Banking structure and the bank lending channel of monetary policy transmission: evidence from panel data methods

This study examines comprehensively the bank-lending channel of monetary policy for Zambia using a bank-level panel data covering the period Q1 2005 to Q4 2016. Specifically, the study investigates the effects of monetary policy changes on loan supply by commercial as well as the effect of bank-specific factors on response of loan supply to monetary policy shocks. In addition, the study investigates whether the level of bank competition does affect the bank-lending channel. Using a dynamic panel data approaches developed by Arellano-Bond (1991), the results indicate that a bank-lending channel exists in Zambia. In particular, the results show that is loan supply is negatively correlated with policy rate implying that following monetary policy tightening loan supply shrinks. Further, the results indicate that size, liquidity and bank-competiveness have effects on credit supply while capitalization has no effect. Specifically, the results show that bank size has negative effect on credit supply while liquidity and market power are found to enhance credit supply. Most importantly, the results showed that bank-specific factors and bankcompetiveness is responsible for the asymmetrical response of banks to monetary policy. Specifically, the results showed that larger banks, banks with more market power, well-capitalized banks and liquid banks respond less to monetary policy tightening and vice-versa.


Introduction
In the aftermath of the global financial crisis, monetary policy has come to include financial system stability to its traditional objectives of price stability and supporting full employment. To achieve these objectives therefore, it is important for policy makers to have clear understanding of the mechanisms through which their actions affect real variables as well as the factors that influence these channels. This is more so important for developing countries, such as Zambia, where monetary policy is still considered ineffective (Tahir, 2012;Mishra, P., Montiel, P., and Spilimbergo, A., 2010).
Literature reveals that the propagation mechanisms of monetary policy is affected by a number factors (Chileshe and Akanbi, 2016;Mishra et al., 2010). Among factors that influence the proper functioning of the propagating mechanisms, include the structure of the economy, legal and financial structure in a specific country. In this regard, a good understanding of the channels of monetary policy as well as factors that influence them is key to designing tailored monetary policy strategies. In addition, such information is important for designing the necessary reforms needed to improve the effectiveness of monetary policy. Further, the 2007-2010 global financial crisis revealed the importance of the financial sector, especially banks, in monetary policy transmission.
In view of this, there is renewed interest among researchers and policy makers in understanding the role banks play in the transmission of monetary policy.
A number of empirical studies have investigated Zambia's monetary policy transmission (Chileshe et al., 2014;Zgambo and Chileshe, 2015;Mutoti, 2006;Simatele, 2004;Simpasa et al., 2015;Chileshe and Akanbi, 2016). All studies mentioned above, with an exception of Simpasa et al. (2017) as well as Chileshe and Akanbi (2017) uses aggregated time series analysis data to investigate monetary policy transmission. Evidence from these studies suggest that the exchange rate channel is the strongest in Zambia while there is only weak evidence for the bank lending channel (BLC) and no evidence for the interest rate channel. Simpasa (2017) specifically investigates the BLC using bank level data and the effects of bank specific characteristics on monetary policy transmission. This study finds that the BLC is stronger among larger banks than smaller and medium banks. On other hand, Chileshe and Akanbi (2016) investigate the effect of market structure on the interest rate channel. They find evidence in support of the hypothesis that monetary policy transmission is adversely affected by lack of competition in the financial sector.
In this study, we are interested in assessing the bank-lending channel in Zambia. In particular, it answers the following questions: Is there a bank-lending channel in Zambia?
ii) Does market structure affect the effectiveness of the bank-lending channel in Zambia?
iii) Does bank-specific factors such as capitalization, liquidity and size affect the banklending channel?
Although Simpasa et al. (2015) recently provided evidence on the BLC using panel data; however, this study differs in some important aspects. Unlike the study by Simpasa et al. (2015), this paper also investigates the effects of market structure on the bank-lending channel. Specifically, it investigates the effect of bank competition on the bank-lending channel, an issue that has not been tackled using the Zambian data. In addition, this study extends the dataset used by Simpasa et al. (2015) by four years or 16 quarters.
To tackle its objective and effectively answer the research questions, this study is divided into five sections. Section 2.0 provides theoretical and empirical literature on monetary policy transmission whereas section 3.0 overviews Zambia's financial sector and its implications for monetary policy transmission channels. Section 4.0 discusses the empirical methodology, data sources and the variables used in the study while section 5.0 concludes and provides some policy implications.

Literature review
A lot of theoretical and empirical literature exists on the effects of monetary policy and its channels. In this section, theoretical and empirical literature on monetary policy transmission are reviewed. However, the focus of empirical literature review is on studies that focus on the banklending channel using panel data methods.

Interest Rate Channel
According to the interest channel, an increase in money supply leads to a decrease in the real interest rate due to the Keynesian assumption of sticky prices thereby inducing an increase in investment and consumer spending and consequently aggregate demand (Mishkin, 1996). This channel implicitly assumes that the central bank is able to influence long-term real interest rates through manipulation of short-term real interest rates. Mishkin (1996) notes that this suggests the rational expectation hypothesis of the term structure of interest rates holds true. Often in literature, the interest rate channel is referred to as the hallmark of the "Money View".

Credit Channel
The credit channel is not seen as departure from the traditional interest rate channel but an enhancement of it (Buktiwiecz and Ozdogan, 2009). The credit channel explains the impact of monetary policy via the effects of informational asymmetry between the lender and the borrower (Mishkin, 1996). The credit view posits that due to informational asymmetries, two channels of monetary transmission arise: one that operate through the effects on bank lending and another that affect the balance sheet of economic agents. The bank lending channel is based on the assumption that financial intermediaries are best suited to solve problems of informational asymmetry in credit markets while the balance sheet channel is based on the effects of monetary policy on the net worth of firms and hence their collateral (Simatele, 2004). The bank-lending channel operates through the quantity of loans supplied by banks to households (Dabla-Norris and Floerkemeier, 2006).
Expansionary monetary policy increases liquidity in the banking system-enabling banks to supply more loans for investment and consumer spending resulting in increased aggregate demand and consequently economic activity. The bank-lending channel is likely to be more effective in an economy where there are many bank dependent firms with no access to capital markets.
On the other hand, existence of informational asymmetries between borrowers and lenders makes the role played by commercial banks as financial intermediaries to be important and thus comes in the balance sheet channel (Tahir, 2012). Existence of asymmetric information gives rise to moral hazard and adverse selection. There is consensus in literature that banks have a comparative advantage in assessing the balance sheets of borrowers there by mitigating adverse selection and moral hazard (Tahir, 2012;Bernanke and Gertler, 1995). Literature identifies several means through which monetary policy can affect balance sheet of agents. For example, expansionary monetary policy increases the net-worth of agents through increase in stock prices, reduction in debt servicing costs and increased sales. The increase in net-worth reduces the probability of moral hazard and adverse selection thereby enabling them to access loans. Easier access to loans increases borrowing leading to increased consumer spending and investment, and consequently economic activity. It is important to emphasize here that all the other channels operate mostly through the credit channel.

Exchange Rate Channel
The exchange rate channel is one of the primary transmission channels in open economies, especially those with flexible exchange rate regimes. Monetary policy can influence the exchange rate through interest rates (the popular uncovered interest rate parity condition), direct intervention in foreign exchange markets or through inflationary expectations (Dabla-Norris and Floerkemeier, 2006). In this channel, monetary policy affects economic activity (output) through net exports.
For example, expansionary monetary policy, leads to a fall in domestic interest rates relative to the foreign interest rates inducing capital outflows leading to a depreciation of the local currency thereby making exports cheaper resulting in increased net exports and consequently aggregate demand and output (Mishkin (1996(Mishkin ( , 2001).
The strength of the exchange rate channel is affected by several factors such as the exchange rate regime, sensitivity of the interest rates, the size and openness of the economy, degree of capital mobility and the degree of expenditure switching between domestic and imported goods (Boivin et al., 2010;Mishra et al., 2010;Tahir, 2012).

The Asset Price Channel
Monetary policy affects asset prices such as bonds, equity and real estate, changing firms' stock market values and household wealth. Changes in stock market values and household wealth in turn affect aggregate demand. The asset price channel of monetary policy transmission is assumed to operate through two mechanisms, namely: the Tobin's (1969) Q-theory of investment and life cycle theory of consumption (Ando and Modigliani, 1963). The Tobin's Q theory works as follows, expansionary monetary policy increases the demand for equities (either by the Keynesian or by Monetarist argument), raising equity prices and thereby boost market value of firms relative to the replacement cost of capital. This will result in increased investment and therefore output.
Furthermore, higher equity prices also raise the net-worth of firms and households and hence improve their credit worthiness and access to funds, the effects of which would partly reflect the balance sheet channel of monetary policy (Afandi, 2005). On the other hand, in the life cycle model of consumption monetary policy changes affect economic agents' long-term wealth and therefore, alter their consumption pattern. The basic premise of Ando-Modigliani theory is that consumers smooth out their consumption over time and this consumption depends on lifetime resources and not only current consumption (Mishkin, 1996). Expansionary monetary policy, which lowers interest rates, changes consumers' portfolio composition in accordance with the risk of each asset class. In this case, a decrease in the interest rates encourages people to reduce their holding of interest earning deposits and bonds and substitute them with equity/stocks, thereby increasing stock prices (Afandi, 2005). Given that a major component of wealth is in common stocks, the increase in stock prices increases their wealth resulting in higher consumption expenditure and hence output.
Tahir (2012)  shallower and uncompetitive in addition to high levels of macroeconomic instability.

The Expectations Channel
In modern macroeconomic theory, there is a consensus that expectations play a key role in shaping the behavior of economic agents. Although there is a consensus on the importance of expectations, economists differ on how these are generated. Friedman and other monetarists, postulate adaptive expectations while the new classical school lead by Lucas and the New Keynesian School argue for rational expectations.
Since economic agents are forward looking and rational, the expectation channel is in effect fundamental to the working of all channels of monetary policy transmission. Theoretically, this channel is mainly operational in developed economies with well-functioning and deep financial markets (Davoodi et al., 2013). For example, if economic agents expect future changes in the policy rate, this can immediately affect medium and long-term interest rates. Further, monetary policy can influence expectations of future inflation and thus price developments. Inflation expectations matter in two important areas. First, they influence the level of the real interest rate and thus determine the impact of any specific nominal interest rate. Second, they influence price and money wage-setting behavior and feed through into actual inflation in subsequent periods.
Similarly, changes in the monetary policy stance can influence expectations about the future course of real economic activities by affecting inflationary expectations and the ex-ante real interest rate and guiding the future course of economic activities.

Empirical Literature
The overall objective of this study is to assess the bank-lending channel (BLC) in Zambia, especially the effect of bank specific characteristics on monetary policy transmission. Therefore, in this subsection we provide empirical evidence on the BLC from across the global.

a) Studies from the USA
The difficulty associated with empirically identifying the BLC from aggregate data prompted researchers to turn to disaggregated panel data. Kashyap and Stein (1995) pioneered work on the BLC using a panel data from the USA for the period 1973 to 1991. Kashyap and Stein (1995) examines the effect of monetary policy tightening on the volume of loans provided by banks. In particular, they tested the hypothesis that smaller banks respond more to monetary policy than larger banks. Their results indicate that following monetary policy tightening the volume of loans decline for small banks but not for larger ones. However, their study suffers from a limitation, as it does not take into account the effect liquidity holdings as a buffer to monetary policy shocks.
In response to the limitation, Kashyap and Stein (2000) controls for differences in liquidity across banks. They analyse the differences in bank response to monetary policy controlling for capital and liquidity using panel data for the period 1976 to 1993. Their results suggests that monetary policy has a strong impact on bank lending on banks with less liquid balance sheets-i.e banks with less securities to asset ratio. This response was also found to be stronger among smaller banks.
Although their results support the existence of a 'bank lending channel', they are unable to provide statements about its quantitative importance. Kishan and Opiela (2000) explores the credit channel of monetary policy transmission focussing more on the BLC using data from 1980 to 1995. Specifically, they test for the shifts in loan supply by segregating according to asset size and capital leverage ratio. In the study, monetary policy is represented a federal funds rate and the Bernanke-Mihov (1995) indicator. In general, their empirical results suggest that monetary policy has larger effects on small under-capitalised banks.
In a later study, Kishan and Opiela (2006) extends the data to 1999 while allowing for both expansionary and contractionary monetary policy. Using only the federal funds rate as a measure of monetary policy, they find that contractionary monetary adversely affect loan by undercapitalised banks but expansionary policy has no effect on loan supply.
In a recent study, Olivero, Li and Jeon (2011) investigates the effects of bank consolidation on the bank lending channel using bank-specific panel data from Asian and Latin American countries for the period 1996 to 2006. In addition, they also investigate the role bank-specific factors play in the relationship between the BLC and consolidation. Their results indicate that bank consolidation weakens the bank-lending channel resulting in less effective monetary policy. Furthermore, they find that the adverse effects of bank consolidation on the bank-lending channel are most conspicuous on small banks. Hence, they argue that from a policy perspective it is important for the authorities to take keen interest of consolidation efforts in the banking sector to arrest the adverse effects on monetary policy effectiveness.
Another recent study was by Gambarcota and Marquez-Ibanez (2011) who uses bank specific panel data for the period Q1 1999 to Q4 2009 from Europe and USA. The objective of the study was to investigate the effects of bank specific characteristics on the bank-lending channel during the financial crisis. Using dynamic panel data methods for over 1000 banks, they find that banks business models had significant impact on the supply of credit or the bank-lending channel.
Specifically, they find that short-term funding and securitisation activity have significantly changed the way banks react to monetary policy shocks. In addition, they find that banks with a higher level of non-interest income sources tend to be highly constrained during a financial crisis.

b) Studies from Europe
Comprehensive studies of the BLC on European countries started to appear towards the end of the 1990s. One of the earliest studies was undertaken by Bondt (1998) who investigates the credit channels in some European countries focussing on the role of bank specific factors. The study uses short-term interest rates and a monetary conditions index as measures of monetary policy stance.
Using a panel data covering the period 1990-1995, empirical results provides evidence for the existence of the credit channel in Europe. Specifically, the bank-lending channel is found to be strong in German, Belgium and Netherlands, while the balance sheet channel is strong in German and Italy. In the UK, the BLC is non-existent or dominated by loan and deposit demand factors of bank customers. Brismiss, Kamberoglou, and Simigiannis (2003) empirically analyses the role that bank lending has in monetary policy transmission using bank level data from Greece. They utilise two approaches: one in the spirit of Kashyap-Stein (1995)  Blinder yields results that are more satisfactory. In general, they find the existence of a BLC. In addition, bank-specific characteristics were found to systematically shift the loan supply function.
The results showed that large banks could to a certain extent; shield their loan portfolio from monetary policy changes. Similar results hold for the liquid (healthy) banks.
Another country specific study by Ignacio and Martinez-Pages (2001) tests the existence of the bank-lending channel in Spain using a panel data for the period 1991 to 1998. Their empirical results suggest that there is no evidence for the existence of the bank-lending channel in Spain.
A study by Altunbas, Fazylov and Molyneux (2002), examined the evidence of a bank lending channel in Europe using panel data for the 1991 to 1999. The authors classify banks according to size and capitalisation to gauge whether these factors have an impact on bank lending. Using changes in short term interest rates as a measure of monetary policy and approach suggested by Kishan and Opiela (2000) they find that undercapitalised banks (of any size) respond more to monetary policy change. Further, their results suggest that there is little evidence that smaller undercapitalised are the conduit of the bank lending. However, they find that the bank-lending channel is much stronger in smaller countries than larger ones. Hence, they conclude by implication that undercapitalised banks operating in smaller countries tend to respond more to monetary policy changes. (2004) investigates the cross-sectional differences in the response of banks to monetary policy and GDP shocks owing to their differences in capitalisation using panel data of Italian banks. Their empirical strategy follows in the Kashyap-Stein (1995) spirit and estimated using the GMM estimator suggested by Arellano and Bond (1991). Their empirical results indicate that well-capitalised banks can better shield their lending from monetary policy shocks because they have easier access to uninsured fund raising. In addition, they find that 'The Bank Capital Channel' has a stronger effect on smaller banks whose balance sheets contain a larger maturity mismatch between assets and liabilities. In addition, they also find that capital influences the way banks react to adverse GDP shocks.

Gambarcota and Mistrulli
A paper by Benkovskis (2008), explores the role of commercial banks in the transmission of monetary policy in Latvia using the panel data method developed by Kashyap and Stein (1995).
Specifically, they estimate a bank loan function that takes into account monetary policy and macroeconomic variables but also bank-specific factors. Their empirical results suggest that there is heterogeneity in the response of banks to monetary policy. In particular, they find that some banks have a statistically significant negative reaction to monetary policy while the others do not.
Further, they find that monetary policy has an effect on small bank, domestically owned, with lower liquidity and lower capitalisation.

c) Studies from Africa
Studies on the BLC using bank level data from research have grown in recent years as more balance sheet data for commercial banks become available. Boughrara and Ghazouani (2009) investigates whether there are differential effects of monetary policy across bank size, capitalisation and liquidity in MENA countries of Jordan, Egypt, Morocco and Tunisia.
Specifically they test for the presence of BLC in MENA countries. Using panel data methods, they find that results are heterogeneous among the MENA countries. Further, they find that less capitalised banks respond more to monetary policy than more capitalised banks in all countries.
Size and liquidity are significant in shaping the response of banks to monetary policy shocks.

d) Studies from Zambia
The review of empirical literature on Zambia reveals very few studies on monetary policy transmission that uses panel data approaches (Chileshe and Akanbi, 2016;Simpasa et al., 2015).
The paper by Chileshe and Akanbi (2016) investigates the effects of bank market power as well as bank-specific factors such as bank size and capitalisation on the interest channel using a quarterly panel data of Zambian banks between Q1 1998 to Q2 2015. Results from the dynamic panel data estimation suggests that competition enhances the interest rate channel. In addition, there is significant positive relationship between the HHI or Lerner Index and the interest rates in both the short-and long-run. However, capitalisation and liquidity have significant positive and negative effect on lending rates in the short-run only, respectively. Simpasa et al. (2015) undertook study similar to this effort by investigating the BLC using bank level data. Specifically, they explore the effect of monetary policy on lending behaviour of commercial banks as well as the effect of bank specific factors in determining the response of banks to monetary policy changes. Using approaches similar to the Kashyap-Stein (1995); they find evidence that the BLC occurs through larger banks while it is moderate in medium sized banks and non-existent in smaller banks. Furthermore, they find price signals rather than quantities are more important in the transmission of monetary thereby supporting the transition of the Bank of Zambia to interest targeting. They also conclude that monetary policy changes will have the desired results if it affects the largest banks in the sample. Although this effort is similar to ours in every effort, it differs in two significant ways. First, this study includes a measure of bank competition to capture its effects on the BLC. Second, Simpasa et al. (2015) exclusively uses exchange rate volatility as an indicator of economic performance. However, an exchange rate is a very information sensitive variable and its volatility may be due to speculative behaviour among players in the foreign exchange market. Hence, in addition to exchange rate volatility the GDP growth is added as a measure of the state of the economy. Finally, this effort extends the data by four years or sixteen quarters making the analysis more rigorous.

Stylized Facts about Zambia's Financial Sector
The financial system in Zambia comprises commercial and non-bank sector regulated by Bank of Zambia (BOZ), pensions and insurance regulated by the Pension and Insurance Authority (PIA), a Stock exchange regulated by the Securities and Exchange Commission (SEC). In terms of size, the banking sector dominates the financial sector in Zambia holding over 70% of total assets over the years (See Table 1). Given its significant size in the financial system, the banking system remains the large source of credit in Zambia averaging over 80% since 2011 followed by non-banks at 12.9%. In essence, the share of credit by institutions under the regulatory control of Bank of Zambia is 98.7% as at 2015.
Although this could imply that the bank-lending channel might be stronger it could be hampered by the small size of the population with access to bank credit at only 8.2%, which imply very few individuals are affected by monetary policy changes. Further, although Zambia's capital market has shown growth over the years it remains underdeveloped with only 22 companies listed while market capitalization has been below 25% of GDP (see Table 2 and Figure 1). Table 2 shows that stock market capitalization, as a percentage of GDP at 20.9% is far lower than the sub-Saharan Africa average (46%) as well as the LICs average of 43% as at 2012. This clearly shows that the stock market though growing is still in its infancy. Compared to other developing economies, Table 3 shows that Zambia's financial sector is similar to those of its peers. Specifically, it shows that total deposits and credit in the financial sector is about 28.7% and 25% of the GDP, respectively. Further, although Zambia's access to formal financial institutions by its adult population is better than its peers it remains low at 37.2% while 0 5 10 15 20 the use of formal and informal credit by the adult population is only 8.2% and 15.8% respectively.
It is clear from Table 4 that Zambia's financial sector performance is on average similar to other LICs and SSA regional average.  Massarongo (2012) Furthermore, Zambia's securities market is in its formative stage despite having been in existence since 1995. Available data indicates that there are 10 public bonds and treasury bills outstanding with a market value of approximate of ZMW 19.5 billion or US $ 1.9 billion as at the end of 2015, which is 14.7% of GDP (see Figure 5.11 below). Furthermore, the stock market capitalisation is about 20.9% of GDP as at 2011, which is below the average for SSA, and LICs mean (Table 3 above). However, access to financial services has increased to over 60% in 2015. Furthermore, there were only 14.6% and 13.9% with bank accounts in 2005 and 2009, respectively (see Table 4    The current structure of Zambia's financial system, especially the banking system is the culmination of financial sector reforms, which have been undertaken since 1992. The early years of financial sector reforms saw a progressive entry of new banks in the sector reaching 19 in 1995 from only 12 banks in 1989 (Brownridge, 1996). However, this positive result was reversed during the mid-1990s banking crisis, which saw the collapse of more than eight banks (Simpasa et al., 2017). The banking sector crisis of the mid-1990s created a scope for more prudential reforms that resulted in a more robust regulatory framework (GRZ, 2004;Simpasa et al., 2017).
The economic and structural reforms that have been implemented over the years have been cardinal in attracting foreign banks into the sector. As at the end of 2012, there were a total of 13 foreignowned banks, four privately owned banks, and 2 banks with a government stake. In an effort to improve efficiency, the Government of the Republic of Zambia partially privatised the Zambia National Commercial Bank, the largest domestic bank, through offering a stake to a foreign bank (Rabobank) and offloading part of its shareholding on the Lusaka Stock Exchange to the private sector. Although government and other Zambian shareholders have a majority stake in the bank, management rights are with foreign shareholders (45%). In totality, there are 19 commercial banks operating in Zambia with a combined asset size equivalent to 30% of GDP (Simpasa et al., 2015).
Although there has been an increase in the number of commercial banks in Zambia's banking sector, it remains highly concentrated. The 4-firm concentration, a basic measure of market structure indicates that Zambia's banking sector is not competitive as it shows that four largest banks control nearly two-thirds of all market segments (see Table 5 BOZ database, 1998BOZ database, -2012 Since 1998, a reflection in the level of competition has been most evident in the loans segment followed by the deposits while the assets segment and Bond/securities segment has remained static over the last ten years. Similarly, banking activities have increased over time, as depicted by the composition of the banks' consolidated balance sheet (see Table 5 above). Furthermore, there has been an increase in the banking activities in the country as indicated by the consolidated balance sheet which has shown that the level of assets and liabilities have increased from only 1.5 billion kwacha in 1998 to over 49.6 billion in 2014 (See Table 6 below).  less than 21% of the banking system's total assets against approximately 34% in loans. By the end of 2010, the share of liquid assets in total assets had reached 43.1% while the total loans and advances accounted for 38%. However, by the end of 2012 the share of liquid assets dropped to 36% while loans and advances grew to 47.1%. At end 2015, total loans and advances fell to 43.7%. The rebound in the share of loans follows banks' increased exposure to the private sector following marked improvements in macroeconomic performance, underpinned by low inflation and strong economic growth, especially after the 2007-2010 global economic crisis (Simpasa et al., 2015).
Although there has been a sustained increase in the level of credit extended to the private sector, it remains low even by regional standards. The increase in credit extended to the private sector could be due to a fall in yield rates on government securities. This low ratio of private credit extended to the private could essentially affect the effectiveness of monetary policy, especially the credit and interest rate channels.
In terms of foreign assets, Table 6 shows that before 2006 commercial banks acquired large amounts of foreign assets to hedge against high inflation and a rapidly depreciating domestic currency. However, as conditions have improved, the proportion of claims on foreign financial institutions in total assets by Zambian banks has significantly decreased, reaching 16% in 2006 and then 10% in 2010 and then edging slightly upwards to 12.5% in 2012.
On the liabilities side, deposits account for more than two-thirds of the banks' sources of funds.
Although the bulk of deposits are attributed to the private sector, some large banks also hold substantial amounts of government deposits, which provide a buffer against swings in private sector deposits. On the other hand, shareholders' capital has remained relatively stable over the years, roughly around 9-10% of total liabilities. This level of capitalisation reflects the robustness of the regulatory framework instituted in the aftermath of systemic bank failures in the mid-1990s.
In the continued effort to strengthen the banking sector and improve its resilience to external shocks, the Bank of Zambia (BoZ) increased regulatory capital further in April 2012 and introduced a tiered structure (GRZ, 2012). The minimum capital requirement for local banks was set at K104 billion (US$20 million) while that for foreign banks was set at K520 billion (US$100 million). Before the revision, minimum capital for all banks was K12 billion (approximately US$2 million). It is expected that the new capital requirement will be expected to attract additional resources into the industry and encourage lending to the private sector.

Methodology
In this section, we present the empirical approaches utilized in arriving at the conclusions of the study. In particular, we present the econometric model as well as the specification tests that were utilized. In the last sub-section, the study discusses the dataset used as well as describing the variables in the model.

Econometric Model
This study utilizes a modified empirical specification first employed by Kashyap and Stein (1995) and used in many more empirical works (Gambarcota and Marquez-Ibanez, 2012;Gambarcota and Mistrulli, 2004;Boughrara and Ghazouani, 2009;Opolot, 2013, andSimpasa et al., 2015). In particular, the study modified the Gambarcota and Mistrulli (2004) model in one important way.
In addition to Bank size, capitalisation and liquidity, it also assesses the effect of banks market power on the BLC. This is because market power give banks the ability to set the price of their service well above marginal cost without risking their market share. Specifically, the following empirical model is estimated: Where = 0,1 … , = 1,2, … and = 1,2, … , with m representing the number of lags, N is the total number of banks while T is number of time series observations. is the total loans issued by bank in period to the non-bank sector while represents bank specific variables, namely capitalisation, market power, liquidity and size. , ∅, , , , , , , are the parameters to be estimated, with capturing bank-specific fixed effects while is the white noise error term.
The regression above also includes (in addition to bank-specific variables) exchange rate volatility as in Simpasa et al. (2015) and GDP growth to account for demand effects. These two variables are introduced to allow the model to capture cyclical movements as well as to isolate the effects of monetary policy changes. Hence, they will enable us to gain further insights on the lending channel by reporting the effects of changes in the short-term interest rates changes on other balance items (Gambarcota, 2005). Finally, the non-performing loan are included in the model to account for their cost implications on commercial banks when they need to make loan loss provisions. In

Estimation strategy
The presence of the lagged values of the dependent variable on the right hand side could imply that the error term is correlated with the independent variables thereby violating one key assumption of the Ordinary Least Squares approach (OLS). In other words, estimating both a dynamic panel data model using fixed or random effects could produce biased and inconsistent results. In addition, estimating equation (1) above the endogeneity problem could arise because monetary policy decisions maybe affected by the conditions in the financial sector. To overcome these problems, the Generalized Method of Moments (GMM), proposed by Arellano and Bond (1991) and later refined by Blundell and Bond (1998) is used.
In the GMM literature, it is proposed that the estimation be carried out in first differences in order to get rid of the individual effects as well as using instruments to help in obtaining unbiased and consistent estimates (Benkovkis, 2008). Utilising the orthogonality conditions between lagged values of the dependent variable and disturbances, lagged values of the dependent variable with second and more lags serve as instruments. Further, to deal with the endogeneity problem two approaches have been proposed. First, right-hand side variables enter the model with at-least one lag. Second, lagged levels of predetermined variables such as bank-specific variables are used as instruments. Finally, the strictly exogenous variables such as GDP in first differences are instrumented by higher lags of themselves.

Instrument Selection Tests
To check whether the instruments where chosen properly and the assumptions underlying the model hold, a few tests were proposed (Arellano and Bond, 1991). Consistency of our estimators relies on the fact that the disturbances follow MA (1) process and there is no second order autocorrelation of disturbances. Hence, we use AR (1) and AR (2) (2001), the bank-specific factors are used to proxy potential asymmetric information problems leading to a differential response across banks to a common monetary policy shock.
Measures of size, liquidity, capitalisation and market power (Lerner Index) are calculated as follows: Like in much of the Literature, is the deviation of total assets for bank at time t expressed in natural logs from industry average in the same period. This normalization is done to ensure that the ever-increasing trend in total assets is removed over the sample period thereby making the series stationary (Ehrmann et al., 2001;Simpasa et al., 2015). The bank's liquidity is calculated as a sum of cash, securities holdings and funds at the central bank. The variable LIQR is calculated as the deviation of the ratio of total liquid assets to total asset from the industry average ratio over the sample period. Further, the CAPR is the deviation of the capital to asset ratio from the industry average over a period. Finally, MKTP is the market power of a bank, which represents the ability of a bank to charge a price above the marginal cost. As indicated in sub-section 4.2.1 above, the values of the Lerner Index are obtained from a study by Chileshe and Akanbi (2016) which is more comprehensive.
Macroeconomic performance indicators are included to control for loan demand effects. This is because a fall in credit growth following monetary policy tightening could be due to a fall demand for credit by borrowers (Ehrmann et al., 2001). To capture loan demand effects, the model includes GDP growth and quarterly exchange rate volatility.
It is generally agreed in the literature that developing and emerging economies, there is high-level asymmetry of information leading to moral hazard and adverse selection. In Zambia, for example it is common for banks to profile their clients according to their characteristics when granting loans (Simpasa et al., 2015). Hence, in addition to bank-specific characteristics indicated above we also include a risk variable to capture the effects of credit rationing behaviour among the banks. To capture credit-rationing behaviour, we introduce the non-performing loans (NPLs) as a proxy for credit risk. In

Empirical Results and Discussions
In this section, estimation results of the model presented in equation 1 above are presented. In total, there are five models estimated in the first four models incorporate only one bank-specific characteristic is included while the model in the last column all the variables are incorporated.
Results presented in show that the effects of macroeconomic variables are significant and robust across all the models.
With an exception of the full model, the results indicate that although monetary policy has the expected negative contemporaneous effect on bank lending it is not significant. However, the lagged monetary policy variable has a significant negative effect on credit growth. Specifically, a 100 basis points rise in the policy leads to a fall in credit growth of between 16-25 basis points after one quarter. The lagged relationship between the monetary policy variable and credit growth is expected in the sense that tight or loose monetary policy would first affect the short-term rates before which in turn affect credit rates and consequently growth in credit. In addition, banks may not adjust their lending activities immediately as they may have already negotiated loan agreements, which they have to honour. Furthermore, in all models the loan demand effects represented by GDP and exchange rate volatility are significant. Specifically, the estimation results indicate that growth of GDP has a positive effect on growth indicating a positive elasticity of total loans growth as expected. On the other hand, increased exchange rate volatility has a negative effect on credit growth, which could mean that higher exchange rate volatility could signal an increase in credit risk, especially foreign exchange denominated loans.
Looking at the bank-specific characteristics, results show that bank size, liquidity and market power have significant effects while capitalization has insignificant effects on loan growth. These results imply that size, liquidity and market power are responsible for the asymmetrical response of banks to monetary policy changes in Zambia. Specifically, the coefficient of the bank-size measure is negative and significant in the full model and the model that only incorporates a measure of bank size. This result is similar to those obtained by Simpasa et al. (2015), and Heryan and Tzemeres (2016). This finding is consistent with the argument that small banks that have just started their operations tend to growth their loan books faster than larger well-established banks in order to gain market share. As for liquidity, the results show a positive significant effect on loan growth in Zambia. The positive sign of the coefficient on liquidity is expected in the sense that banks with higher liquidity levels have the ability to extend more credit than banks with lower liquidity levels. In addition, this result could be because of existing regulatory rules in the banking system such as the Basel Committee Rules, which may stipulate that more liquid banks lend their financial resources mostly through bank loans. Finally, the coefficient for market power is positive and significant implying that banks that have more market will supply more loans than those with less market power.
Furthermore, to assess the asymmetrical effects of monetary policy on loan supply, there is need to examine the coefficients of the interaction terms between the bank-specific characteristics and the monetary policy variable (Heryan and Tzemeres, 2016;and Ignacio and Martinez-Pages, 2001). It is argued that significant interaction terms would imply that banks adjust their loan supply heterogeneously in response to monetary policy changes. In case of bank size, the results indicate that the interaction terms between size and monetary policy variable is not significant in the model in which it enters alone but it is significant in the full model. The significant positive interaction term of bank-size and monetary policy variable implies that larger banks are likely to respond less to monetary policy shocks than smaller banks. This result is consistent with theoretical literature on the bank-lending channel, which assumes that lending volume by larger banks are less sensitive to monetary policy shocks than that of small banks. Furthermore, this result is consistent with Kashyap and Stein (1995) prediction that lending volume of smaller banks is more sensitive to monetary policy changes than that of larger banks as well as findings by Simpasa et al. (2015) on Zambia. However, the coefficient of the interaction is close to zero, which could be consistent with the hypothesis that bank-size has no differential effect on credit growth response to monetary policy similar to findings by Ehrmann et al. (2003) among European Countries.
With regard to the interaction term of capitalization and monetary policy indicator, results in table   8 show that it is not significant in the model in which it enters alone while it is significant in the full model. The interaction term is negative implying that banks in good shape in terms of their capital respond to monetary policy more than those, which are in bad shape. These results are similar to those obtained by Heryan et al. (2016) but at variant with those obtained by Gambacorta and Mistrulli (2004). This result could be explained in the sense that well capitalized banks are likely to avoid increasing their exposure to risk following monetary policy tightening which increase the likelihood of loan default compared to undercapitalized banks.
Furthermore, the interaction term of liquidity measure and monetary policy indicator is significant and positive in both the model with only liquidity and the model including all bank specific characteristics. Precisely, this result imply that the loan response of banks with lower liquid assets share is significantly stronger than that of more liquid banks. In other words, banks with higher liquid assets are better able to protect their lending activities from changes in monetary policy consistent with results by Ehrmann et al. (2003), Heryan et al. (2016) and Gambacorta (2005). The underlying reason is that banks with a more liquid balance sheet can use their liquid assets to maintain their loan portfolio and as such are affected less heavily by a monetary policy tightening (Erhmann et al., 2003).  One-step AB One-step AB One-step AB One-step AB One-step AB (1) -7.775*** To capture the effects of bank-specific market power on the bank-lending channel, the model includes an interaction term of the Lerner Index and monetary policy indicator. The coefficient of the interaction term is positive in all cases but significant in the model that only includes the market power to explain asymmetrical response to monetary policy changes by commercial banks. The positive term imply that banks with more market power are better able to shield their lending activities from monetary policy changes than banks with less market power. This result is consistent with findings by other studies (Chileshe and Akanbi, 2016;Li and Lee, 2015). The reason for this could that in the absence of vigorous competition among banks; banks find it relatively easy to access alternative sources of liquidity to finance their lending activities such as certificate of deposits and interbank loans. In other words, banks having less market power find it more difficult to access alternative sources of funding. Hence, high-level competition means that changes in monetary policy influence the availability of funds so that changes in monetary policy will more directly influence the supply of bank loans.

Conclusions and Policy Implications
In this study, an attempt was made to provide evidence for the existence of the bank-lending channel in Zambia using a panel data covering the period Q1 2005 to Q4 2016. Specifically, the study attempted to examine the response of bank loans to monetary policy changes. In addition, the study attempts to evaluate the effects of bank specific factors of capitalization, size, liquidity and market power on the bank-lending channel. In other words, it attempted to provide evidence for the source of asymmetrical response of banks to monetary policy changes.
The results obtained in this study show that there is a bank-lending channel in Zambia. The results generally show that loan growth is negatively related to monetary policy implying that following monetary policy tightening loan supply generally falls. However, the size of the coefficients are smaller compared to those from more developed countries.
To analyze the asymmetrical responses of banks to monetary policy changes, the study followed a procedure prescribed by Kashyap and Stein (1995). The results generally show that size, liquidity and market power have effects on the supply of credit by commercial banks while bank capitalization is found to have no effect. Most importantly, the results in this study showed evidence to support the argument that bank specific factors of size, capitalization, liquidity and market power explains the asymmetrical response of banks to monetary policy changes.
Specifically, the results show that bank size has a negative effect on credit or loan supply, which is consistent with argument that smaller banks tend to lend more in order to, gain more market share. Liquidity and market power, on the other hand are found to enhance the supply of credit or loan supply by commercial banks. Concerning asymmetry, results showed that larger banks, banks with more market power, well capitalized and liquid respond less to monetary policy tightening and vice versa.