.038602164 -0.039181112 2009/6/30 -0.06462056 0.039654847 -0.024965712 2009/9/30 -0.06023765 0.038495667 -0.021741988 2009/12/31 -0.05716822 0.039587274 -0.01758095 2010/3/31 -0.06338061 0.039057695 -0.024322912 2010/6/30 -0.06019271 0.038566432 -0.021626274 2010/9/30 -0.05602546 0.039122827 -0.016902637
36 2010/12/31 -0.0517756 0.041071307 -0.010704289 2011/3/31 -0.04845928 0.039978043 -0.00848124 2011/6/30 -0.04791896 0.039605242 -0.008313717 2011/9/30 -0.04806012 0.040431653 -0.007628472 2011/12/30 -0.04632797 0.044399244 -0.001928727 2012/3/30 -0.04209789 0.041598796 -0.000499096 2012/6/29 -0.04258529 0.040684268 -0.001901019 2012/9/28 -0.04011335 0.040688667 0.000575319 2012/12/31 -0.0421074 0.045184415 0.00307701 2013/3/28 -0.04147989 0.043387134 0.001907239 2013/6/28 -0.03857115 0.043306902 0.004735756 2013/9/30 -0.03987199 0.043140444 0.003268454 2013/12/31 -0.03586141 0.045896444 0.010035034
As we can see that a large percentage of its asset side liquidity creation ratio is negative since this bank holds most of its assets in liquid form. And when banks possess enormous liquid assets, when liquidity shortfall comes, such as depositors ask for their demand deposits, they are capable of meeting this requirement using liquid assets like cash or cash equivalents, additionally, even though the cash holding is not sufficient, marketable short-term assets can be translated into cash quickly without incurring large discount to their intrinsic value. Under these circumstances, capital’s role in solving liquidity problem becomes less useful. What’s more, in the light of deposit crowding-out effect, when banks hold more capital in balance sheet, the amount they could raise as deposit decreases, although this may mitigate the mismatch of maturity between assets and liabilities, in the mean time it also exposes banks to more fluctuation in value, because deposit’s market value is stable in most cases, while for capital, for example, common stock, its value will be closely dependent on the underlying company’s fundamentals, and capital is less liquid compared to banks’ liquid assets. If there is material negative information about the banks’ operation or management prevailing in the market, the value of common
37
stocks will decline sharply within short time period, and this shrink in value is reflected in banks’ balance sheet quickly. In a word, capital’s contribution to reduction of liquidity risk depends on banks’ balance sheet structure, if banks have a lot of liquidity in hand, the marginal utility of capital’s effects in deal with liquidity shortfall will decrease, furthermore, the unstable characteristics of capital’s value sometimes even causes liquidity problem, especially for asset side originated liquidity risk.
