macroeconomic uncertainty and bank lending: the case of ukraine

Economic Systems 36 (2012) 279–293

Contents lists available at SciVerse ScienceDirect

Economic Systems

journal homepage: www.elsevier.com/locate/ecosys

Macroeconomic uncertainty and bank lending:The case of Ukraine

Oleksandr Talavera a,*, Andriy Tsapin b, Oleksandr Zholud c

a University of East Anglia, Norwich NR4 7TJ, United Kingdomb National University of Ostroh Academy, Ukrainec International Center for Policy Studies, Ukraine

A R T I C L E I N F O

Article history:

Received 29 June 2010

Received in revised form 18 June 2011

Accepted 19 June 2011

JEL classification:

G21

G28

P34

Keywords:

Banks

Macroeconomic uncertainty

Ukraine

Banks’ balance sheets

A B S T R A C T

This study investigates the link between bank lending behavior and

country-level instability. Our dynamic model of bank’s profit

maximization predicts a non-monotonic relationship between bank

lending and macroeconomic uncertainty. We test this proposition

using a panel of Ukrainian banks over the 2003Q2–2008Q2 period.

The estimates indicate that banks decrease their lending ratio in

times of substantial economic volatility, which could be explained

by higher risk aversion of bank managers. Additionally, small and

least profitable banks are less likely to be affected by changes in the

macroeconomic environment compared to their large and most

profitable peers. This outcome is robust with respect to the different

measurements of macroeconomic uncertainty.

� 2011 Elsevier B.V. All rights reserved.

1. Introduction

The determinants of capital structure have attracted considerable attention in the financial andeconomic literature. In a seminal paper, Modigliani and Miller (1958) argue that funds are alwaysavailable for positive net present value investment projects and firm value and financial structure areindependent. Internal and external finance can be viewed as perfect substitutes in a world with perfectcapital markets and without information asymmetries, transaction costs, or taxes. However, the realworld is imperfect and the determination of optimal capital structure is often subject to financialfrictions in the banking sector, also called the ‘‘arterial system of the economy.’’ Banks play anintermediary role by financing relatively illiquid assets, such as long-term commercial loans, from

* Corresponding author. Tel.: +44 1603 593415.

E-mail address: [email protected] (O. Talavera).

0939-3625/$ – see front matter � 2011 Elsevier B.V. All rights reserved.

doi:10.1016/j.ecosys.2011.06.005

http://dx.doi.org/10.1016/j.ecosys.2011.06.005

mailto:[email protected]

http://www.sciencedirect.com/science/journal/09393625

http://dx.doi.org/10.1016/j.ecosys.2011.06.005

O. Talavera et al. / Economic Systems 36 (2012) 279–293280

more liquid liabilities, such as short-term deposits. Hence, the availability of bank loans may haveimportant effects on fixed capital investment, and, consequently, on economic growth (Rousseau andYilmazkuday, 2009).

The level of uncertainty about inflation or money growth has a direct impact on the risk in financialmarkets (Krkoska and Teksoz, 2009) and the banking system in particular. Naturally, there is anemerging body of theoretical and empirical literature focused on banks’ behavior under a constantlychanging macroeconomic environment. This research is motivated by the fact that changes inmacroeconomic uncertainty, partially influenced by monetary policy, affect firms’ costs of obtainingexternal finance and their investment dynamics. In this paper, we explore the relationship betweensupply of financing and the variations in the macroeconomic environment. Specifically, we askwhether banks change their lending behavior in response to changes in macroeconomic uncertainty.

Several papers have analyzed the interaction between the macroeconomic environment and bankasset management. Jimborean (2009) investigates the effects of monetary policy on bank lendingchannels for 10 Central and Eastern European countries. The author finds evidence of functioning banklending channels, but only through small banks. Stein (1998) develops a model of bank asset andliability management and concludes that monetary policy affects bond-market interest rates onlybecause of imperfections in the banking sector. Kashyap and Stein (2000) show that the impact ofmonetary policy on bank lending behavior is particularly strong for small American firms with lessliquid balance sheets. Among other macroeconomic environment factors, uncertainty is a substantialfactor affecting bank capital structure. Baum et al. (2003) suggest that macroeconomic uncertaintyplays an important role in explaining banks’ lending decisions. They find that growth of total loans hasa positive relationship with uncertainty proxies. Besides growth of loans a number of papersinvestigate whether ‘‘second moments matter’’ by considering the relationship between the varianceof lending and macroeconomic uncertainty (Baum et al., 2009; Calmes and Salazar, 2006). It is foundthat when macroeconomic uncertainty increases, the cross-sectional volatility decreases, whichsuggests a herding behavior in the banking industry. Quagliariello (2009) also finds that not onlyaggregate shocks but also idiosyncratice shocks lead to herding.

None of these papers addresses the issue examined here, namely the relationship between theasset structure of banks and macroeconomic volatility.1 Therefore, the contribution of our paper istwofold. First, we develop a theoretical framework that provides predictions about links between banklending and macroeconomic uncertainty. Second, the model predictions are empirically tested byapplying the GMM estimator on a panel of Ukrainian banks.

Our work extends the theoretical models of Abel (1983) and Hahm (1996). According to our setup,bank managers choose optimal levels of labor and deposits from business agents to maximize thebank’s value, equal to an expected present value of profit. The banks face convex costs of borrowingadjustment and are irresolute about the price on credit resources because of future demand shocks.The model predicts a non-monotonic relationship between bank lending and macroeconomicuncertainty. However, due to the specific features of underdeveloped financial markets (e.g. lowinterest elasticity of the demand for loans),2 we predict a negative effect of loan demand volatility onthe lending decisions of banks.

To test the model’s predictions, we apply the System GMM estimator (Blundell and Bond, 1998) to apanel of Ukrainian banks over the 2003Q2–2008Q2 period. These data are hand-collected from theofficial monthly newsletter of the National Bank of Ukraine (NBU) ‘Visnyk NBU’, which provides in-depth information on the structure of banks’ assets, liabilities and capital. After screening proceduresour data include 2777 quarterly bank observations. Since the impact of uncertainty may differ acrosscategories of banks, we also consider splits of the sample on large and small banks as well as on mostand least profitable banks.

Our main empirical findings can be summarized as follows. We find strong evidence of a negativeassociation between the optimal level of bank lending and six out of eight measures of macroeconomicuncertainty. This outcome could be explained by precautionary behavior of bank managers in moreuncertain times. There are also differences in sensitivity of lending with respect to macroeconomic

1 We use the terms uncertainty and volatility interchangeably.2 See, for example, Claeys and Vander Vennet (2008), Bhaumik and Piesse (2008), and Weill (2011).

O. Talavera et al. / Economic Systems 36 (2012) 279–293 281

volatility among banks’ size and profitability subsamples. It is shown that small (least profitable)banks are notably less affected by uncertainty compared to larger (most profitable) counterparts.

These results provide useful insights into financing fixed capital decisions. Inconstancy of themacroeconomic environment affects the supply of external financing, which influences firms’investment decisions. Therefore, our estimates suggest that the transmission mechanism of monetarypolicy is much more complicated than formulated in standard models which ignore the interaction ofbank asset structure and macroeconomic stability.

The rest of this paper is organized as follows. The next section provides an overview of theUkrainian banking sector. Section 3 presents the theoretical framework. Section 4 describes the dataand illustrates econometric results. Finally, Section 5 concludes.

2. Review of the Ukrainian banking sector

The origin of the present Ukrainian banking system dates back to the pre-independence period.3

During Gorbachev’s perestrojka times, the Communist Party of the Soviet Union initiated economicreforms which affected all sectors of the economy and the banking sector in particular. Prior to thereform there were only four banks in the Soviet Union: State Bank (Gosbank), Construction Bank(Strojbank), Saving Bank (Gostrudsberkassy), and Exportbank (Vneshtorgbank). The role of banks inthat system was limited as there was no real difference between credits and subsidies. State Bankoften financed inefficient and unprofitable state programs. The modern Ukrainian banking system wasonly born in 1991 when the law ‘‘On Banks and Banking’’ was adopted. Since then the number of banksincreased dramatically from 76 during the first year of Ukrainian independence to 230 in 1995. Thissignificant increase could be explained by the low barriers to entry in those days. However, in thefollowing years the churn rate was fairly high until the banking sector gained more stability in theearly years of the new century.

Starting from January 1998, the banking system of Ukraine has been transferring to theinternational accounting and statistics standards. However, all International Financial ReportingStandards (IFRS) have not been approved so far. Unlike the rest of the economy, the banking systemgenerally adopted international accounting standards, thus this sector can be viewed as closest to thestandards of developed countries. Moreover, during 2001–2008, the NBU implemented severalsignificant actions to enhance the lending capacity of the banking system. In particular, the NBU raisedthe minimum capital adequacy ratio, tightened the requirements to lending regulations andintroduced a new risk assessment methodology. The strong growth and perspectives of furtherexpansion ‘‘on the heels’’ of Central European countries led to the subsequent purchase of several largeand medium banks by foreign investors in 2005–2007.

In order to receive additional funds for credit expansion, banks actively borrowed abroad in the2000s. In December 2003, Ukrainian banks started tapping international capital markets withEurobonds and syndicated credits. According to the NBU’s data on external debt, the banking systems’external indebtedness as of end-2003 was USD 1.746bn, while at the end of the second quarter of 2008(before the banking crisis) it reached USD 38.45bn. More importantly, the share of loans in liabilitiesincreased from 16.4% in end-2005 to 30.5% in mid-2008, becoming the second largest item in liabilitiesfollowing personal deposits. The Ukrainian banking sector is still significantly underdeveloped in thisrespect compared to the banking sectors of Eastern European countries. However, a reliable and stronglegislative framework still has to be established to suppress corruption, business and political pressureon domestic banks’ activities.

3. Theoretical model

3.1. Model setup

The theoretical model proposed in this paper is based on the bank value optimization problems andrepresents a generalization of the firm investment models by Abel (1983) and Hahm (1996). The

3 For a historical overview of the Ukrainian banking sector, see Baum et al. (2008) and the references therein.


present value of the bank, V(Lt, Bt�1) is equated to the expected discounted stream of profits, pt, whereb is the discount factor4:

VðLt; Bt�1Þ ¼ maxX1t¼1

1

1 þ bEt�1½pt� (1)

Profit of the bank is defined as revenues from lending activities minus operating expenses:

pt ¼ rltlt � wLt � rbBg

t (2)

where lt is the amount of loans lent with rlt interest rate. To produce credits lt, the bank hires labor Lt,

and borrows money Bt�1 from perfectly elastic input markets. These factors of production cost w andrb, respectively. Additionally, the model implies increasing convex costs incurred by adjustment ofborrowing. The cost of adjustment function has constant elasticity g>1. The loan production isdescribed by a Cobb–Douglas function:

lt ¼ Lat B1�a

t�1 (3)

where a is the output elasticity of labor. It is assumed that 0<a<1.Banks are indecisive about the output price as the credit interest rate is determined by the

interaction between sector supply and demand for loans Ldt ¼ LS

t ¼ nlt . Ldt represents the sector-wide

demand for credit resources, while there are n identical banks which are perfect competitors in thecredit market. Thus, the demand for loans is the main source of uncertainty and can be parameterizedas a downward sloping function.

rlt ¼ ðLd

t Þ�1=e

pt (4)

0<e<1, where e is the elasticity of demand and pt is a stochastic shifting variable. An increment ofdemand for credits reflects in a shift to the right of the demand curve, increasing pt. A geometricrandom walk process for pt is considered to model the uncertainty in the demand for credits.5

log ptþ1 � log pt � N �s2

2; s2

� �(5)

A differentiation of (1) with respect to Lt and Bt�1 gives the following first order conditions:

arltL

a�1t B1�a

t�1 � w ¼ 0 (6)

1

1 þ bEt ½ð1 � aÞrl

tþ1Latþ1B�a

t � � grbBg�1t ¼ 0 (7)

Plugging (6) and (3) into (4) yields the equilibrium interest rate for credits.

rlt ¼ ðnBt�1Þ�1=e w

a

� �a=ð1�aÞept

" #1=½1þða=ðð1�aÞeÞÞ�

(8)

Optimal borrowing with the stochastic shifting variable is received due to combining (5)–(8)6:

Bt ¼1

grb

1 � a1 þ b

aw

� �aðe�1Þk 1

n

� �k

pekt eð1=2Þekðek�1Þs2

" #1=ðkþg�1Þ

(9)

4 The bank index i is suppressed except when needed for purposes of clarity.5 See Hahm (1996) for a similar modeling strategy.6 If log pt � Nðm p; s2

pÞ, then Eð pÞ ¼ expðm p þ s2p=2Þ and Eð pSÞ ¼ expðsm p þ s2=ð2s2

pÞÞ. This gives Eð pektþ1Þ ¼ pek

t eð1=2Þekðek�1Þs2,

where k=1/[(1�a)e+a].


The expressions for optimal borrowing (9) and interest rate in equilibrium (8) are inserted into thetransformed Eq. (3):

ltþ1 ¼ ðBtÞ1�ak 1

n

� �ak aw

� �ðað1�akÞÞ=ð1�aÞpaek

tþ1 (10)

Using (5) after some transformations, we receive the expression to characterize the impact ofuncertainty on bank loans:

ltþ1 ¼1 � a

grbð1 þ bÞ

� �ð1�akÞ=z 1

n

� �ðk½ð1�akÞþaz�Þ=z aw

� �ðað1�akÞ½kð1�aÞðe�1Þþz�Þ=ðð1�aÞzÞ

� pek½1þaðz�kÞ�

zt eð1=2Þs2 ½ðek�1Þ=zþaðaek�1Þ� (11)

where z=k+g�1.Finally, the partial derivative of borrowing with respect to uncertainty is:

@ltþ1

@s2¼ 1

2ltþ1

ðek � 1Þð1 � akÞz

þ aðaek � 1Þ� �

(12)

The sign of the partial derivative of this expression @lt+1/@s2 can be either positive or negative,depending on the peculiarities of economies and financial markets. An important implication of themodel results is that the relationship between uncertainty in demand for loans and bank lending isnegative in case of inelastic demand (e<1).

This appears to be very plausible for the Ukrainian economy because availability of financial fundsdoes not notably affect loan interest rates.

3.2. Empirical model

To track the link between bank lending and macroeconomic uncertainty, we estimate the followingeconometric specification:

L

Kit¼ m0 þ m1

L

Kit�1þ m2

PKitþ m3Kit þ m4tt�1 þ f i þ eit (13)

where i and t denote bank and quarter, respectively. Our dependent variable is the loans to capitalratio L/Kit. The right hand side variables include important bank-specific characteristics. The proxiesfor the key variable of interest, macroeconomic uncertainty, tt�1 are discussed in the next subsection.To control for economies of scale in loan production we include the natural logarithm of net worth ofbank Kit. The performance of a bank is measured by the bank’s profit to capital ratio, P/Kit. We controlfor bank size and profitability because Mertens and Urga (2001) argue that in Ukraine small and largebanks demonstrate different cost- and profit-efficiency that could affect lending behavior.7 Finally, eit

is the error term, while f 0i is the bank-specific effect.

3.3. Identifying macroeconomic uncertainty

The literature points out various candidates for macroeconomic uncertainty proxies such as themoving standard deviation (Ghosal and Loungani, 2000) or standard deviation across 12 forecastingterms of the output growth and inflation rate in the next 12 months (Driver and Moreton, 1991).

However, as in Driver et al. (2005) and Byrne and Davis (2005), we stick to a GARCH model tomeasure our first proxy of macroeconomic uncertainty. This approach suits our case better becausedisagreement among forecasters may not be a valid uncertainty measure and may containmeasurement errors. Finally, conditional variance is a better candidate for uncertainty compared tounconditional variance because it incorporates the previous period’s information set. This

7 We also experimented by dropping the profitability measure from our main regression specification and received

quantitatively similar results.

Table 1GARCH proxies for macroeconomic uncertainty.

M1 M2 M1–M0 M2–M1 CPI PPI

Constant 0.002 0.005*** 0.000 0.002 �0.005*** 0.001***

(0.002) (0.001) (0.004) (0.002) (0.002) (0.002)

AR �0.240*** �0.182*** �0.254*** �0.148 �0.080** �0.708***

(0.092) (0.064) (0.081) (0.108) (0.035) (0.028)

ARCH(1) 0.615*** 0.819*** 0.174*** 0.935*** 11.042*** �5.684***

(0.131) (0.107) (0.055) (0.081) (0.764) (0.272)

GARCH(1) 0.451*** 0.306*** 0.786*** 0.373*** �0.002 0.026

(0.107) (0.053) (0.056) (0.050) (0.005) (0.016)

Constant 0.000** 0.000*** 0.000 0.000*** 0.000** 0.000***

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

Log likelihood 282.35 340.94 224.30 332.85 45.81 41.69

Obs 198 187 187 187 197 197

* Significant at 10%.** Significant at 5%.*** Significant at 1%.

Table 2Correlation of macroeconomic uncertainty proxies.

tM1 tM2 tM2–M1 tM1–M0 tCPI tPPI zbipower

tM2 0.81

tM2–M1 0.67 0.40

tM1–M0 0.73 0.85 0.56

tCPI �0.12 �0.17 0.06 �0.02

tPPI �0.02 �0.02 0.03 0.08 0.91

zbipower 0.38 0.43 0.22 0.16 �0.36 �0.42

zh0.59 0.25 0.20 0.10 �0.09 �0.12 0.15

Note: The t2 measures are conditional variances, derived from GARCH estimations using monthly data, z measures are

calculated using daily data.


macroeconomic uncertainty identification approach closely follows Baum et al. (2006). As banksdetermine the optimal loan to total capital ratio in anticipation of future macroeconomic shocks, thedifficulty of evaluating the optimal amount of lending increases with the level of macroeconomicuncertainty.8

We draw our series for measuring macroeconomic uncertainty from the monthly growth ofmonetary aggregates (M1 and M2), demand deposits (M1–M0), time deposits (M2–M1), as well asconsumer price index (CPI) and producer price index (PPI) series. The first two series are available on amonthly basis from the National Bank of Ukraine. The price indices are produced by the State StatisticsOffice. A generalized ARCH (GARCH(1,1)) model is built for all these series, where the mean equation isan autoregression. Details of the estimated model are described in Table 1. We have significant ARCHand GARCH coefficients for most time series. The conditional variances derived from these GARCHmodels are averaged to the quarterly frequency and then employed in the analysis as alternativemeasures of macroeconomic uncertainty. Notably, we use not only lagged but also weightedconditional variances of variable.9 The introduction of arithmetic lag proxies allows us to capture thecombined effects of contemporaneous and lagged levels of uncertainty.10 In addition to the

8 While in the existing literature the loans to assets ratio is more widely used, different normalization does not change the

results notably because the capital-to-assets ratio usually changes in a very narrow band.9 The weights 0.7, 0.2 and 0.1 correspond to s2

t , s2t�1, and s2

t�2 for the macroeconomic indicators.10 Some caveats should be noted in the approach described above. The choice of a particular proxy for generating

macroeconomic uncertainty might be dependent upon the choice of the model and might exhibit significant variability over

specifications.

Table 3Definitions and descriptive statistics.

Variable Definition m s

Lending/Capital Credits and accounts receivable to total own capital ratio 3.895 2.291

Profit/Capital Profit/Loss in accounting period normalized by total own capital 0.040 0.041

Capital Total own capital 11.094 0.949

tM1 Weighted conditional variances of M1 growth 0.006 0.002

tM2 Weighted conditional variances of M2 growth 0.003 0.001

tM1–M0 Weighted conditional variances of (M1–M0) growth 0.012 0.005

tM2–M1 Weighted conditional variances of (M2–M1) growth 0.003 0.001

tCP1 Weighted conditional variances of CPI growth 11.047 33.456

tPP1 Weighted conditional variances of PPI growth 8.739 20.715

zbipower Bipower variation measure of uncertainty (Ghysels et al., 2006) 0.086 0.043

zhUncertainty measure is based on Merton (1980) 1.041 3.679

Note: This table reports descriptive statistics for Ukrainian banks. The time span is from 2003Q2 to 2008Q2. The number of

observations is 2777. m is mean, while s is the standard deviation.


conditional variances of the macro series, two volatility proxies are estimated employing daily PFTS(stock exchange) index returns. The first measure is based on Merton (1980).11 In order to employ theMerton (1980) methodology, we first take the squared first difference of the daily changes in returnsdivided by the square root of the number of trading days. This difference is defined as the dailycontribution to annual volatility. This approach provides a more representative measure of theperceived volatility while avoiding potential problems such as the high persistence of shocks. Second,using absolute returns, we utilize the bipower variation measure of uncertainty described in Ghyselset al. (2006).

As can be seen from Table 2, there are three distinct groups of uncertainty proxies: monetary(M1, M2, demand deposits (M1–M0), time deposits (M2–M1)), price indices (PPI, CPI) and stockindices. Correlation within each group is high, but correlation between proxies from differentgroups is low, which is hardly surprising, bearing in mind the nature of these series. Therefore, wecan use them for the composition of complimentary proxies, which should demonstrate therobustness of our results.

Ideally, other proxies could also be exploited for uncertainty measurement (e.g. labor, industrialproduction or gross domestic product), but most of these series are either too short or unreliable. Forexample, the reliability of labor data is questionable due to substantial hidden unemployment issues.In the case of the real GDP or industrial production series, even the State Statistics Office’s ownpublications inform that monthly series are calculated unsatisfactorily and, therefore, cannot be usedin an econometric analysis. More reliable data are available only on a quarterly and annual basis,which is not satisfactory for a GARCH estimation.

4. Empirical implementation

4.1. Data set

The data items loans, profits, capital and total assets are utilized to construct bank-specificvariables. We use quarterly data on all Ukrainian banks’ balance sheets, which are published in theofficial NBU’s ‘Visnyk NBU’.12

The balanced NBU dataset has 2777 observations from 2003Q2 to 2008Q2. Variables include in-depth data on the structure of banks’ assets, liabilities and capital. While even the larger sample givessatisfactory results, it is better to screen the data before starting empirical investigations. Bank levelvariables are winsorized at the most extreme (top and bottom) 1% level of the distribution on anannual basis to alleviate the influence of extreme observations. After the exclusion of newly arrived

11 The daily index series are taken from the PFTS website, http://www.pfts.com.ua.12 Referred to henceforth as the NBU data set.

http://www.pfts.com.ua/


and closed banks the sample contains 171 banks. In order to work only with long time series forindividual banks, we exclude all banks which have less than half time points.13 The basic descriptivestatistics and data definitions are available in Table 3.

To examine the effects of macroeconomic uncertainty on groups of banks having similarcharacteristics the bank data are divided into small and large banks. A bank is defined as SMALL if itsaverage yearly assets are below the mean, otherwise it is considered as LARGE. Similarly, we categorizebanks as most profitable and least profitable or non-profitable. A bank is defined as MOST PROFITABLE

if its average over-the-years net profits are above the mean, otherwise it is considered as LEAST

PROFITABLE.

4.2. Results for all banks

In this section we investigate the extent to which lending behavior responds to volatility in themacroeconomic environment. Our analysis starts with the evaluation of the full sample of Ukrainianbanks, and later we look at differences in results across sub-samples based on banks’ capital andperformance measures.14

Estimates of the optimal bank capital structure measures usually suffer from endogeneityproblems, and the use of instrumental variables may be considered as a possible solution. To calculateour econometric models we utilize a two-step GMM-SYSTEM dynamic panel data estimator. TheGMM-SYSTEM, unlike the usual GMM, uses not only transformed equations but combinestransformed equations with level equations (Blundell and Bond, 1998). Lagged levels are consideredas instruments for transformed equations and lagged differences are used as instruments for levelequations. The models are estimated employing a first difference transformation to remove theindividual bank effect.

The reliability of our econometric methodology depends crucially on the validity of its instruments.We check it with Sargan’s test of overidentifying restrictions, which is asymptotically distributed as x2

in the number of restrictions. The consistency of estimates also depends on the serial correlation in theerror terms. We present test statistics for second-order serial correlation. The results are estimatedusing finite sample correction (Windmeijer, 2005). Since there is some degree of simultaneity in thedecisions of bank managers, we consider all bank-level variables as endogenous, while country levelcharacteristics are assumed to be exogenous. The matrix of instruments for the all banks estimationincludes L/Kt�3 to L/Kt�4, P/Kt�2 to P/Kt�3, Kt�2 to Kt�3 and DL/Kt�3 to DL/Kt�4, DP/Kt�2 to DP/Kt�3

and DKt�2 to DKt�3.The outcomes of estimating Eq. (13) for all banks are given in Tables 4 and 5. The columns of Table 4

represent the final result of the two-step GMM System estimator with weighted conditional varianceof four different monetary parameters: M1 monetary aggregate growth, M2 monetary aggregategrowth, domestic currency demand deposits growth (M1–M0), and time deposits growth (M2–M1).The sign of all proxies is in line with the theoretical expectations and three of them (M1, M2 and timedeposits) are significant at the 1% level. However, the insignificance of the measure based on demanddeposits can be explained by a shift from demand deposits to time deposits because of the resumptionof trust to the banking system.

The estimation results suggest the existence of a significant negative relationship between a bank’slending and macroeconomic uncertainty measured with proxies based on monetary aggregates. Thestatistically significant coefficients vary from �60.05 to �32.43 for time deposits and M1 measures,respectively. The difference is caused mainly by the different nature of the proxies and the degree inwhich they can be managed by authorities. Another important outcome is that the persistence in theoverall credits to capital ratio in period t�1 is also observed, which suggests that on a quarterly basis,inertia is very important. Another statistically significant coefficient in all specifications – the profit tocapital ratio – ranges from �8.29 to �7.86. Table 5 presents the result with the weighted conditionalvariance of the consumer price index (CPI) and producer price index (PPI), as well as two different

13 Series can have a maximum of 15 time points. All banks that have less than seven time points are newly entered banks.14 Similar estimates were made using the alternative data set from the Association of Ukrainian Banks (AUB). The results were

quite similar, thus we report only the results of one data set to avoid confusion.

Table 4Determinants of bank lending: all banks, monetary proxies.

(1) (2) (3) (4)

Lending/Capitalt�1 0.930*** 0.927*** 0.916*** 0.920***

(0.049) (0.048) (0.050) (0.048)

Profit/Capitalt �8.294*** �7.935*** �7.863*** �8.139***

(1.801) (1.695) (1.695) (1.735)

Capitalt 0.284*** 0.294*** 0.280*** 0.287***

(0.069) (0.068) (0.073) (0.068)

tM1,t�1 �32.434***

(11.174)

tM2,t�1 �54.796***

(15.560)

tM1–M0,t�1 �1.261

(3.171)

tM2–M1,t�1 �60.047***

(21.035)

Constant �2.217*** �2.351*** �2.332*** �2.267***

(0.760) (0.741) (0.821) (0.752)

AR(2) p-val 0.30 0.27 0.38 0.29

Sargan p-val 0.75 0.75 0.69 0.73

Obs 2777 2777 2777 2777

Note: Asymptotic robust standard errors are reported in parentheses. The matrix of instruments includes L/Kt�3 to L/Kt�4, P/Kt�2

to P/Kt�3, Kt�2 to Kt�3 and DL/Kt�3 to DL/Kt�4, DP/Kt�2 to DP/Kt�3 and DKt�2 to DKt�3. ‘‘Sargan’’ is a Sargan–Hansen test of

overidentifying restrictions (p-value reported). ‘‘AR(k)’’ is the test for kth order autoregression. Capital is the natural logarithm of

total own capital. Lending/Capital is defined as the credits and accounts receivable to total own capital ratio. Profit/Capital

represents the Profit/Loss in accounting period normalized by total own capital.

* Significant at 10%.

** Significant at 5%.*** Significant at 1%.


possible proxies based on the stock index. As can be seen, proxies based on the stock index aresignificant, while neither PPI nor CPI is. The latter could be explained by problems with measurementinflation. This macroeconomic indicator is often underreported because of political reasons.

4.3. Results for subsamples of banks

Having established the presence of a negative role for macroeconomic uncertainty on bank’slending, we next investigate whether the strength of association varies across groups of banks withdifferent characteristics. Table 6 presents two measures of macroeconomic volatility. Panel A showsthe results with weighted conditional variance of M2 growth, while Panel B reports estimates withMerton’s uncertainty measure.15 Both panels show outcomes for small/large and most/least profitablebanks.

There are interesting differences across the large and small banks subsamples which are reportedin Columns (1) and (2), respectively. Employing two measures of uncertainty, we observe negative andstatistically significant links between lending and macroeconomics volatility for both categories ofbanks. However, large banks appear to be more sensitive to changes in the macroeconomicenvironment. This finding suggests that small banks might be less able to change their behavior overtime in response to changes in the uncertainty. Additionally, their lending depends heavily onavailable resources. It could also be explained by the fact that small banks in Ukraine are often referredto as ‘‘pocket banks’’ due to an extreme concentration of credits in one (usually affiliated) entity.

Columns (3) and (4) of Table 6 report the results for most and least profitable banks. The effect ofthe M2 measure is negative and statistically significant for both categories of banks. However, the

15 Results for other measures of uncertainty follow the patterns of the presented estimates. Only inflation based proxies

appear to be statistically insignificant, which could be explained by the severe mismeasurement of inflation in Ukraine.

Table 5Determinants of bank lending: all banks, non-monetary proxies.

(1) (2) (3) (4)


(0.049) (0.049) (0.046) (0.044)

Profit/Capitalt �7.625*** �7.540*** �8.285*** �8.097***

(1.785) (1.716) (1.650) (1.610)

Capitalt 0.285*** 0.281*** 0.270*** 0.293***

(0.069) (0.068) (0.073) (0.069)

tCPI,t�1 �0.000

(0.001)

tPPI,t�1 �0.001

(0.001)

zbipowert�1 �0.958*

(0.554)

zht�1 �0.009*

(0.004)

Constant �2.424*** �2.363*** �2.074** �2.531***

(0.746) (0.743) (0.844) (0.752)

AR(2) p-val 0.37 0.34 0.38 0.33

Sargan p-val 0.68 0.73 0.91 0.90

Obs 2777 2777 2777 2777



overidentifying restrictions (p-value value reported). ‘‘AR(k)’’ is the test for kth order autoregression. Capital is the natural

logarithm of total own capital. Lending/Capital is defined as the credits and accounts receivable to total own capital ratio. Profit/

Capital represents the Profit/Loss in accounting period normalized by total own capital.* Significant at 10%.** Significant at 5%.*** Significant at 1%.

Table 6Determinants of bank lending: results for subsamples.

Size Profitability

Large Small Most Least

Panel A: M2 based measure of uncertainty


(0.068) (0.121) (0.066) (0.077)

Profit/Capitalt �8.572*** �20.643*** �7.869*** �11.125**

(2.176) (6.467) (1.919) (5.265)

Capitalt 0.228** 0.150 0.287*** 0.269**

(0.096) (0.145) (0.100) (0.120)

tM2,t�1 �84.580*** �84.273*** �72.726*** �49.604**

(25.952) (27.102) (25.319) (23.967)

Constant �1.112 �0.770 �1.997* �2.110

(1.280) (1.604) (1.127) (1.270)

AR(2) p-val 0.48 0.15 0.72 0.12

Sargan p-val 0.79 0.44 0.85 0.21

Obs 1415 1318 1459 1318

Panel B: Merton’s measure of uncertainty


(0.063) (0.118) (0.060) (0.073)

Profit/Capitalt �8.563*** �17.202*** �7.825*** �13.572***

(2.127) (5.563) (1.742) (4.060)

Capitalt 0.233** 0.191 0.306*** 0.247**

(0.096) (0.137) (0.099) (0.124)

zht�1 �0.013* �0.006 �0.015** �0.007

(0.008) (0.007) (0.007) (0.006)


Table 6 (Continued )

Size Profitability

Large Small Most Least

Constant �1.462 �1.491 �2.489** �2.042

(1.239) (1.493) (1.114) (1.296)

AR(2) p-val 0.55 0.15 0.51 0.12

Sargan p-val 0.94 0.49 0.96 0.22

Obs 1415 1318 1459 1318



overidentifying restrictions (p-value reported). ‘‘AR(k)’’ is the test for kth order autoregression. Capital is the natural logarithm of

total own capital. Lending/Capital is defined as the credits and accounts receivable to total own capital ratio. Profit/Capital

represents the Profit/Loss in accounting period normalized by total own capital.* Significant at 10%.** Significant at 5%.*** Significant at 1%.


sensitivity of lending to macro uncertainty is higher for most profitable banks (�72.73) compared totheir least profitable counterparts (�49.60). The difference is even stronger for results with Merton’smeasure of uncertainty. Only most profitable banks are likely to decrease their lending ratios inresponse to increased stock market volatility.

Thus, we receive empirical confirmation of our analytical hypothesis. An increase in the level ofmacroeconomic uncertainty leads to narrowing of bank lending. This behavior could be explained byhigher risk aversion of bank managers in more uncertain times. The result is robust because differentproxies yield the same outcome. We show the dissimilarities in behavior of small and large banks andof more and less profitable banks. Different groups of banks have different sensitivity to changes in themacroeconomic environment as measured by different proxies. This can allow for a shift of lendingfrom one group of banks to another if only one measure of uncertainty changes.

5. Conclusions

This paper investigates the link between commercial banks’ lending and macroeconomicuncertainty. Managers maximize the bank value, equal to a discounted stream of bank’s profits. Basedon theoretical predictions, we claim that higher uncertainty can lead to lower lending in emergingmarket economies. The empirical predictions of this model on the sample of Ukrainian banks areexamined. Using eight alternative measures of macroeconomic uncertainty, we find out that banksdecrease their supply of loans when the volatility of macroeconomic variables increases. Consistentwith the value-maximizing model, we receive significant evidence that banks extend credit supplywhen macroeconomic uncertainty decreases. This effect remains after controlling for size,profitability, and capital. Moreover, there is a distinct sensitivity of contrasted groups of bankswith respect to different proxies for macroeconomic volatility. The result is achieved for subsamplesbased on size and profitability of banks.

We believe that this evidence sheds light on three sets of questions. First, our unique dataset allowsus to investigate the effects of macroeconomic uncertainty on banks’ lending in a country with anunderdeveloped financial sector. The estimated effects of macroeconomic uncertainty negativelyinfluence bank lending, which is also consistent with the predictions from the dynamic model of bankvalue maximization in case of inelastic demand for credits. Moreover, some macroeconomicuncertainty proxies have marginal or no impact on some groups of banks. Second, our resultscontribute to the existing literature of a bank lending channel for monetary policy.16 Banks affectbank-dependent borrowers’ ability to finance their investment projects through this channel. There issubstantial evidence for the effects of monetary policy on banks’ balance sheets (Kashyap and Stein,1995). When a bank’s financial situation reflects the borrowers’ financial statement or switching costs

16 See Rajan and Parulkar (2008) and Elbourne and de Haan (2009) for a detailed description of monetary policy channels.


are small, the effects of a bank lending channel on monetary policy is minimal (Hubbard et al., 2002).Third, if there is a negative effect of macroeconomic uncertainty on bank’s lending behavior, one canfind out how the riskiness of the whole system changes. This should allow for better bank supervision,thus minimizing the effect of external shocks.

Our research has important policy implications. According to Nier and Zicchino (2005), a decreasein loan supply may reduce aggregate investment, therefore amplifying macroeconomic fluctuations.These consequences are not confined to particular countries and particular times. When banks curtailtheir lending, companies are unable to obtain funds and may be forced to default on their obligations.Moreover, as shown by Dell et al. (2005), scarcity of funds may lead to early liquidation of long-terminvestments, which affects the long-term growth trend as well.

This research is the first attempt to study and test the effect of changes in macroeconomicuncertainty on bank lending in an emerging market country. The results of this investigation cannot beconsidered a definitive answer to what is the appropriate policy for the NBU or other state agenciesthat supervise the financial sector, except to convey the general notion that they have to decrease thelevel of macroeconomic uncertainty whenever possible.

While taking a look at the sensitivity of bank lending to macroeconomic uncertainty, someunexplored topics are left for future exploration if available data permit. Among the mostinteresting parts is the investigation of herding behavior of banks. Researchers could check whathappens with cross-sectional distribution of bank-lending ratios when macroeconomic uncertaintyincreases.

Acknowledgments

We are grateful to Barry W. Ickes, Wojciech W. Charemza, Mark E. Schaffer, and the input ofparticipants of the EERC seminars, Verein fur Socialpolitik meeting, Munich, 2007, and the FirstConference of Portuguese Economic Journal, Azores, 2007 for thoughtful comments and suggestions.The standard disclaimer applies. Tsapin and Zholud acknowledge research support from theEconomics Education and Research Consortium (Grant No. R04-1081).

Appendix A. Theoretical model

The theoretical model proposed in this paper is based on the bank value optimization problems andrepresents a generalization of the models by Abel (1983) and Hahm (1996). The present value of thebank, V(Lt, Bt�1), is equated to the expected discounted stream of profits, pt, where b is the discountfactor:

VðLt; Bt�1Þ ¼ maxX1t¼1

1

1 þ bEt�1½pt� (A1)

Profit of the bank is defined as revenues from lending activities minus operating expenses:

pt ¼ rltlt � wLt � rbBg

t (A2)

where lt is the amount of loans lent with rlt interest rate. To produce credits lt, the bank hires labor Lt,

and borrows money Bt�1 from perfectly elastic input markets. These factors of production cost w andrb, respectively. Additionally, the model implies increasing convex costs incurred by adjustment ofborrowing. The cost of adjustment function has constant elasticity g>1. The loan production isdescribed by a Cobb–Douglas function:

lt ¼ Lat B1�a

t�1 (A3)

where a is the output elasticity of labor. It is assumed that 0<a<1.As the credit interest rate is determined by an interaction between sector supply and demand for

loans Ldt ¼ Ls

t ¼ nlt , banks are indecisive about the output price. Ldt represents the sector-wide demand


for credit resources, while there are n identical banks which are perfect competitors in the creditmarket. Thus, the demand for loans is the main source of uncertainty and can be depicted as adownward sloping function.

rlt ¼ ðLd

t Þ�1=e

pt (A4)

0<e<1 where e is the elasticity of demand and pt is a stochastic shifting variable. An increment ofdemand for credits reflects in a shift to the right of the demand curve, increasing pt.

A geometric random walk process for pt is used to model the spread in demand shocks.

log ptþ1 � log pt � N �s2

2; s2

� �(A5)

Differentiation of (A1) with respect to Lt and Bt�1 gives the following first order conditions:

arltL

a�1t B1�a

t�1 � w ¼ 0 (A6)

1

1 þ bEt ½ð1 � aÞrl

tþ1Latþ1B�a

t � � grbBg�1t ¼ 0 (A7)

From (A6) the optimal level of labor is defined as:

Lt ¼ Bt�1arl

t

w

� �1=ð1�aÞ

(A8)

Similarly to Eq. (7) from Hahm (1996), we get the optimal level of Bt

Bt ¼1 � a1 þ b

� �1

grb

aw

� �a=ða�1ÞE ðrl

tþ1Þ1=ð1�aÞ� �� ð1=ðg�1ÞÞ

(A9)

Plugging (A8) and (A3) into (A4) yields the equilibrium interest rate for credits.

rlt ¼ ðnBt�1Þ�1=e w

a

� �a=ðð1�aÞeÞpt

" #1=½1þða=ðð1�aÞeÞÞ�

(A10)

We rewrite (A9) using (A10)

Bt ¼1

grb

1 � a1 þ b

aw

� �ðaðe�1ÞÞ=ðð1�aÞeþaÞ 1

n

� �1=ðð1�aÞeþaÞE pe=ðð1�aÞeþaÞ

tþ1

� �" #1=ðð1=ðð1�aÞeþaÞÞþg�1Þ

(A11)

If log pt � Nðmp; s2pÞ, then Eð pÞ ¼ expðm p þ ð1=2Þs2

pÞ and Eð psÞ ¼ expðsmp þ ðs2=2Þs2pÞ. This gives

Eð pektþ1Þ ¼ pekt eð1=2Þekðek�1Þs2

, where k=1/((1�a)e+a). Substitution of the latter into (A11) allows us to

express optimal borrowing with the stochastic shifting variable

B ¼ 1

grb

1 � a1 þ b

aw

� �aðe�1Þk 1

n

� �k

pekt eð1=2Þekðek�1Þs2

" #1=ðkþg�1Þ

(A12)

We rewrite (A3) using the optimal level of labor hired by the bank to express the level of loans forperiod t+1:

ltþ1 ¼ Btarl

tþ1

w

!a=ð1�aÞ

(A13)

The expressions for optimal borrowing (A12) and interest rate in equilibrium (A10) are inserted into(A13):

ltþ1 ¼ ðBtÞ1�ak 1

n

� �ak aw

� �ðað1�akÞÞ=ð1�aÞpaek

tþ1 (A14)


Using (A5) after some transformations, we receive the expression to characterize the impact ofuncertainty on bank loans

ltþ1 ¼1 � a

grbð1 þ bÞ

� �ð1�akÞ=z 1

n

� �ðk½ð1�akÞþaz�Þ=z aw

� �ðað1�akÞ½kð1�aÞðe�1Þþz�Þ=ðð1�aÞzÞ

� ptðek½1þaðz�kÞ�Þ=zeð1=2Þs2 ½ðek�1Þð1�akÞ=zþaðaek�1Þ� (A15)

where z=k+g�1.Finally, the partial derivative of borrowing with respect to uncertainty is:

@ltþ1

@s2¼ 1

2ltþ1

ðek � 1Þð1 � akÞz

þ aðaek � 1Þ� �

(A16)

Despite a complexity of the expression, the sign of derivative @lt+1/@s2 could be either positive ornegative, which depends on the ek term, and e in particular. Apparently, a negative relationshipbetween credit interest rate uncertainty and bank lending results for inelastic demand, while apositive effect of the uncertainty can be observed for the case of an elastic demand curve.

References

Abel, A.B., 1983. Optimal investment under uncertainty. American Economic Review 73, 228–233.Baum, C.F., Caglayan, M., Ozkan, N., 2003. The role of uncertainty in the transmission of monetary policy effects on bank lending.

Boston College Working Papers in Economics 561, Boston, MA.Baum, C.F., Caglayan, M., Ozkan, N., 2009. The second moments matter: the impact of macroeconomic uncertainty on the

allocation of loanable funds. Economics Letters 102, 87–89.Baum, C.F., Caglayan, M., Ozkan, N., Talavera, O., 2006. The impact of macroeconomic uncertainty on non-financial firms’

demand for liquidity. Review of Financial Economics 15, 289–304.Baum, C.F., Caglayan, M., Schaefer, D., Talavera, O., 2008. Political patronage in Ukrainian banking. The Economics of Transition

16, 537–557.Bhaumik, S.K., Piesse, J., 2008. Does lending behaviour of banks in emerging economies vary by ownership? Evidence from the

Indian banking sector. Economic Systems 32, 177–196.Blundell, R., Bond, S., 1998. Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics

87, 115–143.Byrne, J.P., Davis, E.P., 2005. Investment and uncertainty in the G7. Review of World Economics 141, 1–32.Calmes, C., Salazar, J., 2006. Variance macroeconomique conditionnelle et mesure de dispersion des actifs dans les portefeuilles

bancaires. In: Racicot, F.E., Theoret, R. (Eds.), Finance computationnelle et gestion des risques. Presses de l’Universite duQuebec, Quebec, pp. 687–700.

Claeys, S., Vander Vennet, R., 2008. Determinants of bank interest margins in Central and Eastern Europe: a comparison with theWest. Economic Systems 32, 197–216.

Dell, A., Detragiache, E., Rajan, R., 2005. The real effect of banking crises. IMF Working Paper 05/63, Washington, DC.Driver, C., Moreton, D., 1991. The influence of uncertainty on aggregate spending. Economic Journal 101, 1452–1459.Driver, C., Temple, P., Urga, G., 2005. Profitability, capacity, and uncertainty: a model of UK manufacturing investment. Oxford

Economic Papers 57, 120–141.Elbourne, A., de Haan, J., 2009. Modeling monetary policy transmission in acceding countries: vector autoregression versus

structural vector autoregression. Emerging Markets Finance and Trade 45, 4–20.Ghosal, V., Loungani, P., 2000. The differential impact of uncertainty on investment in small and large business. The Review of

Economics and Statistics 82, 338–349.Ghysels, E., Santa-Clara, P., Valkanov, R., 2006. Predicting volatility: getting the most out of return data sampled at different

frequencies. Journal of Econometrics 131, 59–95.Hahm, S., 1996. Uncertainty and investment in equilibrium. International Economic Journal 10, 71–82.Hubbard, G., Kuttner, K.N., Palia, D.N., 2002. Are there bank effects in borrowers’ costs of funds? Evidence from a matched

sample of borrowers and banks. Journal of Business 75, 559–573.Jimborean, R., 2009. The role of banks in the monetary policy transmission in the new EU member states. Economic Systems 33,

360–375.Kashyap, A., Stein, J., 1995. The impact of monetary policy on bank balance sheets. Carnegie-Rochester Conference Series on

Public Policy 42, 151–195.Kashyap, A., Stein, J., 2000. What do a million observations on banks say about the transmission of monetary policy? American

Economic Review 83, 78–98.Krkoska, L., Teksoz, U., 2009. How reliable are forecasts of GDP growth and inflation for countries with limited coverage?

Economic Systems 33, 376–388.Mertens, A., Urga, G., 2001. Efficiency, scale and scope economies in the Ukrainian banking sector in 1998. Emerging Markets

Review 2, 292–308.Merton, R.C., 1980. On estimating the expected return on the market: an exploratory investigation. Journal of Financial

Economics 8, 323–361.Modigliani, F., Miller, M., 1958. The cost of capital, corporate finance, and the theory of investment. American Economic Review

48, 261–297.


Nier, E., Zicchino, L., 2005. Bank Weakness and Bank Loan Supply. Mimeo, Bank of England, London.Quagliariello, M., 2009. Macroeconomic uncertainty and banks’ lending decisions: the case of Italy. Applied Economics 41, 323–

336.Rajan, R.S., Parulkar, M., 2008. Real sector shocks and monetary policy responses in a financially vulnerable emerging economy.

Emerging Markets Finance and Trade 44, 21–33.Rousseau, P.L., Yilmazkuday, H., 2009. Inflation, financial development, and growth: a trilateral analysis. Economic Systems 33,

310–324.Stein, J.C., 1998. An adverse selection model of bank asset and liability management with implications for the transmission of

monetary policy. RAND Journal of Economics 29, 466–486.Weill, L., 2011. How corruption affects bank lending in Russia. Economic Systems 35, 230–243.Windmeijer, F., 2005. A finite sample correction for the variance of linear efficient two-step GMM estimators. Journal of

Econometrics 126, 25–51.

macroeconomic uncertainty and bank lending: the case of ukraine

Documents