Bank-Firm Relationship Analysis Based on Co-occurrence in News Texts:Evidence from Listed Real Estate Companies in China
Enwei ZHU,Jing WU,Hongyu LIU
2019-01-01
Abstract:The bank-firm relationship has long been of interest in the financial literature. Before a lending decision, the commercial bank typically requires the company to provide its "hard" information, such as financial statements and credit score data, to mitigate the problem of information asymmetry and reduce the potential default risk. However, this approach is less effective if the hard information is insufficient, especially for small companies. On the other hand, a good bank-firm relationship can help to transmit "soft "information, such as the character and reliability of the company's owner, which may also improve the company's credit financing capacity. However, quantitatively measuring the strength of the bank-firm relationship remains a challenge in the financial literature.In the context of the big data paradigm, we introduce the co-occurrence analysis approach and develop an innovative indicator for bank-firm relationships based on their co-occurrence in news texts. The co-occurrence analysis is a text-mining method that has been extensively applied for knowledge discovery in many domains. The key idea is to use the co-occurrence frequencies of given keywords in a corpus to represent their relationship strength. Regarding the bank – firm relationship, a good relationship can manifest not only in frequent business cooperation, but also in non-business interactions, such as mutually organizing or participating in public events, because both formal and informal communications can effectively deliver the company's soft information to the bank. Meanwhile, these mutual events are likely to appear in the news andthus be documented by online media if they are important enough. Hence, we propose to use the news co-occurrence frequency of a specific bank-company pairing to quantify their relationship strength. Relative to most traditional measurement approaches, the news co-occurrence indicator can comprehensively reflect both business and non-business bank-firm relationships, and has the advantage of data availability.Following this idea, we use the news co-occurrence indicator to quantitatively measure the relationships between 238 listed real estate companies and 19 major domestic banks from 2009 to 2014 in China, using Baidu News as the source of the news text. We test the validity of this indicator by comparing it with the corresponding bank's occurrence frequencies in the corresponding company's annual reports and its chief sources of debt financing.We then examine the effects of the bank – firm relationship on the company's credit financing capacity. To alleviate the potential endogeneity problem, we introduce two instrumental variables for the bank – firm relationship, namely the number of senior executives in the company with a banking background and the geographical distance between the headquarters of the company and the bank, and then estimate the models using the 2 SLS method. The empirical results suggest that a stronger bank-firm relationship can significantly improve the company's credit availability, but is not generally effective in reducing the cost of credit. In addition, such patterns are heterogeneous across companies; in particular, the bank – firm relationship effect is more pronounced for smaller or less profitable companies from the perspective of either credit availability or credit cost. Furthermore, a company can improve its credit availability more efficiently by choosing to further strengthen its strongest bank relationship rather than establishing relationships with new banks. Strengthening the relationship with SOE versus non-SOE banks achieves similar effects. Further empirical analysis shows that the news co-occurrence measuring approach can be extended to other industries, and corresponding empirical models suggest similar conclusions.This study is among the first to introduce big data and text-mining technology to research on the bank-firm relationship. It contributes to the literature in two ways. First, we propose an innovative and effective bank-firm relationship indicator based on their co-occurrence frequencies in news text, which can comprehensively reflect the relationship from both the business and non-business perspective. Hence, it can serve as an important complement to more traditional measures of the bank – firm relationship. Second, our empirical findings provide not only new evidence for the effects of the bank-firm relationship on a company's credit financing capacity, but also additional insights into the heterogeneities of such effects across companies and banks.