The Relationship between Liquidity Risk and Failure of Commercial Banks in Kenya

This study sought to establish the relationship between liquidity risk and failure of commercial banks in Kenya in the years 2013 to 2016. Additionally, the study endeavoured to establish the effect of capital adequacy, asset quality, management quality, earnings, sensitivity to market and size on the failure of banks in Kenya. To achieve this goal, secondary data was collected from the websites of operational banks while data for failed banks was collected from reports published by the central bank of Kenya, corroborated with publications in past years newspapers. Panel logit regression was used to analyze the data using Eviews 9.5 student version. The results of the regression revealed that there was a positive and significant relationship between liquidity risk and bank failure, implying that liquidity increased the likelihood of failure. The study also found a positive and significant relationship between bank failure and asset quality and earnings indicating that they increased the likelihood of failure. The study found a negative and significant relationship between bank failure and management quality and sensitivity to market implying that they decreased the likelihood of bank failure. Capital adequacy and bank size were found to have insignificant relationship with the failure of commercial banks in Kenya. These findings are valuable to managers in understanding how the variables of the study increase or decrease the likelihood of failure so that they may come up with appropriate strategies for managing the various risks facing their banks


LIST OF TABLES
Banks play an indispensable function in a country's financial system and economy as a whole through offering intermediary and liquidity services (Heffernan, 2005). The function of financial intermediation inherently exposes banks to liquidity risk through the activity of transforming the maturity of short-term liabilities and demand deposits into long term maturity assets in form of loans (Venkat & Baird, 2016). Farag, Harland, and Nixon (2013) define liquidity risk as the possibility of a colossal number of depositors and investors withdrawing amounts held in their accounts at once, thereby depleting a bank's funds thus prompting it to sell off its assets at unfavourably low prices in order to remain afloat. According to Kaufman (1996), a bank is deemed to be a failure if the market price of its assets is diminished to an extent that it is less than the market price of its liabilities. Daley, Matthews and Whitfield (2008) contend that bank failure includes closure, bankruptcy, supervised merger, or direct government assistance. Minamihashi (2011) consider a bank to be a failure if it suspends issuance of new loans or credit to its clients. According to , liquidity, undercapitalization, safety, soundness, and fraud are some of causes of bank failure.
Theoretically, liquidity risk and bank failure can be linked through the following four theories selected from previous studies because purportedly they can explain the most proximate sources or reasons for failure of commercial banks. The liquidity preference theory, which posits that entities hold cash because they want to transact with it or because they speculate some opportunity which they wish to take advantage of or because they wish to be precautious about some uncertain future (Keynes, 1936). The moral hazard theory which posits that when an entity is not obliged to bear fully the consequences of its behaviour, it may be tempted to accept more risk than that which is necessary (Zeckhauser, 1970). The theory of herd behaviour which argue that individuals and even institution tend to mimic the behaviour, strategies and trade of others even when their common senses dictate that they should do otherwise (Banerjee, 1992). Lastly, the study is also anchored on the financial contagion theory, which argues that the failure of a single bank can precipitate failure of other banks that come into contact with it in normal business transactions thus occasioning a banking crisis (Kaufman, 1992).
Kenya had 43 commercial banks on 31st December 2012, which form the cohort population of this study (CBK, 2012). During the period under study, three banks were acquired and three others were put in receivership due to liquidity problems and malpractices (CBK, 2015a(CBK, , 2015b(CBK, and 2016. The number of fully functional banks in Kenya had reduced to 39 by the end of the study period on 31st December 2016. In the year 2017, Mayfair Bank and Dubai Islamic Bank were licensed (Central Bank of Kenya, 2017). Nonetheless, banking business in Kenya is robust with large banks controlling 80 percent of the market while small and medium bank share the remaining 20 percent (Deloitte, 2016). The fact that some banks have failed in Kenya during the period under study provide enough cases of failure and survival making it an opportune context to investigate failure of banks and liquidity risk. Pyle (1999) posit that banks face many risks including market risk, credit risk, operational risk, performance risk and liquidity risk. Liquidity risk has been defined by Wu and Hong (2012) as that risk resulting from a bank's inability to meet payment obligations promptly and cost-effectively. Banks (2005) expound liquidity risk as the uncertainty that a bank may incur loss due to a lack of cash or its equivalents or that it may suffer economic loss in its attempt to procure the cash vital for its operations.

Liquidity Risk
According to Farag et al. (2013), liquidity risk can take two forms: Funding liquidity risk, which results from the bank having insufficient cash and collateral to settle debts owed to counterparties and customers immediately; and market liquidity risk which is the possibility that the bank's assets cannot be cashed quickly without incurring large discounts.
Liquidity is important to banks because it compensates for expected and unexpected fluctuations in their financial position besides providing funds for their growth (Van Greuning & Brajovic-Bratanovic, 2009). If a bank suffers from a liquidity shock and it fails to repay depositors and other creditors amounts owed to them punctually then it may be declared cash-flow insolvent (Farag et al., 2013). The importance of liquidity risk (both funding and market) is underscored by the fact that it has potential to cause severe liquidity spirals (Gomes & Khan, 2011). Severe liquidity crisis may arise when numerous depositors withdraw their savings at once leaving the bank without funds causing what is known as a bank run. Such bank runs can even cause "healthy" banks to fail affecting the entire economy (Diamond & Dybvig, 1983). Another reason why liquidity risk is important as observed by Acerbi and Scandolo (2008) is that it can explode market and credit risks in addition to transforming loss in one bank into a systemic and contagious crisis.
According to Banks (2005), the two most frequently used measures of liquidity risk are liquidity ratios and liquidity gaps. Empirical studies reviewed in this study reveal there is no single measure of liquidity risk that suits all studies. Wu and Hong (2012) used a combination of government securities ratio, brokered deposits ratio and the difference between treasury bills rate and the interbank lending rate as proxies for liquidity risk.
Liquidity coverage ratio and the net stable funding ratio have also been used to quantify liquidity risk by Cucinelli (2013) and Muriithi and Waweru (2017). Ogilo and Mugenyah (2015), and Wekesa (2016) adopted the loans to deposits ratio to measure the same. Berger and Bouwman (2009) propounded the Berger and Bouwman comprehensive liquidity measure (BB) which has been used by Fungacova, Turk and Weill (2015) and Zheng, Cheung, and Cronje (2016). Liquidity risk is measured in this study by ratio of loans to deposits.

Bank Failure
According to Kaufman (1996), a bank is deemed to be a failure if the market price of its assets is diminished to an extent that the market price of its assets is exceeded by that of its liabilities. Thomson (1992) and Heffernan (2005) posit that banks should be considered to be failures if they have been closed, declared bankrupt, have been forced to merge with a healthy bank, or have been directly assisted by the government. A bank is considered to be a failure if it is closed and it cannot redeem all its notes at par or face value (Rolnick and Weber, 1984).
Bank failure consequences can be financial, economic, and social or even political (Okeahalam, 1998). As stated in Ragalevsky and Ricardi (2009), bank failure can adversely affect its depositors with balances exceeding prescribed deposit insurance limits, vendors, staff, landlords, borrowers, other banks and its counterparties.
Notwithstanding, it can damage the economy and erode public confidence on the entire banking sector. Additionally, Minamihashi (2011) postulate that bank failure affects client firms of failed banks since they cannot borrow from the failed banks causing them to suffer from a credit crunch something that may stagnate the activities of such clients. According to Sarkar and Sriram (2001), bank failures tend to compromise investor confidence and deplete resource of deposit insurance Corporations. Müller and Trümpler (2004) opine that failure of banks precede a significant drop in output and an upsurge of unemployment. Macey and Miller (1988) argue that the government's stake in the financial stability of banks and the fact that many people perceive healthy banks as essential for a stable economy makes bank failure a matter of concern and hence important to study. Another reason why bank failure may warrant a study is the resultant loses. James (1991) found out that failed banks lose a substantial value of their assets averaging to about 30% while bank closure cost averaged to 10% of the banks' assets.
According to Bouvatier, Brei and Yang (2013), bank failure is measured by an indicator or dummy variable which indicates whether an event has or not happened. According to Skrivanek (2009), indicator variables are used in regression to assign either "1" or "0" to members of two mutually exclusive categories. This study will measure bank failure according to Zheng, Cheung, and Cronje (2016) where a binary performance variable is adopted to signify whether a bank fails in a particular financial year. As such a bank that fails within a period 12 months will be flagged failure and given a score of one, otherwise it is flagged surviving and given a score of zero.

Liquidity Risk and Failure of Commercial Banks
Matz (2011) postulate lack of liquidity to be the most immediate cause of bank failure alongside credit problems. He argues that overreliance of banks on short-term funding exposes them to liquidity risk and eventually contributing to bank failure. Banks (2005) concurs with the above view that many global bank failures of the past few decades have been attributed to the combined effects of credit risk and liquidity risk. Illiquidity follows insolvency and lack of sufficient funding liquidity is a sign of banks in serious financial difficulties (CBK 2013a). Inadequate management of liquidity is attributable to the bank failure crisis of 2007 since many banks failed notwithstanding adequate capitalization (Basel Committee on Banking Supervision, 2013). Inadequate and untimely handling of liquidity risk in one bank can potentially trigger a crisis of confidence affecting the entire banking industry (CBK, 2013). Wu and Hong (2012) explain that liquidity risk affects bank failures through both the idiosyncratic channel, through which banks fail due poor liquidity risk management practices and the systematic channel through which risk is inherent in the aggregate market and thus cannot be eliminated through diversification. According to Diamond and Rajan (2005), liquidity risk can also cause bank failure by shrinking the common pool of liquidity thereby creating or worsening aggregate liquidity shortages. Bank failure due to aggregate liquidity shortage stems from the fact that banks cannot finance their projects or operation through the interbank borrowing market because of insufficient amount of liquidity that has resulted in the destruction of existing lending relationship (Wagner, 2007). Theoretically failure of banks and illiquidity are related by theories like: bank contagion theory (Kaufman, 1992), herding behaviour theory (Banerjee, 1992), moral hazard theory, (Pauly, 1968) and liquidity preference theory (Keynes, 1936).

Research Problem
Commercial banks play a crucial role of creating liquidity for their clients by channelling funds from surplus units to deficit units (Madura, 2015). The creation of liquidity for clients through advancement of loans reduces the recipient's liquidity risk, but generates liquidity risk to the bank (Fiedler, 2011). This exposure to risk according to Matz (2011) emanates from banks' reliance on short-term deposits to finance long term loans. Yimer (2016) argues that as much as inadequate liquidity can cause bank failure through bankruptcy, excess liquidity can equally lead to bank failure because of poor profitability since liquid assets yield little or no interest income to the bank. Several studies have been done with a view to establishing the relationship between liquidity risk and bank failure. Hong and Wu (2013) found that systemic funding liquidity risk predicted the 2008 and 2009 banking crisis significantly. Fungacova, Turk and Weill (2015) found that creating excess liquidity increases the chances of bank failure. Zheng, Cheung and Cronje (2016) found a negative and significant relationship between failure risk and liquidity risk. Locally, Cheserek (2007) investigated determinants of bank failures in Kenya but did not consider liquidity risk as one of the determinants. Other local studies on liquidity risk such as Maaka (2013), Kibuchi (2015, Muriithi and Waweru (2017) and Musembi, Ali and Kingi (2016) studied liquidity risk as it relates to financial performance. Another set of studies notably Ogilo and Mugenyah (2015) and Kimathi, Mugo, Njeje, and Otieno (2015) investigated the determinants of liquidity risk.
From the above studies the association between the concepts of liquidity risk and bank failure is mixed in that both inadequate and excessive liquidity can lead to bank failure through different channels. The review also shows that there is a contextual gap to be filled in that none of the local studies has examined how liquidity risk and bank failure are associated. This study seeks to fill these gaps by answering the question: Is there a relationship between liquidity risk and failure of commercial banks in Kenya?

Research Objective
To find out the relationship between liquidity risk and failure of commercial banks in Kenya the period 2013 to 2016.

Value of the Study
This study will be of value to bank managers and professional analysts in understanding the association between liquidity risk and commercial bank failures in Kenya. The study will help managers and professionals formulate liquidity risk management strategies that are informed with research. Such strategies will minimize bank failure due to liquidity issues.
In a situation where banks are failing rapidly like they are doing in Kenya, policies, regulations and statutes need to be scrutinized keenly to surmount such crisis. The study will be useful to policy makers and regulators in formulating comprehensive guidelines to banks on matters of liquidity to ensure a reduction or elimination of incidents of bank failure. In particular it will help regulators in developing both offsite and onsite monitoring tools that cover liquidity risk in detail and from the assets, liabilities and off balance sheet activities.
This study will be beneficial to scholars in that it seeks to establish the relationship between liquidity risk and bank failure from a Kenyan context. This study will generate knowledge on the applicability of liquidity and bank failure theories originated in other counties to the local banking environment. The study will also be useful to local scholars in that it intends to use and popularize logit regression analysis of variables as an alternative to the overly used simple and multiple linear regression analysis.

Introduction
This chapter reviews the theoretical and empirical literature on liquidity risk and failure of commercial banks besides presenting the conceptual framework showing how the concepts relate with each other.

Theoretical Review
This section reviews some of the theories that explain how liquidity risk explains failure of banks as suggested by prior studies. In particular, we shall review herd behaviour, financial contagion, moral hazard, and liquidity preference theories.

Herding Behavior Theory
Banerjee (1992) defines herd behaviour as the act of everyone imitating what everyone else is doing, in spite of the fact that their private information may be suggesting that they should do something quite different. The rational herding theory explains liquidity risk and bank failure in two ways. Firstly it is the herding behaviour of depositors. This may happen when numerous depositors simultaneously apportion a very high probability to the bank potentially losing its assets and they resort to simultaneous withdrawing of their deposits exposing the bank to a run (Kaufman, 1987).
Secondly, herding behaviour can emanate from bank management when they make same or similar risk-taking, management, and asset holding decisions. Herding also can manifest at the time when these banks sharing the same information or facing similar circumstances end up making decisions that are similar, or when they mimic each other's lending behaviour (Liu, 2014). The effect of this managerial herding behaviour is that firms with similar trading strategies will plunge into similar perils during liquidity crisis and no bank will be salvaged by counter party funds from the other bank.
As a result there will be a systemic liquidity risk. Pedro (2014) argues that uncertainty and herd behaviour are causes of financial crises. Herd behaviour leads to asset bubbles and irrational panic that in turn lead to quick reduction of prices in the financial markets and as such banks in liquidity problems end up selling their assets at a loss which eventually leads to illiquidity and bank failure (Ionescu, 2012). Kaufman (1992) proposed the bank contagion theory with the following characteristics:

Bank Contagion Theory
it occur with fastness, then spreads more broadly within the industry, resulting in several failures thereby occasioning large losses to depositors and may even spread beyond the banking industry thereby damaging the financial system and the macroeconomy as a whole. Masson (1998) postulates that the term contagion should be applied when a crisis situation at one point triggers a crisis at other points in a manner that macroeconomic fundamentals cannot explain it. Eichengreen, Rose, and Wyplosz (1996) established that contagion mostly affects banks that trade with each. Calomiris and Mason (1997), posit that the contagion theory can explain bank failure in that when depositors lack information about the incidence of an observable shock across banks they may have motivation to withdraw their deposits and wait until such information asymmetry is resolved. This may precipitate aggregate disturbances that may collapse several banks or the whole banking system. As Diamond and Rajan (2005) posit bank failure leads to a reduction in the common pool of liquidity, thereby precipitating or worsening aggregate liquidity shortages that could lead to a contagion of failures and a total breakdown financial system.

Moral Hazard Theory
The theory of moral hazard, attributed to Pauly (1968) but formalized by Zeckhauser (1970) refers to change of behaviour that takes place once financial consequences of possible peril have been shifted to another through insurance. According to Dowd (2009), moral hazard issues results from a situation where an agent is responsible for managing the interests of another, but the first party has inclination to foster his own interests first. Madura (2015) defines moral hazard for banks as the act of banks taking more risk simply because their depositors are protected by insurance and purports that that many of the banks that failed in the 1980s and 1990s were a result of taking excessive risks. As Umar and Sun (2016) indicate, moral hazard can also lead to bank failure through sales agents and bank manager taking excessive risk in lending in a bid to make commissions, improve performance and compete with other banks by advancing loans to clients resulting to higher non-performing loans (NPLs). (2016) give two forms of moral hazard problems that can lead to bank failure. The first is managerial rent-seeking in which managers pursue their own benefits through investing in 'pet projects' by not monitoring loans sufficiently. The second type of moral hazard is a result of diverging interests of shareholders and creditors in which shareholders make risky loans but end up shifting the risk to depositors. The last moral hazard channel is where directors give themselves, other insiders and businesses or organizations associated to them huge loans, which in most cases will end up being nonperforming loans and consequently paralyzing the bank leading to its closure. This was presumed to be the case in the failure of Imperial bank of Kenya. A strong relationship between loan growth rate and failure of banks indicates moral hazard problem in lending.

Liquidity Preference Theory
The liquidity preference theory (Keynes, 1936) holds that people need money because they have some expenditure to finance with it, that is the transaction motive, or because they speculate that interest rates may be higher in the future (the speculative motive) and wish to hold it until the interest rates have risen or because they are uncertain about the future (precautionary motive) and do not want to risk (de Carvalho, 1999). The precautionary motive theory (Keynes, 1936) argues that banks should hold cash balances for unforeseen contingencies. Thus banks should have extra cash and its equivalents over and above what is needed for day to day transactions so that in case there is an unanticipated demand for cash they can be able to surmount it without resorting to impromptu sale of asserts which may lead to selling assets at unreasonably high and unbearable discount rates, nor to buying credit at uneconomically exorbitant interest.

Determinants of Bank Failure
This section reviews the determinants of failure of commercial bank. The study, based on reviewed studies, will utilize bank fundamental analysis factors such as asset quality, earnings, capital adequacy, management capability, liquidity and market sensitivity. In addition the study includes bank size since a number of studies reviewed have shown that bank size is significantly related to bank failure. In total seven determinants are reviewed as below: The first determinant is liquidity risk which is the independent variable of the study.
The liquidity shortage hypothesis of bank failure argues that banks fail because of a liquidity shock that causes them to be unable to meet their contractual debt obligations as they fall due (Fungacova, et al. 2015). This illiquidity that impairs the ability of a bank to honour its obligations promptly is known as liquidity risk. The inclusion of this variable as a determinant is informed by studies such as Hong and Wu (2013), Wu and Hong (2012) and Ugoani (2015).
Another determinant of bank failures based of reviewed studies is capital adequacy which is included as the first control variable. A bank is said to have adequate capital if that capital is sufficient to absorb any shocks that it may suffer (Kibuchi, 2016).
Besides, the capital base helps depositors assess the riskiness of the bank and signals the bank's continued ability to honour its obligations (State Bank of Pakistan, 2001).
Adequate capital is important for banks because it affects the returns for its shareholders, and also because a certain minimum capital is required by regulatory authorities for banks to continue operating legally (Mishkin & Eakins, 2012).
According to Kaufman (1996) when the capital base of a bank is diminished until it has a negative net worth then it should be deemed to be a failure. On the contrary banks with sufficient capital can absorb shocks due to large losses and their time to insolvency is long. Based on previous studies bank failure and capital adequacy are negatively related with each other. Capital adequacy will be measured by the ratio of equity to total assets.
Asset quality or the magnitude of loans and advances impairment makes our third determinant and is included as the second control variable. As Muriithi (2016) posit, a deterioration of asset quality increases credit risk which in turn reduces the banks' expected profits. Deteriorating value of asset value has far reaching effects as losses will eventually be written-off against the bank's capital base jeopardizing its earning capacity leading to insolvency (State Bank of Pakistan, 2001). Waweru and Kalani (2009) found a significant relationship between nonperforming loans and bank failure.
Management capability, the fourth determinant, is the ability of the 'board of directors and management of banks, in their respective roles, to identify, measure, monitor, and control risks to ensure the safety, soundness, and efficiency of an institution's operations and comply with applicable laws and regulations (Federal Reserve System, 1996). The proxy used in this investigation is the ratio of total cost to total income in accordance with Zheng et al. (2016). Alternative measures include the ratios of total expenditure to total income and operating expense to total expense (State Bank of previous studies such as Cheserek, 2007 andBabar &Zeb, 2011, this study will measure earning by using return on equity (ROE), that is, net income after taxes to total equity as its proxy.
Sensitivity to market is also included as determinant number. Sensitivity to market measures how a bank responds to unfavourable changes in interest rates, foreign exchange rate, and such like factors (Babar & Zeb, 2011). According to Federal Deposit Insurance Corporation (2007) market risk for a bank affects capital and earnings through changes in interest rates. These adverse changes in interest rate in turn reduce earnings which also affect a bank's liquidity and capital adequacy negatively.
According to Babar & Zeb (2011), the ratio of total securities to total assets can be used to measure sensitivity to market risk and this is the measure adopted by this study.
Lastly we include bank size because according to Zheng et al. (2016), the relationship between bank failure and liquidity risk is influenced by bank size and it can be negative for large and positive for small. Zheng et al. posit that bank size can be used to test both the moral hazard theory and the precautionary motive of the liquidity preference theory, in that if the relationship between liquidity and bank failure is negative for large banks the moral hazard theory is confirmed and if the relationship between bank failure and liquidity risk is positive for small banks then the precautionary motive is confirmed.
This study will be measure bank size by natural logarithm of total assets in line with Zheng et al. (2016).

Empirical Review
Cheserek (2007)  Separate descriptive statistics were produced for each group and compared by statistical package Stata 11. The analysis showed that the sample of failed banks had greater risk of failure than the sample of surviving banks. The study found gross domestic product growth rate to exert significant influence on bank failure together with liquidity, earnings, loan to assets ratio, loans to deposit ratio, deposits to total assets ratio, gross revenue ratio, core capital to total risk weighted assets, bank size, management quality and nonperforming loans to total loans. logit regression results of the study showed that the relationship between liquidity and failure risk depends on the size of the bank. The empirical results supported both the moral hazard and precautionary motive theories. That is to say higher liquidity is associated with lower probability of bank failure for the moral hazard theory and high liquidity may lead to higher probability for bank failure for large banks for the precautionary motive theory. Muriithi and Waweru (2017) studied the effect of liquidity risk on financial performance of all 43 registered commercial banks in Kenya using secondary data from financial statements as available at the central bank of Kenya. They measured bank performance using return on equity ratio while liquidity risk was measured by the liquidity coverage ratio (LCR) and the net stable funding ratio (NSFR). Using random effects panel data analysis and generalized method of moments, the results indicated that NSFR was negatively correlated with profitability in both the short and long run.
They also found out that LCR had insignificant effects on the performance of commercial banks in Kenya. They generalized that liquidity risk has a negative effect on performance of banks and as such managers should pay greater attention to liquidity risk management

Conceptual Framework
The conceptual model in figure 2.1 below shows how the independent variables, the control variable and the dependent variable are related. The dependent variable will be measured by the probability of bank failure will be either one if a bank fails or zero if it does not fail in a particular year. The independent variable, liquidity risk, is represented by the ratio of liquid assets to customer deposits. The control variables, capital adequacy, asset quality, management capability, earnings, sensitivity and bank size are included to control for omitted variables error in the regression.

Research Design
The study subscribed to a quantitative descriptive research design to investigate if there was a relationship between liquidity risk and failure of commercial banks in Kenya. A descriptive study according to Saunders, Lewis, and Thornhill (2011), seeks to give an accurate profile of events, persons, or situations by giving answers to who, what, where, when or how questions.
The study adopted a quantitative descriptive research design because the phenomena under scrutiny had already happened at the time of the study, and the researcher was not in control of the independent variables to manipulate them as is the case in experimental study. The second reason for choosing a descriptive approach was because the research objective was finding if there was a relationship between the liquidity risk and bank failure as opposed to whether the independent and control variables were the cause of the failures.

Population of the Study
The population of the study was all the 43 commercial banks licensed and operating in Kenya on 1 st of January 2013. A census study approach was adopted because the total number of banks in Kenya is few compared with other countries such as USA that have

Data Collection
The study relied on secondary data extracted from the audited annual financial statements and other disclosures of all commercial banks licensed to operate in Kenya as posted in their respective websites or gleaned from the annual bank supervision reports of the central bank of Kenya, for the period 2013 to 2016. Secondary data was used because it is concrete, objective and does not change with time.
Data collected from each statement included book value of total assets, liquid assets, total equity, total deposits, total income, gross loans and advances to customers, gross nonperforming loans, total operating expenses, net profit or loss before tax, and total securities. The study used the secondary data collection form shown in appendix 3 to summarize data for each bank. This data was used to compute the various financial ratios shown in Appendix 4 that were then used for correlation and logit regression analysis.

Data Analysis
The study employed logistic (logit) regression analysis to establish the relationship between bank failure and liquidity risk. Logit regression was selected because of the binary nature of the dependent variable bank failure. Secondly, multiple regression could not be used because the assumptions of assumptions of linearity, constant variance, absence of special causes, normality, and independence of the test data cannot be in met a binary variable like bank failure (Fang, 2013) .The study used Huber/White Quasi-Maximum Likelihood (QML) method to assess the standard errors of the logit regression. The data was analysed by Eviews 9.5 student edition.

Analytical Model
The analytical model of the study consists of bank failure as the dependent variable and liquidity as the independent variable. However to control for omitted variables regression error, capital adequacy, asset quality, management capability and sensitivity to market and bank size were used as control variables. The logit regression model employed to analyse the effect of liquidity risk on bank failure is specified below.

Tests of Significance
The significance of the coefficients was tested by the Wald test while goodness of fit and robustness were tested by the Huber-White method at 5% significance levels. Pvalues were used to discriminate significant and insignificant regression variables.

Operationalization of Variables
The table below show how the variables of the study were operationalized.   Asset quality (column 5 of Table 1) represented by the ratio of gross nonperforming loans to gross loans and advances had a mean of 0.096 or 9.6 percent. The maximum and minimum ratios were 51.4% and 0 respectively. The standard deviation of asset quality was 0.084 or 8.4 percent. The skewness of asset quality was 1.923 and 7.814 is the kurtosis.
Management quality shown on the sixth column of Table 1 and measured by the ratio of total operating expenses to total income had a mean and median of 0.448 and 0.416 respectively. The maximum and minimum ratios were 1.402 that is 140.2 percent, and 0.116 or 11.6 percent respectively. The standard deviation of management quality was 21.9 percent and skewness and kurtosis were 1.475 and 6.588 respectively.
Earnings quality, shown on column 7 of Table 1 and measured by the ratio of net profit (or loss) before tax to total equity (ROE), had a mean and median of 0.147 or 14.7 percent and 0.195 or 19.5 percent respectively. The maximum and minimum earnings observed were 0.436 or 43.6 percent and -0.414 or a loss of 41.4 percent respectively.
The standard deviation, skewness and kurtosis for return on equity were 0.171, -1.225 and 4.382 respectively.
Sensitivity to market (SM), column 8 of Table 1, measured by the ratio of total securities to total assets, had a mean of 0.228 or 22.8 percent and a median of 0.191 or 19.1 percent. The maximum sensitivity to market observed was 0.636 or 63.6 percent and the minimum observed was zero. The standard deviation was 13.5 percent and skewness of 0.883 and kurtosis of 3.893.
Column 9 of Table 1 shows total assets in Ksh. Million. Although the natural logarithm of total assets was used to measure bank size, the descriptive statistics presented are those of total assets. The average size of commercial banks in Kenyan is 75.45 billion.
The maximum size observed was 504.78 billion and the smallest has a size of 2.9 billion. The standard deviation of size is 96.5 billion. Bank size has a skewness of 1.938 and kurtosis of 6.875.

Correlation Analysis
This section gives results of correlation analysis and shows the degree of linear association between independent and control variables to avoid duplication of variables. As shown on the second column of Table 4.2 the relationship between failure and the explanatory variables is low with the highest being -0.083 for failure and sensitivity to market and the lowest being 0.014 between failure and loan to deposit ratio (LAD).
Column 3 of Table 4

Regression Analysis
In order to find how the independent variables explain failure of commercial banks in Kenya over the study period panel logit regression analysis was carried out and the results are shown on Table 4.3 below. Source: Author, 2017 As shown in Table 4.3 the coefficient for liquidity risk or loans to deposit ratio (LAD)

Discussion of Research Findings
This section discusses the descriptive statistics, correlation and regression analyses presented above. It interprets and compares the findings with previous studies.

Descriptive statistics
The study findings indicate that the mean of liquidity risk is above the regulatory minimum ratio of 20 percent. This implies that a majority of Kenyan banks are compliant and hence the banking sector on overall does not suffer from liquidity risk problems despite some few banks having liquidity ratios below regulatory minimum.
From theory, high liquidity can lead to bank failure through low profitability, while low liquidity may lead to failure through liquidity shortages and regulatory sanctions (Fungacova, et al., 2015). The four years mean of capital adequacy is also above the regulatory minimum of 14.5 percent. This suggests that on average Kenyan banks comply with the CBK capital requirement regulations. This indicates that the capital adequacy of Kenyan commercial bank is health despite a few incidences of undercapitalization. CBK, (2016c) requires banks to comply with the ICAAP document and ensure that their total capital levels are adequate, consistent with their strategies, business plans, risk profiles and operating environment on a going concern bases. Banks with very low capital adequacy ratio face the risk of regulatory sanctions and excessive monitoring. A high ratio of capital adequacy implies that a bank can absorb large unexpected losses and hence faces less risk of closure due to bankruptcy. On the contrary a lower ratio indicates inability to absorb shocks in profitability and therefore has higher probability of closure due to bankruptcy risk (Kibuchi, 2015).
Notwithstanding, the 4 years asset quality ratio is higher than the 31 st December 2016 rate indicating that Kenyan banks are in the process of reducing the rate of nonperforming loans. Waweru and Kalani (2009), in a previous study found that nonperforming loans were a major cause of bank failure. Equally, the four years mean of management quality (MQ) of is higher than that of the year ended 31 st December 2016. The fact that the 2016 alone efficiency ratio is lower implies that banks are cutting on operating expenses hence improving on efficiency. Yirgu (2017) found management quality, proxied by operating expenses to total income, to be negatively and significantly related to bank distress. As argued by Canicio and Blessing (2014), the higher this ratio is the higher the likelihood of the bank to fail.

Correlation Analysis
There is a positive correlation between bank failure and liquidity risk, capital adequacy, asset quality, and earnings. This implies that increase in these variables increase the likelihood of bank failure. Similarly, a negative correlation between bank failure and management quality, sensitivity to market and bank size implies that increase in these variables decreases the likelihood of bank failure.
A positive correlation between liquidity risk and capital adequacy, earnings, and sensitivity to market implies that as banks take more liquidity risk; capital adequacy, earnings and sensitivity to market tend to increase as well. On the other hand a negative correlation between liquidity risk and asset quality, management quality and bank size implies that as liquidity risk increases these variables tend to decrease. Spearman rank correlation also shows a positive correlation between capital adequacy and sensitivity to market which means that as capital adequacy increases so does sensitivity to market increase. Conversely, a negative correlation between capital adequacy and asset quality, management quality, earnings and size implies that as capital adequacy increases these variables also decrease. Similarly, a positive correlation between asset quality and management quality indicates as one of these variables increase so does the other. On the contrary a negative correlation between asset quality and earnings, sensitivity to market and bank size indicates that as asset quality (or nonperforming loans) increases earnings, sensitivity to market and size deceases as well.
Spearman rank correlation also shows a negative correlation between management quality and earnings, sensitivity to market and bank size implying that as management quality or operational costs increase earnings, sensitivity to market and bank size decrease. A positive correlation between earnings and sensitivity to market and bank size implies that as earnings increase sensitivity to market and bank size also increase.
Finally, a positive correlation between sensitivity to market and bank size means that as sensitivity to market increases bank size increases.
To summarize it all, the spearman correlation rank order analysis does not show any strong correlation between the variables that could alarmingly affect regression analysis. As such all variables were incorporated in the logit regression analysis.

Model Goodness of Fit
According to Gujarati and Porter (2009)

Regression Analysis
The fact that liquidity risk has a positive coefficient implies that increase in liquidity increases the likelihood of failure. According to Zheng et al. (2016), the relationship between liquidity risk and bank failure can be negative if failed banks suffer from the moral hazard problem or positive if the banks are pursuing the precautionary motive of the liquidity preference theory. The fact that the coefficient is positive and significant at 95% confidence level indicates that failed banks had stocked piled liquid asset for precautionary reasons. This finding is contrary to Canicio and Blessing (2014) and Yirgu (2017), who found liquidity to be significant and negatively correlated with bank crisis or failure. The finding is however consistent with the findings of Sahut and Mili (2011) and Pena (2016) who found liquidity to be positively correlated with banking crisis, and Zheng et al. (2016) who found a positive correlation between failure of small banks and liquidity risk. The finding is also consistent with Berger and Bouwman (2014) who through trend analysis found that liquidity creation tends to be higher prior to financial crisis. Given that the banks that failed in Kenya were small banks, this finding is considered valid.
Capital adequacy has a positive coefficient implying that increase in equity in the

Summary of Findings
The purpose of the study was to investigate the relationship between liquidity risk and failure of commercial banks in Kenya. Additionally, other variables that have been empirically established to have relationship with bank failure were included as control variables. Consequently, the relationships between bank failure and CAMES were also investigated.
The study found that there was a significant positive relationship between bank failure and liquidity risk implying that liquidity risk increased the likelihood of failure of commercial banks in Kenya in the years 2013 to 2016. This finding empirically supports the precautionary motive of the liquidity preference theory. The study also found a positive and significant relationship with asset quality and earnings indicating that they increased the likelihood of failure for the banks studied. The study found that management quality had a significant negative relationship with bank failure. This implies that management quality reduced the likelihood of bank failure. The study further found that sensitivity to market was negatively and significantly related to bank failure indicating that sensitivity to market reduced likelihood of bank failure. The study however did not find a significant relationship between bank failure and two of the control variables viz. capital adequacy and bank size.
The Mcfadden R-squared of the regression was 0.327 indicating that jointly, liquidity risk, capital adequacy, asset quality, management quality, earnings, sensitivity to market and bank size explain 32.7 percent of failure.

Conclusion
The study found a positive and significant relationship between liquidity risk and bank failure. According Zheng et al. (2016)  The study also found that management quality had a negative and significant relationship with bank failure. Since this was measured by the ratio of operating expenses to total income, then it implies that proper management of operating expenses has potential to reduce the chances of bank failure. As such, Kenyan commercial banks need to check on their efficiency in order to survive bank failure.
Earnings had a significant positive relationship with bank failure indicating that banks that failed were making profit prior to their failure. This indicates that bank managers need to understand the negative effect of pursuing short term profit as the main goal instead of pursuing shareholder wealth maximisation as discussed in the theory of the firm by Jensen and Meckling (1976). A positive and significant relationship between bank failure and earnings provides empirical support that bank managers of failed banks had not pursued shareholders wealth maximization as the main goal of the firm but had rather sought to maximize short-term profits. This leads to the conclusion that just as making losses can lead to bank failure through bankruptcy risk, excess profits can equally cause bank failure through excessive loans default risk.
The study's regression results show that sensitivity to market had significant negative relationship with bank failures, implying that holding more securities reduced the likelihood of bank failure. Bank should therefore have more liquid securities to reduce their chances of failure. The fact that both earnings and asset quality had a significant positive coefficient implying that the two variables increased the likelihood of bank failure indicates that moral hazard problem played a major role in the failure of commercial banks in Kenya.
The regression however did not find a significant relationship between bank failure and capital adequacy and bank size. This implies that the banks that failed suffered their predicaments mainly for the way they managed their business and not on size and capitalization. Since capital adequacy was an insignificant variable in the regression, then we can infer that failed banks were not insolvent at the time of failure.

Recommendations
Given that the liquidity risk, asset quality, management quality, earnings and sensitivity to market showed a significant relationship with bank failure, the study recommends that bank managers emphasize on this variables in their day to day management practices. Further, given that the coefficient of earning is positive implying that profitability increased the likelihood of bank failure, the study suggest that managers of banks should be risk sensitive instead of endeavouring to make short-term profits that increase the chances of bank failure. Further, Since liquidity risk has been identified as significant factor related to bank failure the study recommends that the central bank of Kenya implements the Basel III accord to enhance a more detailed reporting on liquidity especially the liquidity coverage ratio (LCR) and the net stable funding ratio (NSFR).
Since asset quality has been found to have a significant relationship with failure of commercial banks in Kenya, this may be an indication of weakness in the credit standards used to screen borrowers. The study recommends that bank managers reassess their current credit standards and consider replacing them with more comprehensive standards. These new and comprehensive credit standards will reduce nonperforming loans and save banks from impending failures.
Further as management quality has a negative and significant relationship with failure of commercial banks. The study recommends that banks should study and find an optimal level of operating expenses that can help in reducing the likelihood of failure.

Limitations of the Study
The main limitation of the study was that it relied on observations spanning over the four years 2013 to 2016 and hence likely to suffer from the small sample bias. This limitation was overcome by doing a census study that incorporated all commercial banks in Kenya. The study was also limited by the fact that at the time of the study the websites of all failed banks were closed hence limiting the availability of financial statements in a timely manner. This limitation was overcome by searching for the financial statement in past newspapers. Further annual reports of one the failed banks for the year 2015 were not incorporated in the analysis for lack of integrity as the auditors had qualified it.
A further limitation of the study is that it was designed to be a descriptive investigation.
As such even if the study found significant relationships between bank failure and LAMES no conclusion can be made that these variables were indeed cause of bank failure. This means that to establish the causality of the variables another study need to be done with the objective of finding the causes of bank failures.
Nonetheless, the relationship of bank failure with macroeconomic variables like inflation, unemployment, interest rate and gross domestic product growth rate were not investigated in this study. Since empirical studies show that these factors can contribute to bank failure, these factors are left for future studies.

Suggestions for Further Research
Since the Mcfadden R-squared is low it means that some explanatory variables were not included in the regression. As such the study should be replicated with additional non-financial determinants of bank failure such corporate governance, corruption and macroeconomic factors. Since the study relied on annual financial statements, it is suggested that the study be replicated with biannual or quarterly financial to capture the effects of the CAMELS on bank failure more proximately. The study concentrated on a narrow window of only four years. The study makes suggestion that a study covering all previous bank failures be done because it might yield more insightful results than those found by this study. A further suggestion is that detailed case studies of each failed bank be carried out to delve into factors that prompted failure of commercial banks in Kenya beyond the CAMELS factors.
The study suggests that a sequel study be carried out to establish whether the relationships shown in this study were also causal in nature or not. It is suggested that a study be carried out that will control for the effect of distressed banks on the relationship between bank failure and the independent variables. Finally this study was retrospective in nature. It is suggested that a prospective study predicting or forecasting bank failures in Kenya be carried out to complement the findings of this study and to provide proactive managerial action.