Transmission of a global financial crisis shock to an emerging economy

1 INTRODUCTION

Banks around the world are interconnected because they provide lending and other financial services, not only to domestic financial and nonfinancial firms but also to customers abroad. On one hand, this interconnection provides liquidity support and reduces the idiosyncratic risk of individual banks via diversification, whereas on the other hand, it also leaves banks vulnerable to systemic shock in the form of a speculative attack, financial panic, or herd behaviour (Kleimeier, Lehnert, & Verschoor, 2008). An important question is whether, and to what extent, this systemic shock is transmitted to the real economy.¹ In this regard, the impact of the last global financial crisis (GFC) of 2007–2009 on real businesses has also attracted considerable attention (see Berg & Kirschenmann, 2015; De Haas & Van Horen, 2013; Giannetti & Laeven, 2012; Ivashina & Scharfstein, 2010; Schnabl, 2012). However, the impact of the GFC on emerging markets has been the subject of relatively few studies, probably due to the limited availability of data.²

As we had access to the relevant data, we study the impact of the GFC on an emerging market. In doing so, we analyse the changes in credit availability to firms in Pakistan as a result of the exogenous economic shock, measured as borrowing outside Pakistan by Pakistani banks.³ Because the banks and financial intermediaries in developed countries were directly and most severely affected during the GFC, we examine whether transmission of capital flows from international banks to local banks in an emerging economy was also affected by the liquidity constraints faced by the international banks (Cetorelli & Goldberg, 2012; Khwaja & Mian, 2008). Pakistan had close trading relationship with the countries from where the GFC originated. Based on 2006 data, United States was the biggest trading partner of Pakistan. Moreover, six European countries were also among the top 10 trading partners of Pakistan.⁴ We reckon that the countries with significant trading relationship with Pakistan were also closely linked to the Pakistani banking sector. Therefore, the main objective of this paper is to study whether the financial shock that originated in those countries impacted the Pakistani economy during the GFC through the credit channel. To that end, we use a unique dataset obtained from the Credit Information Bureau (CIB) of the State Bank of Pakistan, comprising all business loans made to firms by commercial banks in Pakistan, to estimate the transmission of the economic shock via spillover from banks to real firms.

In order to establish the adverse effects of international shocks through the banking system, the Pakistani banking sector provides an ideal setting to study the transmission of international financial shocks, as local banks differ greatly in their international exposure. The international exposure of these banks ranges from zero to 22.3%. We measure this international exposure by the borrowing of Pakistani banks outside Pakistan from other financial institutions and hypothesize that the higher the borrowing outside Pakistan, the higher the financial shock to banks, and hence, those banks transmit the shock to the real economy in Pakistan. Our primary variables of interest are high, intermediate, and low exposure banks, classified as follows. Banks that have over 8% of their total borrowing outside Pakistan are considered as high exposure ones, whereas those with between 4% and 8% borrowing outside Pakistan are classified as intermediate exposure banks, and those with less than 4% borrowing outside Pakistan as low exposure banks.⁵ The higher the exposure of the banks, the more they are expected to transmit the financial shock to the local market.⁶ This classification is similar to that of Schnabl (2012), who classifies banks into these categories and then measures the impact of lending changes by each type of bank on firms in Peru. The uniqueness of our dataset lies in the fact that it covers the period of the GFC perfectly, that is, from April 2006 to March 2011. It contains all the business loans that were outstanding at the month-end with each bank, which allows us to address the identification challenge arising from the economic events that triggered the financial shocks. These shocks may also affect credit demand.

Our findings show that the banking sector in Pakistan was indeed affected by the GFC, as the lending ability of various banks was negatively affected (i.e., the intensive margin), similar to the findings of Forbes (2001), Forbes and Chinn (2004), Khwaja and Mian (2008), Schnabl (2012), and Cetorelli and Goldberg (2012). Our empirical results also indicate that there was transmission of financial shock from international markets to Pakistan. This finding is interesting because Ali (2009) notes that Pakistan has been relatively well insulated against contagion coming from international financial markets.⁷ Our results, however, confirm that lending by Pakistani banks to local firms was reduced. This reduction in lending was reported by the high exposure banks, whereas there was increase in lending by the low/relatively low⁸ exposure banks. We also find that “foreign” banks⁹ transmit the global financial crisis shock and reduce their lending. The findings on the exit from the market of old firms and the extension of credit to new firms (i.e., the extensive margin) indicate that the firms that borrowed from relatively low exposure banks are less likely to exit and are more likely to have survived after the crisis. However, lending to small firms was reduced, irrespective of whether they borrowed from relatively low or high exposure banks. The findings regarding different types of banks reveal that private and big banks reduce their lending; however, contrary to Farooq and Zaheer (2015), we do not find any statistically significant evidence that the granting of Islamic loans increased after the financial shock. We find that high exposure banks reduced their agricultural loans after the financial shock in contrast with Berg and Kirschenmann (2015). However, relatively low exposure banks increased their lending in that regard. This is intuitive, because the relatively low exposure banks are mostly domestic, small banks, which mainly focus on agriculture financing. Our findings on information asymmetry show that bank reputation as a proxy for information asymmetry, as well as the previous interaction of the borrower with the bank, was not able to overcome the information asymmetry problems and banks still reduced lending after a financial shock. However, highly reputed banks with relatively low exposure were able to overcome the information asymmetry problems and increased their lending after the GFC.

Our paper adds to the existing literature in the following ways. First, it complements the rare literature on the transmission of the global financial crisis from developed countries to banks and firms in an emerging economy. Second, we provide a detailed analysis of different bank types (e.g., private and privatized banks¹⁰) and different types of loans (e.g., Islamic, export, and agriculture loans) provided by local banks. By doing this, we are better able to analyse the impact of the global financial crisis on the real economy in Pakistan. Third, our analysis also sheds light on information asymmetry and its impact on lending changes after the recent financial crisis using loan level data. Lastly, our dataset perfectly covers the relevant data period, that is, from April 2006 to March 2011, giving us the possibility to conduct the analysis for both precrisis and postcrisis periods. ¹¹

The remainder of the paper is structured as follows. Section 2 provides the study background, whereas Section 3 describes the data. The methodology adopted in the paper is described in Section 4. Section 5 presents the results, and the conclusion is provided in Section 6.

2 THEORETICAL BACKGROUND AND RELATED LITERATURE

Our paper is related to two strands of literature. The first is the literature on the transmission of financial shocks, which operates through various channels. Next to the asset price channel¹² of transmission, the credit channel operates through information asymmetry in financial markets.¹³ This channel proposes two types of transmission: the bank lending channel and firms' borrowing channel. According to the bank lending channel, banks play an important role in solving the asymmetric information problem; certain borrowers will not be able to have access to credit in the presence of a financial shock to banks. Existing work on the bank lending channel includes Kashyap, Owen, and Stein (1994), Peek and Rossenberg (1997), Kashyap and Stein (2000), Ashcraft (2001), Khwaja and Mian (2008), and Cetorelli and Goldberg (2012). The firm borrowing channel implies that when a financial shock reduces the net worth of firms, it raises the adverse selection problem, leading to a decreased lending to firms. A decline in net worth also increases the moral hazard problem because firms have lower equity at stake and are more likely to engage in risky investment projects. Lenders can observe this and reduce lending to firms (Bernanke, Gertler, & Gilchrist, 1996; Oliner & Rudebusch, 1996). Other papers have analysed the transmission channels of international financial shocks through the banking system. Allen, Hryckiewicz, Kowalewski, and Tümer-Alkane (2014) showed that bank liquidity shocks are transmitted from the parent to the subsidiary, thereby negatively affecting lending by the subsidiaries. Similarly, Adams-Kane, Jia, and Lim (2015) studied the channels that affect credit provision during a financial crisis. Their results indicate that cross-border lending was affected, mainly due to liquidity difficulties faced by international banks and increased uncertainty.

The second strand of literature that our paper relates to is international financial contagion. The literature on this strand can be divided into studies focusing on financial sector linkages for the propagation of economic shock (Bekaert, Harvey, & Lundblad, 2005; Corsetti, Pericoli, & Sbracia, 2005; Derviz & Podpiera, 2007; Forbes & Rigobon, 2002; Jotikasthira, Lundblad, & Ramadorai, 2013) and others that focus on trade linkages (Eichengreen, Rose, & Wyplosz, 1996; Forbes, 2001, 2004; Claessens & Forbes, 2001; Forbes & Chinn, 2004; Blanchard, Dell'Ariccia, & Mauro, 2010). Forbes (2001) describes three channels through which an economic shock can be transmitted to other countries: a competitiveness effect, which has a negative effect on exports because of the adverse impact on relative competitiveness; an income effect, which also has a negative effect because of the adverse impact on income in the importing country and hence a reduction in demand for imports; and a cheap import effect, which has a positive effect because of the positive impact on relative competitiveness and hence an increase in demand for imports. This paper complements the research on the international transmission of financial shocks and international financial contagion.

Overall, our study relates closely to work by Khwaja and Mian (2008), Schnabl (2012), and Cetorelli and Goldberg (2012). Khwaja and Mian (2008) and Schnabl (2012) study the transmission of liquidity shock to real firms by using firm fixed effects to identify the bank lending channel. These studies, however, differ from ours, as that of Khwaja and Mian (2008) analyses the transmission of domestic liquidity, whereas Schnabl (2012) studies the transmission of liquidity shock across countries. We discuss the spillover of financial shock from the developed world to an emerging economy during the global financial crisis and subsequently the transmission of that financial shock to real firms by local banks. Our paper is also different from that of Cetorelli and Goldberg (2012) in the sense that they analyse the financial contagion from banking systems in developed countries to emerging markets, using data on capital flows within global banks, whereas we study the transmission of financial shock from developed banking systems to an emerging market using loan level data.

3 DATA

We use a unique dataset for our empirical analysis, which primarily consists of loans provided by Pakistani banks to local firms. The data were obtained from the Consumer Protection Department of the State Bank of Pakistan, which maintains the domestic credit registry known as the CIB. CIB is an online facility that allows financial institutions to upload and retrieve information through a virtual private network. Since 2006, the State Bank of Pakistan has upgraded the CIB's technical and reporting structure. The new facility has the capacity to incorporate a large number of validation rules to ensure the integrity and accuracy of the submitted data. Banks in Pakistan verify the credit history of a loan applicant with CIB if the application exceeds PKR 500,000. The CIB dataset has already been studied in different contexts by Khwaja and Mian (2005, 2008), Mian (2006), Zia (2008), and Baele et al. (2014).¹⁴ Our data cover the period from April 2006 to March 2011, spanning the period before and after the financial crisis of 2007–2009.

The CIB reports the identity code for each loan contract. On the basis of this, we can identify the industry and total exposure of the borrower. In addition, the database provides information on key loan characteristics, such as maturity, collateral, product name, type of loan, the approved limit, and the remaining outstanding amount. Because the database does not provide banks' financial information, data on bank characteristics such as total capital ratio, total assets, liquid assets, and the number of branches of each bank were obtained from Bankscope.

As we are studying the impact of the financial crisis on the changes in lending ability of the Pakistani banking sector, we divided our data into three samples: precrisis, crisis, and postcrisis. The precrisis period includes data from April 2006 to June 2007; the crisis period ranges from July 2007 to March 2009; and postcrisis is from April 2009 to March 2011. We follow Dungey and Gajurel (2015) in defining the crisis period. They use July 2007 to March 2009 as a crisis period and find that Pakistani banking sector experienced systematic contagion as well as idiosyncratic contagion during the global financial crisis.¹⁵

Because we are interested in the changes in lending before and after the crisis, we start with 100,938 loan observations from 23,188 borrowers.¹⁶ We discard all loans given by any other financial institution except 21 commercial banks, which gives us 71,621 observations. We also discard loans given as a nonfunded facility, which then leaves us with 59,183 observations. Some of our analysis contains all these observations. However, because of the firm fixed effects approach, we filter out the borrowers who borrowed only from one bank and include only those firms that borrowed from more than one bank prior to the crisis. In the next step, we consolidate the borrowing portfolios of all borrowers corresponding to the respective lenders, controlling for the various financial products under consideration.¹⁷ This gives us 33,245 observations; most of our analysis is made on this number of observations, from 11,524 borrowers.

The filtered dataset includes a total of 21 banking institutions in Pakistan. These include eight private, four privatized, two provincial, and two nationalized banks, two foreign banks, and three Islamic banks operating in Pakistan. Our dependent variable is the change in loan i of firm j from bank b (ΔL_ijb). We transform all the loan values to logarithmic form and obtain the difference between the logarithmic values of lending by banks prior to and after the shock, such that a positive value indicates that there has been an increase in lending post financial crisis and a negative value indicates that lending has decreased following the GFC. Our primary variables of interest are high, intermediate, and low exposure banks. As explained above, we classify these as follows: Banks that have over 8% of their borrowing from outside Pakistan are considered as high exposure, and those with between 4% and 8% and less than 4% borrowing from outside Pakistan are classified as intermediate exposure and low exposure banks, respectively. The higher the exposure of banks, the more they are expected to transmit financial shock to the local market. This classification is similar to that of Schnabl (2012), who classifies banks into these categories and then measures the impact of lending changes by each type of bank to firms in Peru. In Table 1, we provide the list of banks in our dataset along with their average market share measured as average total assets as a percent of average total assets of all the banks for the period from 2006 to 2011 in our dataset. We also provide average borrowing outside Pakistan as a percent of total borrowing for the period from 2006 to 2011. We observe international exposure of banks from zero to 22.3%, and there is enough variation in the sample to exploit the transmission of global financial crisis shock.

In addition to our core proxies, we use a number of control variables, which have been shown to influence the change in bank lending to real firms. We focus on bank and borrower/loan characteristics and follow the literature (see, e.g., Khwaja & Mian, 2008; Schnabl, 2012; Baele et al., 2014; Farooq & Zaheer, 2015) when selecting and measuring these controls. Kapan and Minoiu (2013) show that better capitalized and more liquid banks were better able to continue credit supply during the 2007–2008 crisis. Building on this finding, we use the bank capital ratio as a proxy for solvency and the liquid assets to total assets ratio as a proxy for bank liquidity. In addition, we add the natural log of the number of branches as a control variable, as the branch network is one of the factors leading to bank runs during a crisis (He & Manela, 2012). Furthermore, the size of bank may also affect changes in lending, because larger banks are typically considered to be more stable. We use total bank assets as a proxy for the size of the bank. For borrower/loan characteristics, we use size of loan as the proxy for the size of the borrower. In addition to loan size, we also consider other loan characteristics including loan maturity, whether the loan is secured and whether it is a performing loan. Loan maturity can have a negative impact on the change in lending if banks do not want to commit a large amount of funds over a long term. Similarly, security for the loan, and whether the loan is performing, is also important because they may have an effect on the lending ability of banks. Finally, we also use loan-type dummies, bank controls, and firm fixed effects for some regressions as robustness checks. Details of the variables used in the paper and their definitions are given in Table A2.

4 METHODOLOGY

This section outlines the econometric setting that we use to estimate the transmission of bank financial shocks to real firms. This estimation poses a key empirical identification problem arising from changes in credit supply by Pakistani banks and credit demand by Pakistani firms. For example, a reduction in lending by a bank to a firm may reflect a credit supply shortage in the bank, lower credit demand from the firm, or a combination of both. More specifically, if an export-oriented local firm (e.g., Nishat Mills) borrows more from a foreign bank (e.g., Citibank), and its export opportunities decrease following the global financial crisis, then Nishat Mills might reduce borrowing from Citibank, and Citibank's lending may decrease relative to a local bank (e.g., United Bank Limited). To address this potential identification problem, we follow Khwaja and Mian (2008) and Schnabl (2012) and use firm fixed effects and consider the fact that the same firm borrows from more than one bank. Our analysis controls for the changes in credit demand at the firm level by checking for loan relationships within firms, rather than across firms. Thus, if a firm is borrowing from both banks A and B, we check for the changes in borrowing from bank A relative to bank B by the same firm. Using this approach, we are able to identify the changes in credit demand by firms due to the changes in credit supply from banks.

We implement the following strategy for identification using the ordinary least squares (OLS) method. We start with baseline model of Schnabl (2012), which is as follows:

(1)

The dependent variable in Equation 1, ΔL_ijb, is the change in loan i of firm j from bank b, which shows the changes in borrowing by a firm from a bank prior to and after the financial shock. The right hand side of the above equation includes β₁, which are firm fixed effects that absorb all firm-specific credit demand shocks, whereas E_b denotes exposure of the bank to a financial shock. We identify bank exposure in three categories: high, intermediate, and low. In order to include the control variables and other variables of interest, we extend the baseline model 1 and estimate the following model 2:

(2)

β₂ in model 2 is the coefficient of interest and checks the transmission of a financial shock after controlling for credit-demand shocks at the firm level. The key assumption in the analysis is that all firms reduce their lending in equal proportions from all banks after the credit crunch. This assumption is plausible for both high and intermediate exposure banks due to their similar characteristics, but not for low exposure banks. Therefore, we control for bank characteristics (X_bt) in our analysis. The sample is restricted to firms that borrow from more than one bank prior to the shock. As the specification includes firm fixed effects, the coefficient of interest is calculated only for firms that borrow from more than one bank. In addition to the bank characteristics, we also control for borrower/loan characteristics (Y_it) and macroeconomic variables (Z_t) to capture any borrower/loan and macroeconomic factors that might have an impact on changes in loan i of firm j from bank b (ΔL_ijb).

5 EMPIRICAL RESULTS

6 CONCLUSION

As small developing countries may be more vulnerable to exogenous economic shocks due to inherent political instability and financial risk, it is important to analyse the impact of the financial crisis on real businesses in those economies. Moreover, in the absence of fully developed capital and financial markets, most local businesses rely heavily on bank financing in developing markets. The small and medium enterprises that form the core of emerging markets may also suffer due to the shortage of funds from banks and other financial institutions as a result of exogenous economic shocks.

Studying the Pakistani market as a case in point and capitalizing on the unique dataset of all corporate loans, we have identified changes in post-financial shock lending using various bank and loan level variables. We have shown that lending to small firms and lending by private banks decreased, irrespective of the fact whether they borrowed from low or high exposure banks. However, lending by relatively low exposure big banks increased after the global financial crisis. The findings regarding different types of loans revealed that agricultural loans increased after the financial shock (i.e., GFC), both from relatively low exposure banks. Our findings on information asymmetry showed that bank reputation as a proxy for information asymmetry was not able to overcome information asymmetry problems and that high exposure banks still reduced their lending after the financial shock. However, previous interactions of the borrower with relatively low exposure banks were able to overcome these problems and increased their lending after the GFC.

Our work is important from a regulatory perspective as it identifies that emerging markets are open to shocks from the international financial markets, despite their limited linkages with them; hence, mechanisms need to be developed to better cope with any such events in the future. This brings us to areas for further research; first, we would like to check the impact of financial shock on the interbank market of Pakistan, to see whether the liquidity crunch only affected firms, or also interbank liquidity. Another interesting potential area of research could be on the regulatory side, observing the current funding structures of banks in the country and devising a set of guidelines outlining limits to international exposure.