The impact of FinTech start-ups on incumbent retail banks’ share prices

From Uber to Airbnb, disruptions have transformed many industries. Until recently, the financial sector remained largely untouched. This changed with the breakthrough of FinTech firms, which are “companies that use technology for banking, payments, financial data analytics, capital markets and personal financial management” (Huang, 2015). In 2015, global FinTech investment grew by 75%, exceeding the high 22 billion USD amount, and continues to rise (Dickerson et al., 2015). This growing number threatens incumbent retail banks as a new wave of digital banking start-ups has emerged.

This research specifically focuses on digital banking start-ups since they offer the same type of services as retail banks: accepting consumer deposits, facilitating payments, and lending money (Chishti and Barberis, 2016). Their funding serves as an indicator of their potential values. This study uses stock returns to estimate the impact of this funding on the performance of traditional banks. If the returns react negatively, these entrants might disrupt the industry, while a positive effect might imply complementarity. To specify the research, we select the United States because it has the largest FinTech industry, as well as the highest number of FinTech adopters (Ernst and Young, 2014).

Because FinTech is a recent development, very little scientific research has been conducted yet. However, existing research mainly stresses start-ups’ ability to disrupt industries (Chishti and Barberis, 2016; Dickerson et al., 2015). Kauffman et al. (2015) support this view by stating that current financial firms mainly depend on underlying technology innovations and not on their historical market position. This consequently forces traditional banks to enhance their service quality and reduce transaction costs, acting according to strategic necessity rather than competitive advantage (Goh and Kauffman, 2013). Jun and Yeo (2016), however, deemphasize competition and potential substitution by stressing the complementary effect of FinTech. The contradicting views on the future of FinTech and the largely descriptive research motivate us to reassess the role of digital banking start-ups in the traditional banking industry using a quantitative approach.

This paper proceeds as follows. The second section reviews the conceptual and empirical literature on disruptive innovation and FinTech-related studies. The third section presents the data used for the analyses, followed by a methodology description in the fourth section. The fifth section discusses the results two subsections that address the volume of funding and the number of deals. Finally, the final section presents the conclusion from this research as well as its limitations and future research recommendations.

As technology evolves, products, critical success factors, and industry characteristics change (Afuah and Utterback, 1997). This phenomenon has greatly affected the taxi and hotel industry, in which Uber and Airbnb have largely replaced the traditional services by offering decentralized online peer-to-peer platforms (Cannon and Summers, 2014). Since FinTech start-ups follow the same customer-centric approach by offering alternatives such as peer-to-peer lending (Lin et al., 2015; Wang et al., 2015a, b; Wang et al., 2015a, b; Yan et al., 2015), one might expect a shift in the banking industry as well. Researchers apply consumer theory to examine whether digital banking start-ups indeed negatively affect traditional retail banks. This theory states that a new service will act as a complement when utilized jointly with an old service and will serve as a substitute if it can replace the old service by satisfying the same needs (Aaker and Keller, 1990; Frank, 2009). Thus, the services digital banking start-ups offer would benefit the traditional retail banks in the former case and affect the incumbents’ performance in the latter case (Kaul, 2012). It might also be possible to observe no effect when examining stock returns, which could indicate that the complementary and substitution effects offset each other. Other explanations are that the start-ups are simply too small, our measure does not capture FinTech’s value well, or FinTech serves a new channel.

This study used data from the venture finance-data firm CB Insights to identify US digital banking start-ups closely related to retail banks. CB Insights gathers data from different sources, such as regulatory filings (such as Form D), investor websites, company websites, and (social) media. In addition, some investors post data directly to CB Insights.³ Our analysis depends on public disclosure of FinTech deals to be captured in the CB Insights database. Note that it would also be difficult for investors in incumbent banks to change their view on the share price if FinTech deals remain private.

To retrieve the sample, we first applied several filters to companies that engage in FinTech activities and are closely related to retail banks. The main industries we selected included “Internet,” “Mobile and Telecommunications,” and “Financial,” and refined them by selecting industries and subindustries. For “Internet,” we filtered firms for “Internet Software and Services” with “Accounting and Finance” as the subindustry. For “Mobile and Telecommunications,” we selected the “Mobile Software and Services” industry, with “Accounting and Finance” as the subindustry. Lastly, for “Financial,” we applied “Asset/Financial Management,” “Lending,” “Payments,” and “Retail Banking” as filters. This gave us a sample of 558 firms, which we narrowed down to 522 companies by excluding InsurTech. The selected FinTech start-ups are located in the US and received investments from January 2010 to March 2016. We chose this period because most FinTech firms began operating from 2010. CB Insights also provides the volume of funding⁴ committed to these start-ups and the number of deals since January 2010. Unfortunately, the start of FinTech coincides with the establishment of the Dodd-Frank Wall Street Reform and Consumer Protection Act that was signed into federal law on 21 July 2010. The major restructuring of financial regulation may have affected the share prices of traditional banks immediately at its introduction, in which case the effect on our results would be limited. However, a prolonged period in which investors needed to incorporate the new federal law fully into banks’ share prices may influence our empirical results. However, it is not immediately clear how this would affect the month-to-month correlations of FinTech deals and bank prices.

The question may arise as to whether FinTech companies fall under similar regulations as traditional banks. On this point, regulators are somewhat ambivalent. On the one hand, they want to encourage innovation in the banking sector by lowering regulatory costs for start-ups. On the other hand, they are aware that for similar services, such as lending, they should have similar regulations in place to protect consumers and to create a level playing field with traditional banks.⁵ However, financial regulators are still deciding whether to introduce new and stricter regulation for FinTech companies, potentially weakening their threat to incumbent banks.⁶ Buchack et al. (2017) suggest that FinTech companies may be involved in regulatory arbitrage. Regulators may want to prevent this from keeping a grip on the financial system.

As Fig. 1 shows, funding volume is volatile month to month. We used the standardized values of the group mean and standard deviation. We also transformed the absolute amounts of funding into growth rates. The natural logarithm difference between month t and month t − 1 is a proxy for the growth rate of month t. We applied these transformations to the number of deals. Table 1 presents all transformations, and we show our raw data in Appendix Table 4 for reference. We also show the results for a subsample of only “Consumer FinTech” in Panel B, but we see that this subsample is rather small and is therefore less likely to fully capture the impact of FinTech on the retail banking industry.

Original variable	Transformed variable	Formula
FinTech funding volume	Standardized FinTech funding volume	\( {z}_t\kern0.5em =\kern0.5em \frac{x_t\kern0.5em -\kern0.5em \overline{x}}{s},\kern0.5em t\kern0.5em =\kern0.5em 1,2,\dots, n \)
	FinTech volume growth rate	g t   =   ln x t   −   ln x t − 1,  t  =  2, 3, …, n
FinTech number of deals	Standardized FinTech number of deals	\( {z}_t\kern0.5em =\kern0.5em \frac{x_t\kern0.5em -\kern0.5em \overline{x}}{s},\kern0.5em t\kern0.5em =\kern0.5em 1,2,\dots, n \)
	FinTech number of deals growth rate	g t   =   ln x t   −   ln x t − 1,  t  =  2, 3, …, n

We retrieved a list of 138 US retail banks from Credio.com, a finance website that compares US financial services.⁷ We used the Bloomberg database to obtain banks’ monthly total return indices. The total return index (TRI) is the dividend-adjusted stock price that measures both the capital gains and dividend distributions, assuming that dividends are reinvested in the same stock. We present the development of the total return indices of the financial sector and US market in Fig. 2.⁸

Of the 138 retail banks, we have total returns available for 47 banks. The excluded banks are often regional banks. We do not include non-listed banks because they would have only annual accounting returns such as Return-on-Equity (ROE) or Return-on-Assets (ROA) available. With annual data, it would be challenging to separate market-wide effects from the effects of FinTech funding. Hence, we include listed firms with monthly data to which we can link monthly FinTech funding data. Using performance information from non-listed banks could be a valuable extension for future research.

Since we do not have data available for the regional operations of traditional banks or FinTech companies, we cannot provide a more in-depth analysis of which FinTech deal flow affects which bank. This may be a fruitful area of further research. Note that the information we do have about the location of the headquarters may not be that helpful in this regard, as traditional banks have local offices to service clients. Moreover, the FinTech industry seems to be heavily concentrated in either the technology center in California, particularly Silicon Valley, or the financial center around New York, with the US industry concentrated in these locations with about 74,000 and 57,000 people employed in each location respectively (Ernst and Young, 2015). Accenture confirms that New York and Silicon Valley attract the most capital for the FinTech industry (Dickerson et al., 2015). Since FinTechs operate mainly digitally, office locations seem to be even less meaningful when trying to examine geographical relationships. However, since this study concentrates on the industry level, the sample is still representative and the aggregation of the 47 selected banks will nevertheless serve as a reasonable sample for the retail banking industry. Using banks’ TRI, we calculate the stock return for month t as follows:

We assume that the normal return for each incumbent retail bank is well-proxied by either the three-factor model (Fama and French, 1993) or the five-factor model (Fama and French, 2015). We then add the FinTech variable to see whether it has an effect on the stock price above and beyond these three- or five-factor models. The resulting regression equation is:

where RMRF is the excess return of the equity market relative to the risk-free rate, SMB is the return of small minus large capitalization stocks, HML is the return of high minus low book-to-market stocks, RMW is the return of robust minus weak profitability stocks, and CMA is the return of conservative minus aggressive investing stocks. The three-factor model does not include the RMW and CMA factors. Depending on our model, FinTecht represents the standardized FinTech funding volume and the funding volume growth rate, or the number of FinTech deals and the growth in the number of deals.

Since we have a panel data structure, we must determine how to model each parameter. We consider three model specifications: (a) pooled, (b) fixed effects, and (c) free. In the pooled model, we assume that each estimated coefficient is the same for each bank. In the fixed effects model, we assume bank-specific alphas, but keep the other parameters the same for each bank. In the free model, each parameter is bank-specific, except for the FinTech coefficient. We test the 훾-coefficient for statistical significance for each model specification. In addition, we also include the results of a single regression equation on the banking industry returns, in which we do not use information at the individual bank level.

In the Appendix, we also report the results of a Seemingly Unrelated Regression specification, in which we also allow a bank-specific parameter γ. By studying the γ-coefficients on an individual level, we can analyze which banks seemed to be more affected by FinTech.

First, we see that the parameter estimates for the three panel data models and the industry-level regression leads to similar parameter estimates. The three panel data models have higher associated t-values than at the industry-level. This is due to the cross-sectional information used in the panel data setup, while at the industry-level we have only time-series variation.

Based on the panel data models, we see that the funding volume itself is not statistically significant, with t-statistics ranging from 0.03 to 0.98, depending on the panel data model and whether we use the three- or five-factor model as a benchmark. The number of deals is marginally significant, with t-statistics varying between 1.86 and 2.00 for the different panel data models. The results for the growth rates are much stronger; the t-statistics for the panel data models are well above 2.5 for each, indicating statistical significance. Since the coefficients are all positive, this implies that stock returns of incumbent retail banks are higher in periods with increasing FinTech funding or number of deals. Hence, we do not find evidence of creative destruction, but rather that the FinTech industry has a positive spill-over to the traditional retail banking industry.

We show the results of the Seemingly Unrelated Regressions for the growth in FinTech funding in the Appendix in Table 5. These bank-specific estimates of the gamma parameter show that the positive coefficient above is an average across banks and that 16 of 47 banks have negative estimates, although none are statistically significant. Our results are not statistically significant for the level of funding, have different signs across banks, and are not statistically significant at the industry-level, suggesting that our finding of complementarity is fragile.

This paper aims to provide a new perspective on the debate about the impact of US digital banking start-ups on the performance of US retail banks by using quantitative analysis. Based on our empirical results, we reject both null hypotheses incorporating the effect of the growth of the volume of funding and the change in the number of deals, as all coefficients from our panel data models are positive and statistically significant. Therefore, we conclude that although the large sum of investments raised public awareness, the funding of digital banking start-ups is more likely to have a positive effect on the incumbents’ stock returns than a negative effect. This suggests complementarity between FinTech and traditional banking, rather than substitution and disruptive innovation. However, our results are weaker for the level of funding, or if we investigate the banking industry returns.

Although we find positive and statistically significant results, we cannot rule out that our findings are spurious and alternative scenarios are also at play. First, while the FinTech start-ups are growing rapidly, they may still be too small to affect incumbent US retail banks. Within a period of less than five years, it is difficult for consumers to adapt to the new changes newcomers brought or to trust in their online and automatic services. Additionally, retail banks benefit from their ingrained advantages, such as their ability to create credit instantly. Second, the substitute and complementary effects may partly offset each other. On the one hand, successful FinTech firms may have weakened the banks’ dominant position by improving the quality and efficiency of traditional services; on the other hand, banks have taken actions to respond to these challenges, possibly by acquiring FinTech start-ups or setting up their own FinTech affiliates. Third, FinTechs could have established a new channel to distribute financial services, as start-ups and incumbents might have different customer bases. If this is the case, direct competition may remain low, which makes it difficult to observe a clear relationship.

This study has several limitations. First, we used FinTech funding as a proxy for the potential value of the FinTech start-ups because no other data are available. Although we believe that we made reasonable assumptions, we cannot avoid the potentially negative impact of this proxy on the empirical analyses. It would also be interesting to investigate the effect on bank risk taking or competition within the industry, but we lack the necessary data for such a study. Second, the sample might be too small to draw conclusions for the time-series regressions due to the sample period of January 2010 until March 2016, which is rather short. However, these are the only data available. We also could not examine the geographical areas in which FinTech firms and traditional retail banks operate in more detail. Having data available in this area could lead to stronger claims on the causality between FinTech flows and retail bank performance. Similarly, we searched for information to differentiate the reaction of incumbent banks to FinTech developments. Whether banks ignore FinTech, fund their own start-ups, or cooperate with external FinTech firms should affect their stock prices. However, we could not find enough information on the incumbent banks to analyze, such a differentiated view. Another limitation might be that we use monthly data. Since technological innovations follow investments in FinTech firms, there may be a delayed reaction at the moment of the discovery of the innovation rather than at its initial financing. Here, we had to compromise between the frequency of our analysis and the length or our sample period. Using a quarterly frequency would decrease the number of observations and thus limit statistical inference, by definition. Third, as we discussed in the theoretical framework section, the Fama-French three-factor model suffers from shortcomings; for example, it ignores the momentum effect. Although the Fama-French model was improved by introducing two extra factors to create the five-factor model, the momentum effect and the low volatility factor are still of concern (Blitz et al., 2016).

Future research could undertake an extended cross-sectional analysis of the effect of start-ups by studying different countries, because other countries may have a different structure in the retail banking industry and impact outside the US. It would also be worthwhile to expand our sample of listed banks to include non-listed banks, although these studies would have to measure performance through low frequency accounting measures instead. Moreover, examining the effect of different investment stages on the incumbents’ stock returns might provide more insight in the fast-growing FinTech industry.