A Loan-level Investigation of Chinese Credit Guarantee

Abstract

Using a loan-level dataset on a big private-owned credit guarantor in Guangdong province, we study how the guarantor manages risk. In a typical bank-guarantor collaboration agreement, the bank may cherry-pick a borrower to co-sponsor the loan with the guarantor so as to share the default risk, the guarantee fee and the collateral. Borrowers with no bank co-sponsorship should be riskier than borrowers with bank co-sponsorship, so the guarantor may treat these two types of borrowers differently. We find that the borrowers with guarantor’s sole sponsorship pay higher guarantee fee, pledge more collateral, and are granted shorter loan maturity than the borrowers with bank co-sponsorship, and that the loan amount is not affected by the absence of bank’s co-sponsorship.

Keywords:

• Credit guarantor
• risk management
• SME lending

Acknowledgments

We thank the Finance Bureau of Guangdong provincial government for providing us the proprietary dataset. We also thank Viral Acharya, Warren Bailey, Hung-Gay Fung (the editor), Dragon Tang, Wei Xiong, and the discussants in the 7th International IFABS Conference and 2015 China International Conference in Finance for useful comments. All errors are ours.

Declarations of interest

None.

Notes

1 The Chinese banking system imposes significant collateral requirement for all banking deals. For example, when commenting on the reasons why private firms are not served by the Chinese banks, Hu Xiaolian, a vice governor of Chinese central bank, said “The lack of legitimate collateral is the real bottleneck. Approximately 70% of the credit collateral received by commercial banks in China is real estate, while more than 70% of the assets of SMEs are account receivables and inventory.”

2 The aggregate industry statistics reported in this paper are collected from IPO prospectuses of several Hong Kong listed guarantors. They are respectively Hanhua Financial (3903.hk), Guangdong Join-share Financing Guarantee Investment Company (1543.hk), and China Success Finance Holdings (3623.hk).

3 Hancock and Wilcox (1998) study the U.S. Small Business Administration loan guarantee program and find it helps to extend credit during credit market stress. Ono et al. (2013) study the Japanese Emergence Credit Guarantee Program following the 2008 financial tsunami. They find in general the government support program boosts credit availability, yet bank relationships could have perverse effects on the efficacy of public credit guarantees. Gropp et al. (2014) study the elimination of German government guarantee in 2001 as well as its effects on bank lending. They suggest German government guarantee has a moral hazard effect and encourages bankers to loosen their lending standards.

4 Their sample consists of the deals done during 2006-2009 by one of the top three credit guarantors in China. Our sample consists of loans made in Guangdong during 2009‒2013, involving one of the biggest guarantors in the province.

5 Credit guarantors are different from micro- or small-loan companies, who lend out their own funds. Usually, a SME may borrow a small amount from a small-loan company, and may borrow a bigger amount from a bank if a credit guarantor backs up the bank loan. Credit guarantors thus play a more important economic role than small-loan companies.

6 For example, a bank-guarantor collaboration agreement usually specifies a total bank-loan limit that the guarantor can guarantee, and may also define the ceiling of credit exposure of the guarantor to an individual borrower and/or an individual deal.

7 We do not have the data of the interest rate that banks charge. It is likely that they do not charge much higher than the benchmark rate, since they bear little credit risk. Indeed, this is confirmed by an agreement between the big guarantor and a major bank that we have seen.

8 The 18 banks cover the entire spectrum of banking ownership structure in China, including policy banks, state-owned banks, shareholding banks, and city commercial banks.

9 We note that for each loan in our sample, there is a clause of the recourse against the personal property of the controlling shareholder of the borrowing firm as well as any connected enterprises that she may control.

10 Nonetheless, we check whether our results are affected by inclusion of these 15 cases. In unreported analysis, we find that all the later tables, except Table 6 below (testing Hypothesis 2, the collateral effect), are not much affected. We redo that table when these 15 cases are also included, using Tobit analysis. The results show an insignificant difference in collaterals between loans with and without bank co-sponsorship. Note however that the inclusion of these 15 cases can downwardly bias the significance of the collateral effect of zero bank co-sponsorship, since these borrowers, offering no collateral when the guarantee application process begins, tend to have zero bank co-sponsorship. Furthermore, we note that there are 14 extreme-maturity cases (Maturity <12 months or = >60 months) in Table 2 below. We re-run the baseline analysis of Table 7 below after deleting these 14 cases, and find that the coefficient estimate of Sole is robust.

11 It may be argued that 10% bank sponsorship is slightly bigger than 0%, so it should not matter. Yet, later results show strong effects of Sole, suggesting that 10% bank sponsorship already makes a difference.

12 An increase of 0.5 percent in the RRR would reduce by over RMB 300 billion the available liquidity in the banking system, affecting especially those small- and medium-sized banks that have a constant demand for liquidity in the market (Wang & Sun, 2013; Shen et al., 2014).

13 In the equations below, Credit Spread is abbreviated as Spread for brevity.

14 In a setting with multiple endogenous variables, statistically significant instruments in explaining a single endogenous variable can still become weak instruments. That is, if a similar set of IV are used in the first-stage regression (equation system 3) to generate the predicted values of the many endogenous variables, then in the second-stage regression (equation system 4, see below), there will be multicollinearity among the R.H.S. predicted values of the endogenous variables. And given the multicollinearity, some of the endogenous variables may not be properly identified, and their corresponding instruments in such a case become weak.

15 From the initial list of IV candidates, we delete those candidates if their individual t-statistics are not significant at 5%-level in the residual regression suggested by Sanderson and Windmeijer (2016, p. 215). For the remaining candidates, we run the F-tests suggested by Angrist and Pischke (2009, p. 218) and Sanderson and Windmeijer (2016, p. 215) to make sure that they are not weak IV. Weak IV will lead to insignificant F-statistics.

16 Hu Fulin, a prominent entrepreneur, fled the country amid debt crisis in September 2011. Many consider this event to mark the beginning of the Wenzhou credit crisis.

17 Infrastructure- and construction-related firms have an average receivable collection period of 248 days (versus the average of 53 days for the rest of the sample). SOEs have an average receivable collection period of 201 days.

18 The macro-prudential rule issued on August 16, 2006 by CBRC put a cap on a bank's total lending exposure to the real estate sector. This sector is considered by the regulators as riskier (systemically) than other sectors; so although real estate developers are among the most profitable enterprises during the last decade, they find it hard to access bank financing and may be among the borrowers without bank co-sponsorship in our sample. In unreported analysis, we find a high correlation (0.57) between firm size and a dummy variable of infrastructure and real estate business. Out of the largest 30 firms, 27 of them have not received formal bank sponsorship, and 21 of them are in infrastructure and real estate business.