Liquidity shock management: Lessons from Australian banks

Abstract

Prior to the 2007–2009 financial crisis, international banks had an average share of around 65% of the syndicated loan market in Australia. When the crisis hit, the resulting liquidity shock resulted in globally active international banks exiting the Australian market. With limited global operations, the major Australian banks were able to absorb and manage the liquidity shock. This resulted in domestic banks carrying a significantly greater proportion of revolving credit facilities in their syndicated loan portfolios after 2008. Domestic bank willingness and ability to deal with the market disruption and to hold a greater proportion of high liquidity risk revolvers are directly linked to the level of their transaction deposits. Their increased involvement in revolving facilities cannot be fully explained by the certification effect or flight-to-home effect. It is not demand driven and is robust to endogeneity tests.

Keywords

Australian domestic banks competitive advantage liquidity risk syndicated loans

1. Introduction

Liquidity management is of vital importance in the operations, development and survival of a bank. The credit and liquidity crunch of 2007–2009 (Brunnermeier, 2009), which became known as the global financial crisis (GFC), severely tested bank liquidity risk management and provides an ideal research setting to investigate bank reactions to a global liquidity shock. This study focuses on the Australian syndicated loan market and provides evidence of the ability of Australian banks to withstand the systemic liquidity shock caused by the GFC. The results highlight how in a relatively concentrated market, domestic and international banks differ in their funding of revolving loans which carry higher liquidity risk.

The GFC has given rise to numerous studies that explore how banks cope with such an exogenous shock. On one hand, the literature on internal capital markets suggests that global banks should be better at absorbing liquidity shocks. Cetorelli and Goldberg (2012b) document that funds flow regularly between parent banks and their affiliates. Such flows are critical to liquidity management within an internationally operating bank especially when there is a funding shock. Cetorelli and Goldberg (2012a) establish that internal capital markets allowed international banks to absorb foreign-born shocks during the GFC, but in actively reallocating funds among their affiliates those banks contributed to the propagation of the liquidity shock globally.¹ This evidence suggests that foreign banks operating in Australia that are part of large international banking groups may be better able to absorb liquidity risks than domestic banks.

Another view maintains that because the GFC involved a global liquidity shock, international banks may have had to focus their resources on the parent banks. Allen et al. (2014) explore the international transmission of the liquidity shock among multinational banks before and after the GFC. Their results confirm that subsidiaries were dependent on interbank financing for their credit supply prior to the crisis. After the crisis, their deposit growth declined and subsidiaries were unable to rely on parent banks for support. This led to a reduction in their lending during the financial crisis. Jeon et al. (2013) find that a bank’s internal capital market transmits both favourable and adverse shocks to its subsidiaries. Frey and Kerl (2015) show that subsidiaries which relied on short-term wholesale funding were disadvantaged in the crisis, when interbank markets froze.

International banks operating in Australia would have faced similar problems during the GFC. Their deposit growth would have been stagnant and they would have had less support from parent banks during the crisis. Because Australian banks are not ‘international’ in the same sense as globally active banks, they were not called on to supply liquidity to off-shore subsidiaries and branches.² In addition, Australia was relatively insulated from the global shock (Stevens, 2009). For example, Guttmann and Rodgers (2015) find that no single bank in Australia showed any sign of significant liquidity risk during the crisis. For these reasons, we expect to find Australian domestic banks being better placed to deal with the liquidity crisis.

The Australian syndicated loan market during the period surrounding the GFC provides a setting different from that of the United States to investigate the effects of the global liquidity shock. Over the period studied, it had the seventh largest issuance of syndicated loans among developed countries (BIS, 2012), making it an economically important market. About 35% of the funds lent in the Australian syndicated loan market are supplied by the four largest domestic banks,³ with global banks providing the remaining funding. Non-bank funding and funding from smaller domestic banks are minimal.⁴ The Australian market, however, is not as concentrated as the US market where the top three banks account for over 50% of the market by volume (McCahery and Schwienbacher, 2010; Ross, 2010). The Australian syndicated loan market differs from that in the United States because of the absence of a viable secondary market. Consequently, domestic lenders must adopt a ‘buy and hold’ approach to their syndicated lending and their willingness to take on liquidity risk may be affected by this feature of the market. In addition, even though international banks with large internal capital markets may have been better equipped to manage the global liquidity shock, domestic banks were likely to have had liquidity advantages afforded by their large retail deposit bases.

Our findings can be summarised as follows. Access to domestic transaction deposits lowers liquidity risk for Australian domestic banks. This, coupled with their relatively low levels of global activities, allowed them to take on a greater proportion of (higher liquidity risk) revolving credit facilities in their syndicated loan portfolios over the period of 1996–2010. In the period after the GFC domestic banks operating as the lead arrangers in loan, syndicates took on significantly more liquidity risk compared to their international counterparts. Importantly, these findings are robust to controlling for other determinants of loan structure, whether domestic banks are acting as lead arrangers or participating banks in the syndicate, and also to various tests for endogeneity. The results reinforce the view that Australia was relatively insulated from the global liquidity shock and provide evidence of a channel through which the liquidity shock was absorbed.

The remainder of this article is organised as follows. Section 2 provides background on the institutional setting for the Australian syndicated loan market, describes the data used in the study and presents descriptive statistics. Section 3 describes the research approach and documents the results from univariate and multivariate analyses. Robustness tests are presented in section 4, while section 5 concludes the study.

2. Background and data

2.1. The syndicated loan market in Australia

Davis (2011) documents that the four largest commercial banks in Australia captured 76% of the domestic deposit market and 78% of the domestic loan market as at March 2010. Our data show that they funded nearly 35% of total Australian syndicated loans in 2011, representing over 98% of the syndicated loan funding provided by domestic banks. They attract large domestic retail deposits, whereas the vast majority of foreign lenders in the Australian syndicated loan market have a minimal share of Australian retail bank deposits.⁵

The Australian syndicated loan market represents a significant source of funds for corporate domestic borrowers. Figure 1 shows that in 2007, total loan syndications extended to non-financial domestic firms peaked at just over $100 billion,⁶ representing nearly 30% of total commercial lending. While the Australian and global syndicated loan markets saw a significant drop in issuance post-2007, total dollar volume originated for non-financial firms in the Australian syndicated loan market exceeded $80 billion in 2011. With an underdeveloped corporate bond market, the syndicated loan market is an important source of funds for Australian corporations.⁷

Figure 1.

Australian syndicated loan market.

Figure 1 also documents the domestic bank share of the syndicated loan market over the period of our study. Foreign lenders in the market are mostly banks. Within our sample, non-bank foreign lenders funded less than 10% of syndicated facilities in 2010. Both non-bank participation and secondary trading are economically unimportant in the Australian market. Lenders therefore must hold loans until maturity, resulting in important funding and risk consequences for participating banks.

2.2. Data

Historical data on Australian syndicated lending are provided by two companies, Euromoney Institutional Investor PLC’s Dealogic and Thomson Reuters’ Loan Pricing Corporation (LPC), which, respectively, distribute the Loan Analytics and Dealscan databases. Information includes pricing and non-pricing contract terms, as well as lender characteristics for loans to large corporate borrowers from 1993 to the present. Although providing loan information to the market is not a disclosure requirement, financial institutions have a strong incentive to advertise their lending practices. So-called League Tables, which are closely followed by analysts and prospective borrowers,⁸ contain information provided by the lenders including the size of the loan, their role within the syndicate and the proportion held by each lender. Nevertheless, syndicated loan data are not complete and offer researchers only a random sample of market activity (Strahan, 1999).

US studies tend to use the Dealscan database. In this study, we merge the two databases because the number of Australian syndicated loans is significantly less than that in the United States and the databases contain more missing variables for Australian loans. Merging the two databases enables us to maximise the number of observations. Six key variables are used to merge the two databases: the date of the deal, the borrower’s Australian Securities Exchange (ASX) ticker, deal size, tranche size, maturity and the number of lenders.

Our sample includes loans extended to Australian borrowers between 1996 and 2010 from the two databases. Note that borrowers classified as financial institutions are excluded from the sample because the historical leverage ratios of financial institutions bias the observed interaction between borrower financials and loan contract terms. Loans to governments may not fully reflect market fundamentals and are also excluded. Although the databases commence in 1993, prior to 1996, the Australian Federal Government was providing funding to corporations through participation in the syndicated loan market. Their lending entity, the Australian Industry Development Corporation (AIDC), had legislative rights to borrow up to $250 million and lend to corporations in order to support future economic growth.⁹ In 1995, the AIDC was fully privatised, removing government involvement from the Australian market. We therefore collect data from 1996 onwards, retaining only those loans denominated in USD and AUD in order to obtain a consistent sample.¹⁰ All USD loans are converted to AUD using the previous month’s average exchange rate.

Each tranche of the loan is matched with borrower characteristics including rating, size, z-score and leverage. The number of observations is substantially reduced because borrower characteristics are available only for listed firms, and only around 40% of Australian syndicated loans are to listed firms. In addition, we require a loan to report loan characteristics such as price, maturity and amount to be included in the sample. The final sample for the main regression is 616 tranches from 331 deals.

Aspect Huntley’s FinAnalysis database is used to obtain accounting data for ASX listed firms. Macroeconomic data are obtained from the Thomson Reuters’ Datastream database. Information on historical credit spreads is used to capture wider economic impacts on loan terms and structure. Finally, the Australian Prudential Regulatory Authority’s (APRA) monthly banking statistics are used to provide the assets and liabilities of Approved Deposit-Taking Institutions (ADIs) from 1996 to 2010.

2.3. Descriptive statistics

Descriptive statistics for Australian syndicated loans and borrowers between 1996 and 2010 are reported in Table 1. Table 1 documents 616 common sample observations.¹¹ A detailed description of all variables used in this study is provided in Appendix 1.

Table 1.

Descriptive statistics. This table reports descriptive statistics for Australian syndicated loan and borrower information between 1996 and 2010. The total sample consists of 616 observations. Loan and borrower characteristics are reported in Panel A and Panel B, respectively. Market condition variables are reported in Panel C. Syndicate structure and composition variables are reported in Panel D. Deal, tranche and firm size are reported in millions of AUD. Maturity of the syndicated loans is reported in months. For all other variable definitions, see Appendix 1.

Variable	Observation	Mean	Standard deviation	Min	Max
Panel A: Loan characteristics
DEAL ($million)	616	1336	3907	2	55,000
TRANCHE ($million)	616	519	1357	1	20,000
MAT (months)	616	50.85	33	6	312
SECURED	616	0.1948	0.3964	0	1
REV	616	0.4919	0.5003	0	1
Panel B: Borrower characteristics
UNRATED	616	0.6250	0.4845	0	1
FIRMSIZE ($million)	616	8604	16467	38	149,348
LEV	616	0.2818	0.1914	0.0009	0.9063
ZSCORE	616	1.2143	1.1725	−3.9745	11.1633
Panel C: Market conditions
CS (%)	616	1.2470	0.6037	0.0014	3.0643
STP (%)	616	0.2501	0.4101	−0.5455	1.1698
IV	616	0.0970	0.0265	0.0599	0.1985
Panel D: Loan composition
DOMESTIC	616	0.2096	0.2602	0	1
DOMESTIC_LA	616	0.1367	0.2139	0	1
DOMESTIC_PART	616	0.0729	0.1371	0	0.9

Panel A documents loan characteristic variables. The average deal amount is $1.336 billion and the average tranche size is $519 million. Maturity, reported in months, has an average value just over 4 years. Only 20% of reported loans are secured and about half of the loans are structured as revolving loans. Panel B reports the characteristics of borrowers. The average borrowing firm has a market capitalisation of $8.6 billion. Although only listed corporations are examined, 62% of firms do not report a rating for senior debt. The average leverage ratio of borrowers is about 28% and the average z-score is 1.2.

Panels C and D provide descriptive statistics for market conditions and syndicate structure variables. The average credit spread during the sample period is about 1.25%, while the average term premium is 0.25%. On average, the domestic banks fund about 21% of each syndicated loan, made up of 14% as lead arranger and 7% as participating bank.

3. Approach and empirical analysis

3.1. Regression models

Our primary focus is to explore whether the domestic banks have a liquidity advantage especially in the aftermath of the GFC. We start by testing our research question on the entire sample then proceed to splitting the sample into pre- and post-GFC. Following Gatev and Strahan (2006), we estimate the following ordinary least squares (OLS) models

D O M E S T I C_{i} = α_{1} + β_{1} R E V_{i} + γ_{1} X_{i} + ε_{1}

(1)

D O M E S T I C_L A_{i} = α_{2} + β_{2} R E V_{i} + γ_{2} X_{i} + ε_{2}

(2)

D O M E S T I C_P A R T_{i} = α_{3} + β_{3} R E V_{i} + γ_{3} X_{i} + ε_{3}

(3)

The subscript i indexes the loan tranche. The dependent variable DOMESTIC_i is defined as the fraction of the syndicated loan tranche held by Australian commercial banks. Variations in the dependent variable, DOMESTIC_LA and DOMESTIC_PART, are also constructed, which, respectively, capture the proportion of the loan held by domestic banks when they are lead arrangers or participant lenders.

The key independent variable REV is a binary variable which takes unit value if the facility is a revolving credit facility, loan commitment, 365-day facility or a bridge loan in either the Loan Analytics or Dealscan database. These represent loan types with heightened liquidity risk for the lender. Therefore, a positive and significant coefficient on REV indicates that the proportion of the loan funded by domestic banks is positively correlated with liquidity risk. X_i is a vector of control variables including firm- and loan-specific factors and market conditions. The firm-specific factors include information asymmetry proxied by size (LN_SIZE) and whether the firm is rated (UNRATED) as well as credit risk proxied by Altman’s (1968) z-score (ZSCORE), the market-to-book ratio (MKTBK), the leverage ratio (LEV) and fixed assets (FIXD_ASS). Following Dennis et al. (2000), all dollar values of firm characteristics are adjusted for inflation using the CPI index. Loan-specific variables include the size of the loan (LN_TRANCHE), a currency dummy (FX) as well as the maturity in months (LN_MAT) and fixed purpose effects (ACQLBO and PROJFIN). Finally, market conditions include GFC, ACC, DEPIN and WGS, and macro-economic (CS) variables are included to account for full model specification. All variables are fully described in Appendix 1.

3.2. Univariate analysis

We first examine the differences in characteristics between revolving credit facilities and term loans. Table 2 presents univariate analysis of the difference of means for revolving and term loans.

Table 2.

Univariate analysis. This table compares the mean differences of loans which are issued as a revolving credit facility from those issued as term loans. Loan and borrower characteristics are reported in Panel A and Panel B, respectively. Market condition variables are reported in Panel C. Syndicate structure and composition variables are reported in Panel D. Deal, tranche and firm size are reported in millions of AUD. Maturity of the syndicated loans is reported in months. For all other variable definitions, see Appendix 1. The p values are reported in the last column for the differences between revolving credit facilities and term loans. The data period is 1996–2010.

Variable	Revolving credit facility			Term loan			p value
Variable	Observation	Mean	Standard deviation	Observation	Mean	Standard deviation	p value
Panel A: Loan characteristics
DEAL	303	1577	4481	313	1105	3248	0.1343
TRANCHE	303	577	1488	313	463	1219	0.3007
MAT	303	40	19	313	61	41	0.0000
SECURED	303	0.1254	0.3317	313	0.2620	0.4404	0.0000
Panel B: Borrower characteristics
UNRATED	303	0.5545	0.4978	313	0.6933	0.4619	0.0004
FIRMSIZE	303	10053	18662	313	7202	13908	0.0316
LEV	303	0.2712	0.1933	313	0.2920	0.1894	0.0785
ZSCORE	303	1.3204	0.9924	313	1.1116	1.3173	0.0270
Panel C: Market conditions
CS	303	1.3031	0.6214	313	1.1928	0.5818	0.0232
STP	303	0.2850	0.3775	313	0.2163	0.4374	0.0376
IV	303	0.1000	0.0014	313	0.0941	0.0015	0.0000
Panel D: Loan composition
DOMESTIC	303	0.2499	0.2604	313	0.1706	0.2544	0.0001
DOMESTIC_LA	303	0.1563	0.2100	313	0.1177	0.2162	0.0250
DOMESTIC_PART	303	0.0935	0.1568	313	0.0529	0.1115	0.0002

The results in Panel A show no statistical differences in terms of the size of the deal or tranche size between revolving and term loans. The term to maturity of revolving loans is significantly shorter which reflects the nature of the loan. Revolving loans are also less likely to carry collateral. In terms of borrower characteristics, Panel B results show that revolving loans are on average granted to larger, higher quality borrowers.

The results in Panel C suggest that revolving loans are initiated under market conditions reflecting greater credit risk (a greater difference between the long-term corporate bond rate and the long-term treasury bond rate) and where the yield curve is steeper (a greater difference between the 10-year government rate and the 2-year government rate). Both these differences are significant at the 5% level. The results also show that revolving loans tend to be initiated when interest rate volatility is higher (significant at the 1% level). Panel D documents that the fraction retained by the domestic banks is greater for revolving loans generally and where the domestic bank is a participant bank (at 1% level of significance). The fraction retained when the domestic bank is the lead arranger is greater for revolving loans (at a 5% level of significance).

3.3. Liquidity risk and syndicate composition

We now investigate the relation between liquidity risk and the role that domestic banks play in the syndication process, while controlling for borrower characteristics and market variables that may affect the relation. Table 3 presents the results of testing the relation between liquidity risk (REV) and syndicate composition (DOMESTIC) as described by equations (1) to (3).

Table 3.

Liquidity risk and syndicate composition. This table presents OLS estimation of the proportion of the loan held by Australian domestic banks. The model is operationalised as follows: ${DOMESTIC}_{i} = α + β_{1} {REV}_{i} + γ {LC}_{i} + γ {BC}_{i} + γ {CONTROL}_{i} + ε_{i}$ . The dependent variable DOMESTIC is the percentage of the loan amount held by domestic banks. The independent variable of interest is REV, a binary variable which takes unit value if a loan is a revolving credit facility and zero otherwise. LN_SIZE represents the natural logarithm of tranche size. MAT is the number of months to maturity of the loan. SECURED is a dummy variable taking the value of 1 if the loan has collateral and zero otherwise. FS is calculated as (100 × (Total Assets + Market Capitalisation − Total Common Equity)/CPI). It is a proxy for real firm size (Dennis et al., 2000). LEV is the leverage ratio of the borrowing firm, calculated as Total Debt/(Total Assets + Market Capitalisation − Total Common Equity). ZSCORE is the z-score of the borrowing firm, calculated as (3.3 × EBIT/Sales + 1.0 × Sales/Total Assets + 1.4 × Retained Earnings/Total Assets + 1.2 × Working Capital/Total Assets). It is a proxy for default risk (Dennis et al., 2000). UNRATED is a dummy variable which takes unit value if the borrower’s senior debt is unrated and zero otherwise. CS is the credit spread, calculated as the 13-year Australian corporate bond rate minus the 10-year Australian government bond rate, at the end of the month prior to when the loan was issued. STP is a straight-term premium calculated as the 10-year government bond yield minus the 2-year government bond yield. IV is the interest rate volatility in the year in which the loan was extended. Standard errors are clustering adjusted. ***, **, *significance at 1%, 5% and 10% levels, respectively.

	DOMESTIC (1)	DOMESTIC_LA (2)	DOMESTIC_PART (3)
REV	0.0673*** (2.494)	0.0372* (2.062)	0.0301** (1.425)
LN_SIZE	−0.136*** (0.031)	−0.082*** (0.025)	−0.054*** (0.014)
MAT	7.615** (3.689)	7.331** (3.206)	0.284 (1.664)
SECURED	−1.698 (1.080)	−0.711 (0.911)	−0.987* (0.573)
FS	13.084* (7.187)	12.582** (5.811)	0.502 (6.792)
LEV	0.508 (1.006)	−0.333 (0.747)	0.840 (0.575)
ZSCORE	5.304* (2.935)	5.292** (2.452)	0.012 (1.666)
UNRATED	0.643 (2.286)	4.288** (1.885)	−3.645*** (1.236)
CS	2.965 (4.323)	2.640 (3.379)	0.325 (2.421)
STP	−77.055 (65.525)	−91.135** (46.060)	14.080 (41.095)
IV	5.779 (5.434)	2.716 (4.984)	3.063 (2.185)
Wcap	5.674 (5.910)	3.664 (5.500)	2.010 (2.398)
Refinance	−2.327 (6.094)	−4.008 (5.508)	1.681 (2.433)
ACQLBO	47.013* (25.958)	24.816 (22.191)	22.198* (12.659)
Constant	6.729*** (2.494)	3.720* (2.062)	3.009** (1.425)
Observation	616	616	616
Adjusted R²	0.0840	0.0882	0.0591
F-statistic	5.111	4.385	4.313

Results in the first column show that consistent with the univariate results, domestic banks hold a 6.73% greater (significant at 1%) proportion of revolving loans than international institutions in their syndicated loan portfolios. There are two possible explanations for this observation. Domestic banks may be willing to take on greater exposure to liquidity risk due to their competitive advantage arising from domination of the domestic deposit market, while foreign banks may have chosen lower liquidity exposure in Australia to focus on their home market. An alternative explanation is that domestic banks provide the additional monitoring required for revolving facilities and certify the quality of the loans. This would result in domestic banks holding a greater proportion of liquidity risky revolvers than other participating banks within the syndicate. To explore the alternative explanation, we separate DOMESTIC into DOMESTIC_LA, the fraction of domestic banks acting as lead arrangers in the syndicate and DOMESTIC_PART, the fraction of domestic banks acting as participant banks. If the alternative explanation holds, we should observe an increase in exposure of the domestic banks only when they are the lead bank because lead arrangers are usually responsible for monitoring and only lead arrangers can credibly certify the quality of the loan.

The second and third columns in Table 3 provide an empirical examination of this question. When they are lead arrangers, domestic banks hold a 3.72% greater (significant at 10%) proportion of their portfolio as revolvers than international institutions. Moreover, when acting as participant banks in the syndicate, domestic banks hold a 3.01% greater (significant at 5%) proportion than international institutions in revolving loans. Thus, the positive relation between liquidity risk and syndicate composition remains relatively robust for both DOMESTIC proxies. Given that DOMESTIC = DOMESTIC_LA + DOMESTIC_PART, the coefficient of DOMESTIC is expected to be equal to the sum of the other two coefficients. The fact that the coefficient of DOMESTIC_PART is statistically significant indicates that the relative preference for revolving credit facilities by domestic banks over international banks is unlikely to be driven by a demand for certification.

Control variables worth noting include loan size, maturity and borrower opacity. The fraction held by domestic banks reduces significantly as loans become larger; larger loans often require more lenders in the syndicate, and on average, this should reduce the share of each participant including domestic banks. When acting as lead arrangers, domestic banks retain a greater proportion for longer maturity loans, possibly reflecting the informational advantage that domestic banks have over international banks due to their relationship lending with domestic borrowers. Another interesting result relates to borrower rating. Domestic banks appear to retain a greater proportion of unrated borrower loans but only when they are the lead arranger, reflecting their willingness and ability to monitor risky borrowers. Unsurprisingly, they retain a significantly lower proportion of unrated loans when they are not the lead arranger and therefore have little incentive to monitor. Overall, the results indicate that the increase in exposure to liquidity risk of the domestic banks is driven by their competitive advantage in managing liquidity risk. In the following section, we explore if this apparent liquidity advantage is magnified by the exogenous liquidity shock from the GFC.

3.4. Liquidity risk and syndicate composition pre- and post-GFC

Cetorelli and Goldberg (2012a, 2012b) suggest that because of their internal capital markets, global banks should be better at absorbing a liquidity shock than domestic banks. However, domestic banks in Australia have a significant liquidity advantage arising from their strong domestic retail deposit base. Although international banks operating in the Australian market do not have a meaningful market share of Australian retail deposits,¹² large foreign commercial banks may be able to use their home country domestic deposits to fund syndicated loans in Australia. However, during the liquidity and credit crunch of 2007–2008, global US dollar illiquidity led to dislocation in the foreign currency swap market and made it increasingly difficult for foreign banks to fund syndicated loans in off-shore markets.¹³

Adding to these arguments, recent empirical work has highlighted that in syndicated loan markets plagued by increasing systemic risk, foreign lenders will pull funding from off-shore markets (Giannetti and Laeven, 2012). Chui et al. (2010) note the unwillingness of international banks to participate in loan syndicates during the 2008 crisis and offer support for the view that concerns about the soundness of large international banks, together with resulting funding pressures constrained the supply of syndicated loans. De Haas and Van Horen (2012) corroborate this finding by showing that banks suffering a sharp decline in market to book ratio, and those that had to write down subprime assets and refinance large volumes of long-term debt, severely curtailed their lending abroad. Ivashina and Scharfstein (2010) show that during the 2008 crisis, banks with less access to deposit financing and at greater risk of credit-line drawdowns reduced lending more than other banks.

We therefore expect to see differences in the provision of liquidity-intensive loans between domestic and international banks during the 2008 crisis. Table 4 presents the results when we split the sample into loans before and after 2008. We choose the year 2008 because the Australian equity market peaked in November 2007 and by December 2008 had dropped by around 46% (Brown and Davis, 2010).

Table 4.

Liquidity risk and syndicate composition pre- and post-2008. This table presents the multivariate OLS estimation of the proportion of the loan held by the largest four Australian commercial banks given the loan is a revolving credit facility. The model is operationalised as follows: ${DOMESTIC}_{i} = α + β_{1} {REV}_{i} + γ {LC}_{i} + γ {BC}_{i} + γ {CONTROL}_{i} + ε_{i}$ . The dependent variable DOMESTIC is the proportion of the loan amount held by domestic banks. The independent variable of interest REV is a binary variable which takes unit value if a loan is a revolving credit facility and zero otherwise. LN_SIZE represents the nature logarithm of tranche size. MAT is the number of months to maturity of the loan. SECURED is a dummy variable taking the value of 1 if the loan has collateral and zero otherwise. FS is calculated as (100 × (Total Assets + Market Capitalisation − Total Common Equity)/CPI). It is a proxy for real firm size (Dennis et al., 2000). LEV is leverage ratio of the borrowing firm; it is calculated as Total Debt/(Total Assets + Market Capitalisation − Total Common Equity). ZSCORE is the z-score of the borrowing firm calculated as (3.3 × EBIT/Sales + 1.0 × Sales/Total Assets + 1.4 × Retained Earnings/Total Assets + 1.2 × Working Capital/Total Assets). It is a proxy for default risk (Dennis et al., 2000). UNRATED is a dummy variable which takes unit value if the borrower’s senior debt is unrated and zero otherwise. CS is the credit spread, calculated as the 13-year Australian corporate bond rate minus the 10-year Australian government bond rate, at the month prior to when the loan was issued. STP is a straight-term premium calculated as 10-year government bond yield minus 2-year government bond yield. IV is interest rate volatility in the year in which the loan was extended. Standard errors are clustering adjusted. ***, **, *significance at 1%, 5% and 10% levels, respectively.

	Post-2008			Pre-2008
	DOMESTIC (1)	DOMESTIC_LA (2)	DOMESTIC_PART (3)	DOMESTIC (4)	DOMESTIC_LA (5)	DOMESTIC_PART (6)
REV	0.122*** (4.367)	0.122*** (3.705)	0.000 (1.841)	0.0439 (2.962)	0.0001 (2.316)	0.0438** (1.769)
LN_SIZE	4.983** (2.020)	4.772*** (1.650)	0.211 (0.928)	0.451 (1.561)	−1.231 (1.432)	1.682*** (0.528)
MAT	−0.331** (0.148)	−0.250* (0.130)	−0.081 (0.051)	−0.128*** (0.032)	−0.083*** (0.026)	−0.045*** (0.014)
SECURED	12.948** (5.829)	10.617** (5.272)	2.330 (2.889)	4.284 (4.364)	4.512 (3.717)	−0.228 (1.831)
FS	−6.414** (2.592)	−4.928** (2.403)	−1.486 (1.006)	−0.336 (1.183)	0.641 (0.903)	−0.976 (0.680)
LEV	−9.905 (18.376)	8.010 (16.283)	−17.915** (6.939)	19.110** (7.437)	14.897** (5.784)	4.213 (7.304)
ZSCORE	−0.928 (2.955)	−3.212 (2.403)	2.284 (1.552)	0.161 (1.082)	−0.403 (0.829)	0.564 (0.581)
UNRATED	−3.292 (4.959)	−0.690 (4.358)	−2.602 (2.116)	8.886** (3.687)	7.778*** (2.850)	1.108 (2.109)
CS	3.066 (6.444)	1.577 (5.758)	1.489 (2.707)	−3.508 (4.580)	−1.210 (3.314)	−2.298 (2.171)
STP	3.850 (8.199)	7.013 (7.037)	−3.163 (3.878)	−0.359 (5.859)	−3.050 (4.104)	2.691 (3.353)
IV	92.201 (248.277)	15.863 (220.153)	76.338 (116.055)	−63.058 (72.354)	−66.234 (46.947)	3.175 (46.993)
Wcap	21.615* (11.228)	20.922** (9.484)	0.693 (5.176)	3.488 (5.727)	−0.930 (4.997)	4.418* (2.296)
Refinance	17.955 (11.259)	17.171* (10.296)	0.784 (4.837)	5.706 (6.251)	2.437 (5.518)	3.269 (2.669)
ACQLBO	16.042 (15.126)	15.145 (13.262)	0.897 (7.214)	−4.740 (6.140)	−6.299 (5.252)	1.559 (2.479)
Constant	110.043 (67.989)	79.269 (59.905)	30.774 (24.655)	24.935 (29.217)	8.118 (22.878)	16.817 (15.613)
Observation	164	164	164	452	452	452
Adjusted R²	0.114	0.141	0.0550	0.0908	0.0954	0.0799
F-statistics	10.60	6.602	0.800	3.825	3.356	4.281

If commercial banks in Australia are using their balance sheets to fund liquidity-intensive instruments, then their relative percentage involvement in syndicated deals should rise during the period when international banks were exposed to the liquidity and credit crunch. Results from Table 4 (column 4) suggest that prior to 2008, domestic and international banks held similar proportions of revolvers in their syndicated loan portfolios. However, after 2008, domestic banks had a clear advantage over international banks in funding revolving facilities as indicated by the positive and significant (at 1%) coefficient on REV. After 2008, domestic banks hold a 12.2% greater proportion of their syndicated loan portfolio as revolving loans as compared to international institutions. The argument that domestic banks have a liquidity advantage is made even more convincing by the fact that the entire 12.2% difference is driven by domestic banks acting as lead arranger (column 2).

If this outcome results solely from foreign lenders withdrawing from the market in a flight-to-home effect (Giannetti and Laeven, 2012), there should be a similar increase in the proportions held by domestic banks across both revolving and term loans. It is clear from the results presented in columns (1) and (2) that domestic bank participation increases more for revolving loans than for term loans post-2008. However, before 2008, there is no statistical difference in the percentage of revolver versus term loans between domestic and international banks. We therefore conclude that this 12.2% greater proportion of revolving loans in domestic bank portfolios is over and above the flight-to-home effect. Moreover, the domestic banks’ increased share of the market post-2008 is driven entirely by their role as lead arrangers since the coefficient on DOMESTIC_PART post-2008 is zero. Reflecting their ability to maintain their balance sheet liquidity as foreign banks withdrew funding from the Australian market, domestic banks were able to absorb both credit and liquidity risk when systemic risk was increasing.

4. Robustness tests¹⁴

4.1. Liquidity risk management

Recent studies have provided additional theoretical explanations for the unique ability of commercial banks to manage liquidity risk. Kashyap et al. (2002) show that when liquidity demands are not highly correlated between borrowers (loans) and lenders (deposits), a deposit taking institution can reduce their costly capital buffers by servicing both markets. In other words, when a commercial bank has greater access to transaction deposits, it will be able to increase its liquidity risk by extending revolving lines of credit. Antinolfi and Prasad (2008) examine bank liquidity risk and find that financial intermediaries pool capital from depositors to increase liquidity in financial markets.

The increase in domestic bank exposure to revolving loans is shown in section 3 to be consistent with domestic banks actively seeking more liquidity exposure and is not the result of a certification requirement by foreign banks. We provide direct tests of liquidity management by running pooled OLS regressions (Gatev and Strahan, 2009; Kashyap et al., 2002) to estimate if lagged changes in transaction deposits are driving the propensity of domestic banks to fund liquidity-intensive syndicated loans. In untabulated results supporting our main findings, we find that the decision to fund revolving credit facilities by commercial banks in Australia is significantly (at the 5% level) affected by relative changes in the liquidity of their balance sheets and is not driven purely by their market dominance over foreign banks. The regression results also show (consistent with the results in section 3) that during the GFC, domestic banks increased their revolving facilities to an even greater extent.

4.2. Endogeneity tests

An important issue with this literature is the potential endogeneity between the dependent variable and the primary independent variable. In our case, endogeneity between the variable DOMESTIC and the binary choice of term or revolving loan is a potential area of concern. Including a dummy for revolving loans in an OLS regression may not account for the non-random distribution of loan type. Put another way, certain types of borrowers may be more likely to borrow using revolving loans rather than term loans. Effectively, there are unobserved borrower and lender characteristics which influence the choice of loan type. For example, Maskara (2010) argues that revolving facilities are generally awarded to ‘safe’ borrowers. If this is the case, the increase in domestic lenders observed for revolving loans may simply be a reflection of domestic banks choosing to increase their exposure to safer borrowers. Another cause of endogeneity may be the difference in demand for revolving credit facilities. Adverse selection theory suggests that firms with severe asymmetric information problems may use a revolving credit line to signal their quality. Hence, if domestic banks serve that group of borrowers, the observed outcome of higher revolving credit facilities among Australian lenders may simply be demand driven. We address the issue of endogeneity using the propensity score (PS) matching test.

Under laboratory conditions, one would be able to identify every pair of loans from two identical borrowers (or even from the same borrower), one of which is a term loan and the other a revolving loan. The observed difference in DOMESTIC across all these pairs of loans would provide an accurate estimation to answer our research question. While it is not possible to conduct such an exercise in reality, the PS matching technique of Heckman et al. (1997, 1998) provides a robust econometric technique to achieve the same outcome. Bharath et al. (2011) adopt this technique to address the endogeneity between relationship and loan price. Saunders and Steffen (2011) also adopt the PS matching approach to address the endogeneity between borrower listing status and loan price, reporting that the PS matching approach yields robust results when compared to other techniques.

We first estimate the probability of each individual loan being a revolving loan, as a function of the observed loan and borrower characteristics using a Probit model. We then estimate the probability of each loan being a revolving loan to produce a PS, which is matched with that of a similar term loan to examine the difference in the variable DOMESTIC between the two loan types. We find the results using PS matching to be consistent with those reported in Table 3 using OLS, adding further support to the conclusion that domestic banks take on higher exposure to revolving facilities due to their advantage over international banks in having better access to domestic retail funding.

5. Conclusion

This article investigates the risk-taking/management behaviour of the Australian banks in relation to liquidity risk in the Australian syndicated loan market. Overall, we find strong evidence supporting the idea that in the Australian syndicated loan market, domestic banks are more willing than international banks to take on liquidity risk after the GFC. Compared to international banks, domestic banks hold 12.2% more exposure to higher liquidity risk revolving loans in their syndicated loan portfolios following the GFC. This statistically significant result is entirely driven by the domestic banks acting as the lead arranger in the syndicate. Our results also suggest that this outcome is driven by the retail deposits of domestic banks and not by the demand of foreign banks for certification of domestic borrowers. Our results confirm that over and above any flight-to-home bias, domestic banks significantly increased their exposure to liquidity-intensive loans during the GFC.

This article contributes to the existing literature by examining loan-specific risk as a determinant of syndicate composition. The Australian syndicated loan market is similar to the market in the United States in that it is a concentrated market dominated by the four largest domestic banks. It differs from the US market because there is virtually no secondary market and banks must generally hold the loans issued on their balance sheets. Prior studies seek to explain trends in syndicate composition as a function of varying risk aversion among different lenders. The empirical findings of this article suggest that it may be the ability of lenders to absorb risk, rather than their risk aversion, which drives syndicate composition in the Australian market. These findings contribute to empirical research which attempts to explain how syndicates form based on the risk-return profiles of lenders.

While globally active banks may be able to absorb liquidity shocks using internal capital markets, small domestic-focused banking systems with strong deposit bases can also be very effective in managing liquidity shocks. This is a very important lesson from the results of this study. Our findings also have implications for policy makers. The propensity of domestic banks to lend using revolving credit facilities may expose the capital market to systemic risks. The efforts of the Australian Federal Government to shore up the domestic banking system during the GFC clearly had a positive influence on maintaining liquidity in Australian debt markets. It remains to be seen whether that approach will work in a future liquidity crisis.

Footnotes

Appendix

Appendix 1.

Variable definitions.

Variables	Definition	Source
ACC	ACC is a binary variable which takes unit value if the loan was issued in 1996 or 1997 during the Asian Currency Crisis (Esho et al., 2007)	RBA
AISD	The total amount paid by the borrower in basis points for the syndicated loan above the reference rate. It includes all fees and charges as well as interest rates	Loan Analytics and Dealscan
ACQLBO	Variable takes a unit value if the facility is used for an acquisition or leveraged buyout and zero if otherwise	Loan Analytics and Dealscan
CS	The long-term average corporate bond spread minus the long-term treasury bond spread. It is a proxy for macro-economic conditions and the term structure of interest rates	Datastream
DEP/ASS	Total deposits scaled by total assets on a monthly basis for each of the Big4 Australian banks	APRA
DEPIN	DEPIN is a binary variable which takes unit value if the loan was issued during the Australian Government’s federal deposit insurance scheme and zero otherwise	RBA
DOMESTIC	The fraction of the loan retained by the Australian Big4 commercial banks (Gatev and Strahan, 2009)	Loan Analytics and Dealscan
DOMESTIC_LA	The fraction of the loan retained by the Australian Big4 commercial banks when they are lead arrangers only (Gatev and Strahan, 2009)	Loan Analytics and Dealscan
DOMESTIC_PART	The fraction of the loan retained by the Australian Big4 commercial banks when they are participant lenders only (Gatev and Strahan, 2009)	Loan Analytics and Dealscan
FS	The logarithm of the following borrower information: (100 × (Total Assets + Market Capitalisation − Total Common Equity)/CPI). It is a proxy for real firm size (Dennis et al., 2000)	FinAnalysis Datastream
FX	FX is a binary variable which takes unit value if the loan was issued in foreign currency and zero otherwise	Loan Analytics and Dealscan
GFC	The 2007–2008 credit and liquidity crunch became known as the global financial crisis (GFC) in Australia. GFC is a binary variable which takes unit value if the loan was issued in 2008 during the GFC and zero otherwise	RBA
IV	Interest rate volatility for the year in which the loan was extended. We used overnight interbank lending rate to calculate this variable	Datastream
LEV	Leverage ratio of the borrowing firm; it is calculated as follows: Total Debt/(Total Assets + Market Capitalisation − Total Common Equity)	FinAnalysis
LN_SIZE	Logarithm of the size of the tranche, recorded in millions of AUD	Loan Analytics and Dealscan
MAT	The maturity in months of the loan	Loan Analytics and Dealscan
PROJFIN	Variable takes a unit value if the facility is used for project finance and zero otherwise	Loan Analytics and Dealscan
Refinance	Variable takes a unit value if the facility is used for refinance and zero if otherwise	Loan Analytics and Dealscan
REV	Variable takes a unit value if the facility is a line of credit, loan commitment, 365-day facility, revolving credit line or a bridging loan and zero if otherwise. It is a proxy for liquidity risk (Gatev and Strahan, 2009)	Loan Analytics and Dealscan
STP	Straight term premium calculated as 10-year government bond yield minus 2-year government bond yield	Datastream
UNRATED	Variable takes a unit value if the borrower’s senior debt is not rated and zero otherwise	Loan Analytics and Dealscan
Wcap	Variable takes a unit value if the facility is used for working capital and zero if otherwise	Loan Analytics and Dealscan
WGS	WGS, a binary variable takes unit value if the loan was issued during the Australian Government’s wholesale funding guarantee scheme and zero otherwise	RBA
ZSCORE	The z-score of the borrowing firm. It is calculated as follows: (3.3 × EBIT/Sales + 1.0 × Sales/Total Assets + 1.4 × Retained Earnings/Total Assets + 1.2 × Working Capital/Total Assets). It is a proxy for default risk (Dennis et al., 2000)	FinAnalysis

Final transcript accepted 5 June 2016 by Kathleen Walsh (AE Finance).

Funding

The author(s) received no financial support for the research, authorship and/or publication of this article.

Notes

References

Allen

Hryckiewicz

Kowalewski

et al . (2014) Transmission of financial shocks in loan and deposit markets: Role of interbank borrowing and market monitoring. Journal of Financial Stability 15: 112–126.

Altman

(1968) Financial ratios, discriminant analysis and the prediction of corporate bankruptcy. Journal of Finance 23: 589–609.

Antinolfi

Prasad

(2008) Commitment, banks and markets. Journal of Monetary Economics 55: 265–277.

Bailey

Uffelen

Wood

(2012) International activities of Australian banks. RBA Bulletin, December Quarter, pp. 47–55.

Bharath

Dahiya

Saunders

et al . (2011) Lending relationships and loan contract terms. Review of Financial Studies 24: 1141–1203.

BIS (2012) Securities statistics and syndicated loans. Available at: http://www.bis.org/statistics/secstats.htm

Black

Kirkwood

Rai

et al . (2012) A history of Australian corporate bonds. Reserve Bank of Australia Research discussion paper RDP 2012-2009, December. Sydney, NSW, Australia: Reserve Bank of Australia.

Brown

Davis

(2010) Australia’s experience in the global financial crisis. In: Kolb

(ed.) Lessons from the Financial Crisis: Causes, Consequences, and Our Economic Future. Hoboken, NJ: John Wiley & Sons, Inc., pp. 537–544.

Brunnermeier

(2009) Deciphering the liquidity and credit crunch 2007-08. Journal of Economic Perspectives 23: 77–100.

10.

Bruno

Shin

(2015) Cross-border banking and global liquidity. Review of Economic Studies 82: 535–564.

11.

Cetorelli

Goldberg

(2012a) Banking globalization and monetary transmission. Journal of Finance 67: 1811–1843.

12.

Cetorelli

Goldberg

(2012b) Liquidity management of US Global Banks: Internal capital markets in the great recession. Journal of International Economics 88: 299–311.

13.

Chui

Domanski

Kugler

et al . (2010) The collapse of international bank finance during the crisis: Evidence from syndicated loan markets. BIS Quarterly Review 15: 39–49.

14.

Davis

(2011) The Australian financial system in the 2000s: Dodging the bullet. In: RBA conference 2010.

15.

De Haas

Van Horen

(2012) International shock transmission after the Lehman brothers collapse: Evidence from syndicated lending. American Economic Review 102: 231–237.

16.

Debelle

(2008) Market operations in the past year. Speech Dr Guy Debelle, Assistant Governor (Financial Markets), to 2008 FTA Congress, Melbourne, VIC, Australia, 31 October. Available at: http://www.rba.gov.au/publications/bulletin/2008/nov/pdf/bu-1108-4.pdf

17.

Dennis

Nandy

Sharpe

(2000) The determinants of contract terms in bank revolving credit agreements. The Journal of Financial and Quantitative Analysis 35: 87–110.

18.

Esho

Sharpe

Webster

(2007) Hedging and choice of currency denomination in international syndicated loan markets. Pacific-Basin Finance Journal 15: 195–212.

19.

Frey

Kerl

(2015) Multinational banks in the crisis: Foreign affiliate lending as a mirror of funding pressure and competition on the internal capital market. Journal of Banking & Finance 50: 52–68.

20.

Galema

Koetter

Liesegang

(2015) Lend global, fund local? Price and funding cost margins in multinational banking. Review of Finance 19: 1–34.

21.

Gatev

Strahan

(2006) Banks’ advantage in hedging liquidity risk: Theory and evidence from the commercial paper market. Journal of Finance 61: 867–892.

22.

Gatev

Strahan

(2009) Liquidity risk and syndicate structure. Journal of Financial Economics 93: 490–504.

23.

Giannetti

Laeven

(2012) The flight home effect: Evidence from the syndicated loan market during financial crises. Journal of Financial Economics 104: 23–43.

24.

Guttmann

Rodgers

(2015) International banking and liquidity risk transmission: Evidence from Australia. IMF Economic Review 63: 411–425.

25.

Heckman

Ichimura

Todd

(1997) Matching as an econometric evaluation estimator: Evidence from evaluating a job training programme. Review of Economic Studies 64: 605–654.

26.

Heckman

Ichimura

Todd

(1998) Matching as an econometric evaluation estimator. Review of Economic Studies 65: 261–294.

27.

Ivashina

Scharfstein

(2010) Bank lending during the financial crisis of 2008. Journal of Financial Economics 97: 319–338.

28.

Jeon

Olivero

(2013) Multinational banking and the international transmission of financial shocks: Evidence from foreign bank subsidiaries. Journal of Banking & Finance 37: 952–972.

29.

Kashyap

Rajan

Stein

(2002) Banks as liquidity providers: An explanation for the coexistence of lending and deposit-taking. Journal of Finance 57: 33–74.

30.

Loans (Australian Industry Development Corporation) Act (1974) Available at: http://www.comlaw.gov.au/Series/C2004A00212

31.

McCahery

Schwienbacher

(2010) Bank reputation in the private debt market. Journal of Corporate Finance 16: 498–515.

32.

Maskara

(2010) Economic value in tranching of syndicated loans. Journal of Banking & Finance 34: 946–955.

33.

Merrett

(2002) The internationalization of Australian banks. Journal of International Financial Markets, Institutions and Money 12: 377–397.

34.

RBA (2007) Financial stability review. Available at: http://www.rba.gov.au/publications/fsr/2007/mar/pdf/0307.pdf

35.

Rose

Spiegel

(2012) Dollar illiquidity and central bank swap arrangements during the global financial crisis. Journal of International Economics 88: 326–340.

36.

Ross

(2010) The ‘dominant bank effect’: How high lender reputation affects the information content and terms of bank loans. Review of Financial Studies 23: 2730–2756.

37.

Saunders

Steffen

(2011) The costs of being private: Evidence from the loan market. Review of Financial Studies 24: 4091–4122.

38.

Stevens

(2009) Australia and Canada – Comparing notes on recent experiences. In: Canadian Australian chamber of commerce Canada-Australia Breakfast, Sydney, NSW, Australia, 19 May.

39.

Strahan

(1999) Borrower risk and the price and nonprice terms of bank loans. Available at: http://ssrn.com/paper=19276

40.

(2014) Should Australian syndicated loan contracts be standardised? JASSA 1: 19–23.

Liquidity shock management: Lessons from Australian banks

Abstract

Keywords

1. Introduction

2. Background and data

2.1. The syndicated loan market in Australia

2.2. Data

2.3. Descriptive statistics

3. Approach and empirical analysis

3.1. Regression models

3.2. Univariate analysis

3.3. Liquidity risk and syndicate composition

3.4. Liquidity risk and syndicate composition pre- and post-GFC

4. Robustness tests 14

4.1. Liquidity risk management

4.2. Endogeneity tests

5. Conclusion

Footnotes

Appendix

Funding

Notes

References

4. Robustness tests¹⁴