ABSTRACT
This paper empirically tests the effect of FinTech development on bank risk taking using unbalanced bank-level panel data from China for the period from 2011 to 2018. We use the media's attention paid to FinTech-related information to gauge FinTech development. We find robust evidence that the development of FinTech exacerbates banks’ risk taking in general. The heterogeneity analysis further indicates that the asset quality deterioration effect brought about by prosperous FinTech is more salient in banks with larger sizes, lower efficiency, more shadow banking business and more interest-based income. Moreover, the nexus between FinTech and banks’ risk taking is a U-shaped trend, with FinTech initially intensifying and then weakening banks’ risk taking. Moreover, the banks’ responses regarding the U-shaped effect are heterogeneous among different ownership structures. The responses by state- and jointly owned banks are not notable, while those of city banks, foreign banks and rural banks are more sensitive.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
12 Our study begins in 2011, mainly because a large scale of FinTech has emerged since then in China
13 A three-consecutive-year rolling window is applied to determine the σ(ROA)it.
14 Due to the highly skewed character of a Z-score, we use the natural logarithm (1+ Z-score) to smooth higher values (Beck, Jognhe, and Schepens Citation2013). Using (1+ Z-score) instead of simply Z-score avoids the truncation of the Z-score at zero. We denote ln(1+ Z-score) as the Z-score in the latter part of the paper for brevity.
15 When we count news headlines at the technical level, prefixes such as "financial” or “banking" are added before each technology, for example, "financial, big data", "banking, big data", "financial, cloud computing", "banking, cloud computing", etc. Therefore, the word frequency calculated in the dimension of "technical foundation" reflects various technologies applied in the financial field, rather than over a wider scope.
16 In the regression, we aggregate the quarterly indices of each FinTech sub-index to obtain the annual index.
17 For example, when the Z-score is used as the dependent variable, Hausman’s test χ2 statistic is 44.80, and the p-value is 0.0024 when all of the control variables are included. The χ2 measure is significantly different from 0 at 1%.
19 We thank the anonymous referee for this suggestion.