- 0 -
Bachelor’s Thesis BA Track Finance
Bernardus Hendrik (Rick) Hogendorp 11824026 Supervisor: Mihnea Constantinescu
Year: 2019-2020
University of Amsterdam
The effect of bank capital on the monetary policy transmission mechanism in the United States
- 1 -
The effect of bank capital on the monetary policy transmission mechanism in the United States
Abstract
The purpose of this thesis is to research whether a relation between bank capital and the monetary transmission mechanism, represented by the interbank rate, is existent. A panel data analysis will be conducted which will test the effects of determinant of bank capital measured at nine US commercial banks and the interbank rate deployed by the US FED. Those determinants of bank capital are based on the existing literature and it will build further on the solid foundation of the monetary policy making sector and the foundation of optimal capital structure within the banking sector. The conducted research shows that there is no conclusive answer whether there is a direct link between a bank’s capital structure and the deployed monetary policy stance, because the role of bank capital is too connected with the ever changing economic environment, however it can be said that bank capital and bank liquidity contribute to the transmission mechanism since it hinders or enhances the policy stance depending on the maintained levels of capital.
Statement of Originality
This document is written by Student Bernardus Hogendorp who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
- 2 -
Table of contents
0. Abstract and statement of originality………p. 1 1. Table of contents………pp. 2 2. Introduction………pp. 3-4 3. Theoretical framework………pp. 4-6
3.1 Bank Capital(ization)………pp. 4-5 2.2.1 Monetary policy stance………..pp. 5-6 2.2.2 Interbank rates as proxy for the monetary policy stance………...p. 6 2.2.3 Interplay between Bank Capital and interbank market rates activity………p. 6 4. Data, Methodology, and empirical model………pp. 7-12
4.1 Data and measures………pp. 7-8
4.2 Method and presumptive empirical model………...pp. 8-9 4.3 Econometric specification………...pp. 9-12 4.4 Estimating the effect on the overall transmission mechanism……….p. 12
4.5 Testable Hypothesis……….p. 12
5. Results……….pp. 13-16
5.1 Baseline regression………pp. 13-14
5.2 Semi-extensive regression model ……….p. 14 5.3 Extensive regression model………..p. 15 5.4 Implications for the monetary policy transmission mechanism………….p. 16 6. Discussion and conclusion………pp. 17-18 7. References………pp. 19-20 8. Appendix………..pp. 21-27
- 3 -
1. Introduction
During the Global financial crisis (2007-2009), many banks faced liquidity risk in face of a pressing and acute increase in the demand for cash, since the faith in the financial system and the trust in the on average health of financial institutions saw a rapid decline. The financial urgency in which the bank had gotten into was partially caused by bank runs which initiated the disposal of illiquid loans (P. Strahan (2012) pp. 1-2). Part of the main resolves to prevent and check these instabilities within the financial market were the so called liquidity boosts from lenders of last resort (i.e. central banks), but also mainly reserve requirements.
Another crucial factor in the well-functioning of the financial market is the interbank rate, a central bank set on its’ loans, which banks can use to reduce liquidity risk during crises caused by a mismatched between short term liabilities and long term funding projects and the risk of money withdrawal from depositors (X.Freixas A. Martin & D. Skeie (2010) pp. 2-7).
The goal of this paper is to determine whether bank capital has a significant influence/effect on the transmission of the monetary policy stance deployed by the policy makers. The importance of this kind of research lays within the everlasting question on how a monetary policy can be implemented while there are many external factors (such as in this case bank capital) which should be taken into account. The amount of capitalization of commercial banks for example are certainly partly correlated to financial downturns and upturns, which possibly needs to be mitigated and reacted to by a change in the monetary transmission mechanism from out the central bank (N. Martynova (2015)) If the size of this effect can be determined, monetary policy makers can make capital reserve requirements more or less stringent in order to seek a balanced approach to require banks to have sufficient reserves in order to make the transmission mechanism run more smoothly (N. Martynova (2015) ).
The means by which the effect of bank capital on the overall mechanism used by central banks will be a panel regression of different in which different determinants which present the capitalization of different commercial banks within the United States to a
dependent variable which serves as a proxy for the monetary policy stance. Subsequently the effect of bank capital will be extended to the overall effect of the transmission mechanism as a whole. Difficulties encountered within this research are aligned with the estimation of the actual effect on the transmission mechanism since the effect on the chosen proxy can result in an omitted variable bias because the effect on the proxy isn’t self-contained. In order for the
- 4 -
estimation to determine the effect on the transmission mechanism to be relevant and significant, current literature on these matters will be critically examined.
2. Theoretical Framework
This Chapter discusses the theoretical background and a extensive literature review of selected key factors which come at play within the core of this research. Firstly, definitions and possible measures for bank capital will be presented and these measured will be
distinguished by being directly or indirectly related to a banks leverage. Existing research concerning bank capital and bank liquidity are presented consequently. Next, a
comprehensive view on the monetary regulatory framework is introduced in which a
supportive measure is described to serve as proxy for the actual monetary policy transmission mechanism. In the third and final part of this chapter, we describe the juxtaposition of bank capital and the monetary policy transmission mechanism applied by the central bank of a country.
2.1: Bank Capital(ization)
Castiglionesi et al. 2014 shows that banks with limited power to deal with the effects and counter liquidity risk, are more keen to hold more excess capital and have more leverage than banks which have more coinsurance opportunities. ‘
The amount of Capital has an direct influence on the financial state of the market as a whole and it should be an important consideration when choosing a stance within the
monetary policy spectrum (Chen, Minghua 2017). In order to find the most accurate
depiction of a bank’s capital structure the most important financial determinants of a bank’s capital/liquidity have to be found and elaborated on.
The first factor that will come up in mind is the basic bank leverage, since that gives us the most basic idea of the capital structure maintained by bank management. Leverage can be determined by the ratio of the amount of core capital over the amount of total assets maintained in the company.
To advance there will be looked at literature which will support measures which are chosen to be used to conduct the main analysis for this thesis; in the research paper Gropp & Heider (2009) the main determinants of a bank capital structure lay within the bank leverage data such as Tier 1 capital over total assets maintained and the appliance of capital funds over
- 5 -
assets. Furthermore (Estrella, Park, and Peristiani 2000) and (Berger, Herring, and Szego (1995)) advocate for the usage of different capital and liquidity ratios to determine the amount of effective capitalization within a bank which could serve as a buffer in a financial downturn. This thesis analyses the effect of bank capital on the interbank rates stated by the Fed during the great recession, thus the capital and liquidity ratios state by the two aforementioned papers would serve as a good approximation of bank capital. In order to give a comprehensive image in regard to the capital structure the following capital and/or liquidity ratios will be used:
The interbank ratio, the interbank ratio gives us a comprehensive look at the position of the bank within the interbank environment (i.e. net borrower or net lender). This is
important because the capital position of a bank is determined by its’ relative position within the financial interbank system (Bunda & Desquilbet 2003).
To advance the Capital Adequacy ratio (CAR), the CAR is a capital ratio which serves as an overview of a bank’s ability to take a financial blow based on the relevant cash reserves it maintained relative to the capital requirement stated in BASEL III. The CAR therefore serves as an important financial gauge of the capital a bank has maintained overtime (Rime (2001)).
Furthermore the capital funds over total assets ratio and the net loans over total assets ratio. These ratio do explicitly state the capital structure in which a bank operates, it gives us the amount of loans and the amount of capital funds relative to the total assets managed. It therefore serves as a approximation of the leveraged position of a bank within it’s own system (Raharjo, Hakim, Manurung, & Maulana (2014)).
Finally an important exogenous factor to determine a bank’s capital position is the change in loans relative to the previous year, which could result in another capital structure (gamborta and shin (2018)).
2.2.1: Monetary policy stance
A way to classify the monetary stance in the different regions of the world is the interbank market activity these activities can be measured by the repo selling and on the counterpart lending channels Ashcraft, Adam B (2001)
The above mentioned interbank market can’t prevent liquidity risk caused by asymmetrical movement of the commercial financial market. This undiversifiable liquidity risk that cannot be mitigated by policies taken within the interbank market, however when the (Afonso, Kovner and Schoar (2011) pp. 6-12).
- 6 -
Ashcraft &Adam (2001) does show that federal funds become less sensitive to capitalization of financial institutions as it grows further into infinity.
2.2.2 Interbank rates as proxy for the monetary policy stance
When operating in an risk-intolerant environment (i.e. banking) the role of interbank market contracts, which consist of short term commercial bank deposits and short term lending in the form of repos, is the most dominant controlling factor preventing short term liquidity risk (Allen and Gale 2000). Deviation in interbank rates stimulate banks in up- and downward sloping economic prospects (Bonam et al 2018). The 3m interbank rate as lending channel for countries within monetarily integrated zones does contain enough explanatory power to lend itself as a nice variable that could serve as a proxy for the monetary stance at the moment an particular interbank rate is in force (Sunirand (2003)).
2.3 Interplay between Bank Capital, the interbank rate and the interbank monetary transmission mechanism
Observations do make a point that interbank lenders are more likely to be better capitalized and have significantly better credit ratings than interbank debtors (Angelini et al. 2011).
Baker, Malcolm, and Wurgler (2013) state that strict capital(ization) requirement tend to increase the cost of borrowing at the interbank level which means that interbank rates tighten during times in which capital requirements need to be more stringent.
Dimsdale (1994) states that within the interbank market the transmission mechanism in regard to the release of the policy stance is hindered or stimulated by external factors in the macroeconomy in addition to internal decision making based on perspective and self-owned information and monetary ideology. When a change in the interbank rate occurs a change in the deployed monetary policy stance in imminently related to this happening. (Fung & Yuan (2013)). 1
1 Final remark on the theoretical framework: In the final analysis conducted for this paper, the influence on the whole transmission mechanism will be tested as regard to the effect bank capital has on the proxy which serves as the available image in the relevant time period.
- 7 -
3. Data and empirical model
Following sections of this paper will explain in detail the sample data set used and the process of creating a feasible data set which will be used to conduct the main analysis. To start with the clarification of the different variables previously described in the literature section process, for example how we catch the broad capital factor in different smaller variables. Proceeding the foundation of the core model will be presented by the method used to catch the effect this paper is focussed on. Finally, the final econometric specification of the core model, and additionally the assumptions which are made in this thesis.
3.1 :Data and Measures
Dependent variable
To support the claim that bank capital has an effect on the monetary policy transmission mechanism a solid measure which reflects the monetary policy the US FED deploys in order to regulate and watch over bank liquidity. The measure to be used in this paper will be the 3M interbank rate which will serve as proxy for the deployed monetary policy (see section 2.2.2). This report extracts this measure from the FRED database from the St. Louis FED.
Independent variables
In order to Measure the effect of Bank capital on the stated interbank rates there have to exist thorough measures which determine the amount of capital a bank juxtaposed to a bank’s liquidity and leveraged position. For this thesis different variables proclaiming parts of bank capital/capital structure are used. These variables are based on section 2.1 on
bank(capitalization). All these measures are extracted from the BankFocus database. In the table below all variable names with a description and the corresponding variable code can be found.
TABLE 1: VARIABLES TO BE USED TO DETERMINE THE MODEL
Variable name variable description variable symbol
bank_name Name of the US commercial Bank Bank_name
Bank_id Number assigned to each individual bank Bank_id
year Year of the US commercial Year (2006-2011) Year
t time variable assigned to year (t=1, … , 6) t
Net loans/Total assets The net loans in debt divided by the totality of assets loans_assets interbank ratio the ratio formed by "due from bank" over "due by bank" Ibratio capital funds/assets The amount of capital funding reserves over the total assets Capfund_assets Capital Adequacy Ratio a measures to determine the solvability of a company CAR
loans The amount of Loans outstanding loans_assets
Δ loans The chang in loans in regard to the previous year Δ loans
Δln(loans) a numeric adaptation of the delta loans Δln(loans)
GDP growth The growth of the GDP in a specific year GDPgrowth
Total assets All assets a bank has to mangage Totalassets
Tier 1 capital Tier 1 capital Tier1capital
Bank Leverage = tier 1 captial/total assets Bank leverage is determined by tier 1 capital over total assets Bankleverage Bank Size =ln(total assets) a numerica expression of bank size with the fundation in the assets Banksize liquid assets/ total dep and bor. Liquid asset divided by total deposit a Liquid_depbor
- 8 -
Additionally you can find the descriptive statistics in table 2 below.
TABLE 2: DESCRIPTIVE STATISTICS FOR THE MOST IMPORTANT VARIABLES USED IN THE ANALYSIS
3.2 :Method and presumptive empirical model
In order to estimate the effect of bank capital on the interbank rate, which will serve as the market proxy for the monetary policy stance at the same moment a regression has to be run with different determinants of bank capital which reflect the state of the capital position of a bank within the United States. This in order to create a comprehensive image of the capital position and its’ effect on the monetary transmission mechanism.
For the dependent variable which reflects the monetary stance at that specific time the interbank rate stated by the FED is taken. The period in which the interbank rate is being observed has to reflect both a period in which the general economy was flourishing and a period in which an economic downturn was ubiquitous within the US economy. This is done in order to conduct an analysis of changing interbank rates which are the result of a changing economic environment in addition to the monetary ideology the central bank had deployed in that specific period. In order to identify these two periods we look at the state of the US business cycle stated by the National Bureau of Economic Research (NBER) A recession cycle in this thesis is therefore identified as the period between December 2007 and June 2009. According to the Worldbank database on GDP growth of the US economy between December 2007 and June 2009 with growth rates between -0,14% and -2,54%. For the economic upturn a period of with positive GDP growth figures have been chosen, namely the period before the recession commenced (2006-2007) and the period direct behind the great recession (2010-2011).
Since the time period is established a baseline model has to be created in which we interbankr~e 54 2.426248 2.191459 .30333 5.26833 Banksize 54 19.91262 1.189032 17.49388 21.54232 Bankleverage 54 .0685986 .0128792 .0407866 .1010655 GDPgrowth 54 1.361667 2.105326 -2.537 3.513 Δlnloans 54 9.932735 13.28279 -18.53178 19.54174 CAR 54 14.48241 2.664007 10.7 21.99 Capfunds_a~s 54 17.53889 12.49967 7.89 63.8 IBratio 54 454.9148 1020.856 1.35 6730.49 Variable Obs Mean Std. Dev. Min Max
- 9 -
test interbank rates against the most important and fundamental variables which reflects capital. The baseline regression will therefore consist of the measure for bank leverage and the ratio of capital funds over capital assets which will be ran against the interbank rates in the same year. The reasoning behind the baseline stems from the notion that bank leverage is an important factor to consider when determining the capital structure a bank maintains, is an important gauge in order to determine a bank’s financial stability in both financial healthy times and times in which the economy takes a financial downturn.(Estrella, Park, and Peristiani 2000). Additionally the paper states with more support from (Dimsdale 1994) that the capfunds over assets ratio helps to estimate the amount of usable capital within the firm. When both variables are taken into account the following regression model based on existing literature follows:
𝐼𝑛𝑡𝑒𝑟𝑏𝑎𝑛𝑘𝑟𝑎𝑡𝑒
̂
𝑈𝑆𝐴𝑖𝑡= αit + 1(Bank leverage)
it+ 2 (
cap.fundsassets
)
it+ 𝑢𝑖 + εit (1)In the next section the usage of panel data will be explained together with an elaboration on the analysis method to be used. Additionally some extra variables will be added to form a model with additional explanatory power.
3.3 Econometric specification and estimation procedure
The main gripe with panel data is that a vast number of observations that comes within the newly created dataset can be better comprehensively analysed within a multidimensional environment. Additionally the use of Fixed Effects allows the model to be less sensitive to omitted variable bias (OVB). the usage of different analysis tool can measure unobserved time specific and individual bank specific effects which can be both excluded and included in the analysis through. The OVB within these id and time specific variables are variables that could not be directly observed in the
error term.
Therefore in order to look at the fixed effect the method preferred is a panel data analysis in which estimators of bank capital are ran in a regression to the US interbank rates
- 10 -
over different banks within the same time period, The following regression expression will serve as an example:
𝑌𝑖𝑡 = 𝛼𝑖 + 𝛽1𝑋1,𝑖𝑡+ 𝛽2𝑋2𝑖𝑡+ 𝛽3𝑋3𝑖𝑡+ 𝛽4𝑋4𝑖𝑡+ 𝛽5𝑋5𝑖𝑡+ 𝛽6𝑍1𝑖𝑡+ 𝛽7𝑍2𝑖𝑡+ 𝐸𝑡; 𝑖 = 𝐵𝑎𝑛𝑘_𝑖𝑑 = 1, … , 𝑇
In which Y is the estimate of the interbank rate in respect to the estimators of bank capital, X1 till X5, and Z1,2 which serve as the control measures.
To advance on the presumptive baseline model (regression 1), the model has to be elaborated with variables which could be possible determinants of the US interbank rate and controls which will have presumably effect on the interbank rates.
To start on the control/exogenous variables: Habib, Mileva and Stracca (2016) state that the interbank environment is strongly influenced by the economic/financial state of the world or country economy, therefore the first control to be introduced to the model is GDP growth measured in the percentage growth relative to the previous year. Following the same reasoning the dummy recession can be introduced to the model which will be 1 when in recession and 0 when there is constant growth (NBER economic business cycles used). Additionally the model wants to control for bank size, since the size will have an effect on the transmission mechanism within the interbank environment (De Haan and Poghosyan 2011). The measure for bank size is derived from the same paper, namely the size is determined by the natural log of the total assets under management (Ln(assets)). When these variable are added to the baseline model, the following regression is created:
𝐼𝑛𝑡𝑒𝑟𝑏𝑎𝑛𝑘𝑟𝑎𝑡𝑒̂ 𝑈𝑆𝐴𝑖𝑡 = αit + 1(Bank leverage)it+ 2 (cap.funds assets ) it + 𝛽3(𝐺𝐷𝑃 𝑔𝑟𝑜𝑤𝑡ℎ)𝑖𝑡+ 𝛽4((𝐵𝑎𝑛𝑘 𝑠𝑖𝑧𝑒)𝑖𝑡+ 𝛽5(𝑟𝑒𝑐𝑒𝑠𝑠𝑖𝑜𝑛)𝑖𝑡+ 𝑢𝑖 + εit (2)
The next step is to find additional variables, which can be introduced to increase a more comprehensive image of the bank’s capital structure. The first variable to be introduced is the liquid assets over dep. Bor ratio: This variable shows the base level of the amount of liquid assets a bank possesses, which it can use to determine a bank’s relative financial position (Freixas & Jorge (2015)).
To complement that variable the Capital Adequacy Ratio is introduced, since this measure is supposed to give an more extensive image of the size of the capital reserves of a company, since it is a direct measure of a bank’s response to the capital requirements stated in
- 11 -
the BASEL III regulatory framework.
Finally the Gambacorta & S hin paper (2016) states that a change in loans could be an important determinant of the size of changes in bank capital(ization) and this paper would take the natural log of the change in bank loans in order in order to create a more sensible measures consistent with the other variables.
The following regression model in which the previous determinants of bank capital are incorporated (this paper will call this the extensive regression model) will serve as be an estimator for interbank rates in the US:
𝐼𝑛𝑡𝑒𝑟𝑏𝑎𝑛𝑘𝑟𝑎𝑡𝑒̂ 𝑈𝑆𝐴𝑖𝑡 = αit + 1(bank leverage)it+ 2 (cap.funds assets ) it +
3 (Δlnchange in loans) + 5(Liquidass_depbor) + 6 (𝐶𝐴𝑅)it + 𝛽3(𝐺𝐷𝑃 𝑔𝑟𝑜𝑤𝑡ℎ)𝑖𝑡+ 𝛽4((𝐵𝑎𝑛𝑘 𝑠𝑖𝑧𝑒)𝑖𝑡+ 𝛽5(𝑟𝑒𝑐𝑒𝑠𝑠𝑖𝑜𝑛)𝑖𝑡+ 𝑢𝑖 + εit (3)
In the following section the econometric analysis method will be extensively elaborated: The purpose of the regression model which test the coefficients of the bank capital variables within the panel data while keeping/assuming any other effects remain constant will result in the estimation of the isolated effects of each factor related to a bank’s capital(ization)
In order to analyse the effect of the individual variables the id fixed effects the dataset need to be eliminated in order to look at the effect of the interbank rate in the chosen time period. The banks used in the analysis are the nine largest US commercial banks at the time period measured the approach to take in this paper is a time-fixed effect analysis in which the bankid’s are clustered since a comprehensive image with respect to the effect of the individual variables is the only required effect to be analysed (Eller et al. 2006). To advance Lu (2014) proposes the usage of the robust standard errors in order to correct for issues with respect to heteroskedasticity so within this analysis robust standard errors are used. Thereby is the robustness of the extensive regression positively affected by the three controls which were added. The implied Robustness results in more conclusive and comprehensive measured coefficients (Lu 2014). To Proceed the used econometric procedure which is used will be elaborated: Holland and Vieira (2011) do use the Cross-sectional time-series FGLS regression analysis tool to estimate the effect of exchange volatility on the prospects of economic growth with a panel with similar properties to the panel used in this paper.
This thesis will use the same estimation procedure since the core of this research lays within the creation of a comprehensive estimation of the coefficients which vary over time due to exogenous influences of the economy as a whole. The procedure focusses on the
- 12 -
estimation of the actual size of all the coefficients which affect the dependent variable. All regressions are being tested with the same procedure in mind and a F-test will be conducted to look at the additional power of each successive model.
3.4 Estimating the effect on the overall transmission mechanism
In this section the effect estimation on the overall monetary transmission mechanism will be deepened, since this is one of the most important issues to be dealt with in order to answer the research question. Firstly we want to underline that interbank rates are in essence the proxy and determinant of the policy stance deployed by the policy makers within this thesis.
In order to estimate the effect on the transmission mechanism which is highly related to the interbank environment and the interbank rate, a estimation of the effect size of bank capital on the interbank rate has to be estimated by looking at the η2 measure within the ANOVA analysis of the full regression model as seen in (Bakeman & Roger (2005)). This in order to look at the relative effect size of the estimators of bank capital relative to the change in the interbank rate and relative to the effect size of the state of the economy. When we found the relative effect size of bank capital on the interbank rates in each successive year we have found data on, we can see whether maintaining a certain capital level hinders or improves the transmission of the monetary policy stance. This effect can be noticed when bank capital has a significant effect on the increase or decrease of the stated interbank (Bernanke & Mihov (1998)) and (Fung & Yuan (2003))
3.5 :Hypothesis
In this short section the main hypothesis will be stated in order to test it in the following sections and to formulate whether there is conclusive evidence whether bank capital affects the monetary policy transmission mechanism by adjustments to the stated interbank rate Based on existing literature bank capital would affect the monetary policy rate based on the notion that banks, which are sensitive to large worldwide shocks to the economy, does require a capital buffer to protect themselves from bank runs caused by credit risk together with mass unrest. Monetary policy makers at their turn try to stabilise the economy by lowering the interbank interest obligation in order to recapitalise banks to protect them from topple. The hypothesis we can formulate that the coefficients of determinants of bank capital β_hat ≠ 0, juxtaposed to a H0 of β_hat = 0.
- 13 -
4 :Results
In this section the regression models (baseline, semi-extensive and extensive) which are introduced in the previous section are conducted and all the relevant statistical measures are discussed. An important note beforehand that for clarity reasons the intercept term will not be included in the final discussion of measures because these are nugatory and unimportant for this thesis.
In order to test for the effect of bank capital on the interbank stated by the FED, a panel containing the largest US banks in regard to the total assets these entities manage, has been created. Different determinants of bank capital (see previous section) are included for each of those banks in the period ranging from 2006 to 2011, for a total number of
observations of 54 per determinant, times the number of variables which totals 486 observations.
4.1 :Baseline regression (1)
FIGURE 4: REGRESSION OF THE INTERBANK RATE ON BANK LEVERAGE AND CAPITAL FUNDS OVER TOTAL ASSETS (SEE FIGURE A.10 FOR A MORE EXTENSIVE LOOK INTO THE MEASURES)
The results of the baseline regression model is in total summarized in table 4. As mentioned in section 3.3 robust errors are used to prevent heteroskedastic issues to come
up.
As seen in the table the variables of bank leverage and cap_fund over assets are both highly significant with respective coefficients of -156.152 and 0.2254. Recall that bank leverage is in a linear log relation ship which means that the 156.152 coefficient can be transformed to a -1,562 change in the interbank rate when the leverage increases with a 100%. R2 for the baseline model is equal to 0.1489, which means that little of the variance within this model is still explained by the determinants of bank capital. Additionally corr(u_i, X) = 0 cannot be
rho .78173426 (fraction of variance due to u_i)
sigma_e 1.7227754 sigma_u 3.2603592 _cons 9.184991 2.066335 4.45 0.000 5.01783 13.35215 Capfunds_assets .2253883 .0554583 4.06 0.000 .1135461 .3372306 Bank_lev -156.152 28.06778 -5.56 0.000 -212.7561 -99.54792 Interbankrate Coef. Std. Err. t P>|t| [95% Conf. Interval]
- 14 -
rejected by the model which is a positive sign. Finally the wald chi squared measure is equal to 8.92 ; p=0,00115 which means that the null hypothesis for this statistical measure can be rejected and that both factors contribute to the statistical significance. In regard of the baseline regression it is statistically justifiable to reject the null hypothesis, which result in the notion that bank capital could have an effect on the transmission mechanism of monetary policy.
4.2 : semi-extensive regression model (2)
FIGURE 5:PANEL DATA REGRESSION MODEL OF THE SEMI-EXTENSIVE MODEL (SEE FIGURE A.8 FOR A MORE EXTENSIVE IMAGE)
When the second regression is ran in STATA the first measure to look at is the Wald chi2 test in order to identify whether the new variables added will increase the explanatory power of the model χ2
(9) =24.12 p = 0.000, which result in the fact that no variable should be removed for the sake of the explanatory power of the model. As seen in the table above the effect size of both determinants of bank capital are suffering a decrease in effect size, the effect of capfunds assets is not even significant anymore and the effect size of bank leverage is severely decreased in such a way that both confidence intervals don’t overlap each other in a significant way. The effect size of exogenous variables such as bank size and GDPgrowth do contribute a significant amount on the total variance of interbank rates.
The effect of bank size is small though consistent and highly significant with a
coefficient of -3.822 which due to the linear log relation of represents a change of 0,03822 on the interbank rate. A change in the GDP growth percentage has a large and highly significant effect on the interbank rates; a 1% increase in the GDP growth figures within a specific year result in a 0.4127 increase in the interbank rate.
The explanatory power measured by the R2 has a large increase with respect to the base line model, the R2 is equal to 0.4658 though the within measure is still quite low.
rho .91391315 (fraction of variance due to u_i) sigma_e 1.2588862 sigma_u 4.1017629 _cons 83.16071 20.24315 4.11 0.000 42.24777 124.0736 GDPgrowth .4126852 .0952742 4.33 0.000 .2201289 .6052415 recessionNBER -.1898473 .4049454 -0.47 0.642 -1.008272 .6285778 Bank_size -3.821884 1.007754 -3.79 0.000 -5.858631 -1.785137 Capfunds_assets -.0481449 .0695978 -0.69 0.493 -.1888072 .0925174 Bank_lev -62.46471 27.18323 -2.30 0.027 -117.4041 -7.525353 Interbankrate Coef. Std. Err. t P>|t| [95% Conf. Interval]
- 15 -
4.3 :Extensive regression model (3)
FIGURE 6:CROSS-SECTIONAL TIME-SERIES FGLS REGRESSION (FOR A MORE COMPREHENSIVE FIGURE SEE A.9 AND FIGURE A.11 FOR THE BETWEEN-SUBJECT EFFECTS)
In this model the most worth mentioning variables which serve as the determinants of the interbank ratio are: Bank_size, Bank_lev, GDP growth, and the CAR. This last section serves to explain these effects, and their relation to the variables which do not result in significant effects.
A little sideway first to mention the R2 measure which has now increased to a total of 0.7509 and a Wald χ2 of 216.21 with a p value of 0.000 which explains that no variable could be removed in order to improve the explanatory power.
Now back to the model: to start with the two variables from the baseline regression model, namely bank leverage and capfunds_assets. The bank leverage variable is still highly significant in contrast to the regression 1 with a effect of 0,71 on interbank rates relative to a change in 100% in bank leverage, which is fairly large. The effect of capfunds_assets meanwhile decreased and isn’t significant anymore.
Furthermore the exogenous variables of Bank size and GDP growth still remain
highly significant as seen in the table with the respective coefficients of -0.7052 and 0.34. A most noteworthy new variable is the CAR, with a coefficient of -0.3338 (p =0.016) which measures the amount of capital over the general reserve requirement stated in BASEL III.
The other measures for the amount of capital do not give a statistically significant effect on the interbank rate even as the dummy variable for the recession.
In the next section the overall impact on the transmission mechanism is explained using the results derived from this final regression model.
_cons 25.93797 4.779467 5.43 0.000 16.57039 35.30556 Nloans_assets .0184749 .0210879 0.88 0.381 -.0228566 .0598064 Ibratio -.0000345 .0002413 -0.14 0.886 -.0005075 .0004386 Capfunds_assets -.0374865 .0220216 -1.70 0.089 -.0806481 .005675 CAR -.3337925 .13909 -2.40 0.016 -.6064039 -.0611812 Δlnloans -.0010325 .0119722 -0.09 0.931 -.0244975 .0224326 GDPgrowth .3413427 .0780413 4.37 0.000 .1883846 .4943008 Bank_lev -72.38817 27.7827 -2.61 0.009 -126.8413 -17.93508 Bank_size -.7052132 .1912582 -3.69 0.000 -1.080072 -.330354 recessionNBER -.0635462 .3549513 -0.18 0.858 -.759238 .6321455 Interbankrate Coef. Std. Err. z P>|z| [95% Conf. Interval]
- 16 -
4.4 :Implications for the monetary policy transmission mechanism and final remarks
The impact that bank capital has on the transmission mechanism can be estimated looking at the relative contribution of individual capital determinants to the model explaining the US
interbank rate which is the proxy for the monetary policy stance within the US. In order to do that we look at the η2 measure of each individual capital determinant, while
comparing them to the macro-economic factor of GDP growth, in order to measure their relative effect.
To start with bank leverage: Bank leverage has a η2 of 0,06 which means that the overall effect of leverage is relatively low compared to the η2 of GDP growth of 0,62 and following the rule of thumb the effect is very small compared to a large effect of GDP growth.
To continue the most important capital determinant, namely the CAR, has a partial η2 of 0,295 which means that the CAR is surely an important factor to take into account when estimating the overall effect of capital adequacy and liquidity on the monetary transmission mechanism. The effect of the CAR can be marked as large which is can be derived as well from the measured coefficient in the third regression.
There can be said that a specific but fairly important measure of capital (CAR)
mitigates the transmission mechanism by having a hindering effect on the interbank rate when the capital adequacy measure increases since the interbank decreases (increases) when you take into account the contributing effect a upward sloping CAR (downward sloping CAR).
To iterate, the effect of macro-economic factors (such as our GDP growth variable) do contribute a lot in changing the interbank rate and in that way the transmission mechanism deployed by the central bank. In addition it is important to take into account bank capital which mitigates the effect by hindering/stimulating the transmission mechanism at certain levels of capital adequacy/liquidity and partly leverage.2
2 Note: When the interbank rate is estimated by filling in the regression model with our data, you will see that bank capital surely has an impact on the current monetary stance and so a intermediating effect on the transmission mechanism of the monetary policy stance is created.
- 17 -
5 :Discussion and conclusion
The main goal of this thesis was to determine whether bank capital has a significant effect on the monetary policy deployed by policy makers at the US FED. In order to estimate the effect size of bank capital (structure) on the interbank rates which served as proxy for the policy stance of the central bank, a panel data analysis was conducted. The exact estimates found during this research can be found in the result section and the second half of the appendix. To summarize the results the most important determinants will be looked at and subsequently the practical implications of those findings will be looked through.
To start the source of the following claims stems from the extensive regression model (3), since that model has proven itself to be the most informative with the highest explanatory power. The third model shows us that the interbank rate deployed in the US is highly
influenced by the state of the economy, seen in the GDP growth projection in the specific years. (The Recession dummy does not show an effect presumably because the time
dimension used to conduct the analysis). Additionally the bank size of the banks used in the analysis shows us that the total effect is relatively small but still significant presumably because the interbank transmission mechanism is also determined by the size of the assets under management (Ashcraft, Adam B (2001), so the bank size determinant does not
necessarily contribute to a large part of the variance and does not reflect any necessity of bank size in the composition of the interbank rate and thus the deployed monetary policy stance.
Now the coefficients of the determinants of bank capital will be interpreted as being so. The first significant determinant of bank capital is the leverage variable. Bank leverage serves as a nice measure of the amount of capital reserve in respect to the total assets under management. The model suggests that an increase as regard to the leverage level will lead to a decrease in the interbank rate, which could be explained that an increase in bank leverage would lead to more uncertainty as regard to credit risk of lenders and borrowers and the behaviour of bank clients. The final significant variable within the extensive model is the CAR, the CAR reflects a bank’s capital position as regard to the regulatory framework internationally set on capital requirements. When the CAR increases it will be become more easy for a bank to fulfil its’ financial obligations, therefore it seems logical that the interbank rate would increase when a bank is in financial more healthy position. With that same logic ergotism it can be said that bank capital has a link to the enhancement or hinderance of the transmission mechanism, since the interbank rate directly affects the deployed transmission
- 18 -
of monetary policy.
The biggest but which contradicts or at least appease the positive relation between in general more healthy financial position as regard to a bank its’ capital and the stated interbank rate is the notion that it is certainly possible to fall for the simultaneous causality bias in this case, since bank capital is very sensitive to exogenous and external factors from the economy as a whole, since bank capital is correlated to the state of the economy and the state of the individual client within that economic system. This gets supported by the fact that the effect of the other determinants of bank capital such as the interbank ratio and the capital
funds/assets ratio do not show us a statistically significant effect, while the capital funds/assets ratio was highly significant in a model without the exogenous factors.
Therefore there can’t be given a conclusive answer as regard to whether bank capital serves as an important factor that enhances or hinder the transmission mechanism Though it can be said that bank capital possibly serves as an intermediate factor which serves as a link between the overall state of the economy and the interbank transmission mechanism.
- 19 -
References
Ahmad, R., Ariff, M., and Skully, M.J. (2009). The Determinants of Bank Capital Ratios in a Developing Economy. Asia-Pacific Finan Markets 15:255-272.
Ashcraft, Adam B (2001), “New evidence on the lending channel”, FRB of New York Staff Report No. 136
Aiyar, Shekhar, Charles Calomiris, John Hooley, Yevgeniya Korniyenko and Tomasz
Wiedalek (2014a), The international transmission of bank capital requirements: evidence from the United Kingdom, Bank of England Working Paper No.497
Bakeman, Roger. (2005). Recommended effect size statistics for repeated measures designs. Behavior research methods. 37. 379-84. 10.3758/BF03192707.
Baker, Malcolm, and Jeffrey Wurgler (2013), “Do strict capital requirements raise the cost of capital? Banking regulation and the low risk anomaly”, NBER Working Paper No. 19018
Barrios, V.E.J. and Blanco, J. (2003). The Effectiveness of Bank Capital Adequacy Regulations: An Empirical Approach. Journal of Banking and Finance, 27, pp.1935-1958.
Berger, A. N., & Bouwman, C. H. (2013). How does capital affect bank performance during financial crises?. Journal of Financial Economics, 109(1), 146-176.
Berger, A.N, Herring, H.J., and Szego G.P. (1995). The Role of Capital in Financial Institutions.Wharton School Center for Financial Institutions,Working Paper 95-01.
Beltratti, A. and R.M. Stulz (2012), “The credit crisis around the globe: Why did some banks perform better?” Journal of Financial Economics 105, 1-17.
Bunda I., Jean-Baptiste Desquilbet. Bank Liquidity and Exchange Rate Regimes. “Current Challenges, New European Perspectives” 3rd International Scientific Conference, May 2003, Sofia, Bulgaria. pp.24-43. ffhal-00422622
Chen, H., Chen, Q., & Gerlach, S. (2011). The Implementation of Monetary Policy in China: The Interbank Market and Bank Lending. Emerging Markets Economics: Macroeconomic Issues & Challenges eJournal.
Dimsdale, Nicholas. (1994). Banks, Capital Markets, and the Monetary Transmission Mechanism.. Oxford Review of Economic Policy. 10. 34-48. 10.1093/oxrep/10.4.34. Ediz, T., I. Michael, and W. Perraudin (1998). The impact of capital requirements on U.K. bank behaviour. Federal Reserve Bank of New York Economic Policy Review, October 1998.
- 20 -
Estrella, Arturo & Park, Sangkyun & Peristiani, Stavros. (2000). Capital Ratios as Predictors of Bank Failure. Economic Policy Review. 6. 33-52.
Luc Laeven & Lev Ratnovski & Hui Tong, 2014. "Bank Size and Systemic Risk," IMF Staff Discussion Notes 14/4, International Monetary Fund.
Raharjo, P., Hakim, D. B., Manurung, A. H., & Maulana, T. (2014). DETERMINANT OF CAPITAL RATIO: A PANEL DATA ANALYSIS ON STATE-OWNED BANKS IN INDONESIA. Buletin Ekonomi Moneter Dan Perbankan, 16(4), 369-386.
Rime, B. (2001). Capital Requirements and Bank Behaviour: Empirical evidence for Switzerland. Journal of Banking and Finance 25: 789-805.
Gambacorta, L., & Shin, H. S. (2018). Why bank capital matters for monetary policy. Journal
of Financial Intermediation, 35, 17-29.
Siebenbrunner, Christoph & Sigmund, Michael. (2017). Determinants of interbank market rates: theories and empirical evidence. SSRN Electronic Journal. 10.2139/ssrn.3082784. Holland, Márcio & Vieira, Flavio & Silva, Cleomar & Bottecchia, Luiz. (2011). Growth and Exchange Rate Volatility: a Panel Data Analysis. Applied Economics. 45.
- 21 -
Appendix
In the following section you can find all additional and more extensive figures used for this thesis. First you will find P-P plots of all variables used for the regression models to test their normality. Thereafter you will find an extensive and comprehensive view of the STATA and SPSS output used to make estimations of the effect sizes mentioned in the paper.
FIGURE A.1
- 22 -
FIGURE A.3
- 23 -
FIGURE A.5
- 24 -
- 25 - _cons 25.93797 4.779467 5.43 0.000 16.57039 35.30556 Nloans_assets .0184749 .0210879 0.88 0.381 -.0228566 .0598064 Ibratio -.0000345 .0002413 -0.14 0.886 -.0005075 .0004386 Capfunds_assets -.0374865 .0220216 -1.70 0.089 -.0806481 .005675 CAR -.3337925 .13909 -2.40 0.016 -.6064039 -.0611812 Δlnloans -.0010325 .0119722 -0.09 0.931 -.0244975 .0224326 GDPgrowth .3413427 .0780413 4.37 0.000 .1883846 .4943008 Bank_lev -72.38817 27.7827 -2.61 0.009 -126.8413 -17.93508 Bank_size -.7052132 .1912582 -3.69 0.000 -1.080072 -.330354 recessionNBER -.0635462 .3549513 -0.18 0.858 -.759238 .6321455 Interbankrate Coef. Std. Err. z P>|z| [95% Conf. Interval] Log likelihood = -85.63707 Prob > chi2 = 0.0000 Wald chi2(9) = 216.21 Estimated coefficients = 10 Time periods = 6 Estimated autocorrelations = 0 Number of groups = 9 Estimated covariances = 9 Number of obs = 54 Correlation: no autocorrelation
Panels: heteroskedastic Coefficients: generalized least squares Cross-sectional time-series FGLS regression
.
F test that all u_i=0: F(8, 40) = 4.07 Prob > F = 0.0013 rho .91391315 (fraction of variance due to u_i)
sigma_e 1.2588862 sigma_u 4.1017629 _cons 83.16071 20.24315 4.11 0.000 42.24777 124.0736 GDPgrowth .4126852 .0952742 4.33 0.000 .2201289 .6052415 recessionNBER -.1898473 .4049454 -0.47 0.642 -1.008272 .6285778 Bank_size -3.821884 1.007754 -3.79 0.000 -5.858631 -1.785137 Bank_lev -62.46471 27.18323 -2.30 0.027 -117.4041 -7.525353 Capfunds_assets -.0481449 .0695978 -0.69 0.493 -.1888072 .0925174 Interbankrate Coef. Std. Err. t P>|t| [95% Conf. Interval] corr(u_i, Xb) = -0.8992 Prob > F = 0.0000 F(5,40) = 24.12 overall = 0.1437 max = 6 between = . avg = 6.0 within = 0.7509 min = 6 R-sq: Obs per group:
Group variable: Bank_id Number of groups = 9 Fixed-effects (within) regression Number of obs = 54 . xtreg Interbankrate Capfunds_assets Bank_lev Bank_size recessionNBER GDPgrowth, fe
FIGURE A.8FIXED EFFECTS PANEL DATA REGRESSION MODEL OF THE SEMI-EXTENSIVE MODEL
- 26 -
rho 0 (fraction of variance due to u_i)
sigma_e 1.7227754 sigma_u 0 _cons 6.288933 1.585685 3.97 0.000 3.181046 9.396819 Capfunds_assets .0221369 .0226487 0.98 0.328 -.0222537 .0665275 Bank_lev -61.96836 21.98127 -2.82 0.005 -105.0509 -18.88586 Interbankrate Coef. Std. Err. z P>|z| [95% Conf. Interval] corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0115 Wald chi2(2) = 8.92 overall = 0.1489 max = 6 between = 0.0000 avg = 6.0 within = 0.0000 min = 6 R-sq: Obs per group:
Group variable: Bank_id Number of groups = 9 Random-effects GLS regression Number of obs = 54 . xtreg Interbankrate Bank_lev Capfunds_assets
FIGURE A.10TIMESERIES REGRESSION OF THE BASELINE REGRESSION
FIGURE A.11TEST OF BETWEEN-SUBJECTS EFFECTS OF THE EXTENSIVE MODEL
Interbankr~e 0.1192 0.0088 0.1276 -0.5570 0.6922 -0.1264 -0.3646 -0.1778 -0.2169 1.0000 recessionN~R -0.0402 0.1541 -0.0509 0.1658 -0.1669 0.0097 0.2114 0.0221 1.0000 Bank_size 0.1806 -0.2018 -0.6059 -0.1153 -0.1080 0.9486 -0.1958 1.0000 Bank_lev 0.5247 -0.1451 -0.0035 0.2550 -0.2444 -0.2615 1.0000 Totalassets 0.0508 -0.2287 -0.4265 -0.0938 -0.0830 1.0000 GDPgrowth 0.0815 0.0488 0.0905 -0.3982 1.0000 CAR -0.5052 0.2240 -0.1932 1.0000 Capfunds_a~s -0.2329 0.0690 1.0000 Ibratio -0.2844 1.0000 Nloans_ass~s 1.0000 Nloans~s Ibratio Capfun~s CAR GDPgro~h Totala~s Bank_lev Bank_s~e recess~R Interb~e FIGURE A.12CORRELATION TABLE OF ALL VARIABLES USED IN THE MODEL
