# Regulatory constraint and small business lending: do innovative peer-to-peer lenders have an advantage? – Financial Innovation

#### ByÇağlar Hamarat and Daniel Broby

Aug 15, 2022

We use a method that allows us to look at the impact of the regulation at a county level, following the approach taken by Tang (2019). We then apply a difference-in-differences (DiD) approach to obtain our empirical results.

We limited the research period so that the 2008 global financial crisis does not affect the data set exogenously. Our sample period starts after this date and due to using policy change in 2010 as an exogenous shock in our research method, we kept sample period limited to 4 years between 2009 and 2012 in order to mutually coincide the pre and post periods.Footnote 9 This sample was analysed in with a similar empirical method in Tang’s (2019) article where the period is 2009Q1–2012Q2. After the research period was limited to this period, we performed parallel trend analyses to test the robustness of the analyses results, and the results were confirmed.

In order to isolate the regulatory impact, we apply a negative shock at county level to supply of bank loans that leads banks to tighten their lending criteria. In this regard, we consider an arguably exogenous shock to bank small business credit supply that was due to the implementation of the Dodd-Frank Act in June 2010 which is described as the beginning point of the post-shock term. Using small business loan data at bank and county level in regard to the Dodd-Frank Act, we follow TangFootnote 10 (2019) and De Roure (2022) analyses who find that treated banks reduced lending.

In order to provide causal evidence, the Dodd-Frank Act is used as an exogenous shock. The DiD model compares the volume of small business lending one year before and two years after July 21, 2010 (the implementation date of the Dodd-Frank Act). The treatment group are banks that are affected by this regulation and control group are banks that are not affected.

We cannot completely exclude the possibility that time-varying, unobserved market variables, even with the “DiD” technique, simultaneously affect the development of Fintech loans and the position of traded banks before the shock. To alleviate this problem, we present in Fig. 2 findings that show a parallel trend of FinTech lending in both traded and non-traded markets before 2010Q2. We also show that the benefits of treatment began to take effect in the second quarter of 2010. Given the date of the Dodd-Frank, we also examine the impact of other additional regulations in the robustness section, it seems unlikely that other variables are responsible for this trend.

There are two cut-offs for financial institutions according to the Dodd-Frank regulations. The first one is for banks which are exceeding $10 billion in assets that subject to annual stress test and higher disclosure requirements. And the other is one for bank holding companies that are exceeding$50 billion in assets (called “systemically important banks”) that subject to semi-annual stress tests and a far-reaching list of disclosure requirements. However, due to having limited data about bank holding companies, we could not include systemically important banks in the DiD model, which are exceeding $50 billion in assets; therefore, we only use$10 billion as a cut-off and therefore could not apply alternative method Regression Discontinuity.

Firstly, by using equation one, we test and analyse the qualification of existing research related to the Dodd-Frank Act impact on bank level small business lending activity.

$$log (SBLoan)_{i,t} = beta_{i,t} left( {Treated_{i} *DFA_{t} } right) + lambda DFA_{t} + rho Treated_{i} + C_{i,t} + theta_{t} + Pi_{i} + epsilon_{i,t}$$

(1)

where (log(SBLoan)_{i,t}) is originated small business loans (origination volume $1 million or less) by bank (i) in year (t). (Treated_{i}) is a dummy variable that identifies the treatment group, one if the banks with assets over$10 billion threshold which are subject to the Dodd-Frank Act and zero for the banks with assets right below $10 billion threshold and exempted from Dodd-Frank Act. (DFA_{t}) is the treatment dummy that takes the value one from Dodd-Frank Act enactment date (21th July 2010), and zero prior for this date. (C_{i,t}) is a vector of bank-level control variables are defined in Table 2. (theta_{t}) is a variable for the county-year fixed effects and (Pi_{i}) is a variable for bank fixed effects, and both are used to help remove unobserved heterogeneity such as variation in local loan demand due to (county-specific) business conditions and for unobservable bank characteristics. ({epsilon }_{i,t}) is an error term. The four columns of Table 5 report the Dodd-Frank Act impact on bank small business loan volume. According to results, the coefficient of interaction term, Treatedi x DFAt, is negative and highly in all estimations with bank, county and year fixed effects. The results show that small business lending volume in treated banks decreases. In order to check traditional banks’ responses to Dodd-Frank Act in the counties for evaluating small business loan applications, we use the following equation: $$log (SBLoan)_{t,c} = beta_{t} left( {Treated_{c} *DFA_{t} } right) + lambda DFA_{t} + rho Treated_{c} + C_{t,c} + delta_{c} + gamma_{t} + epsilon_{t,c}$$ (2) where (SBLoan_{t,c}) is originated loans to small businesses(loans origination volume$1 million or less) in county (c) in year (t). (Treated_{c}) is a dummy variable that identifies the treated counties and takes the value of 1, if there is a bank with $10 billion assets or over affected by the Dodd-Frank Act and there is low competition according to the C3 and HHI, which are in the top 75th. If the county has a bank asset below$10 billion, and there is high competition in the bottom 25th, it is defined as a control county and takes 0. Counties other than the 75th and 25th percentile are not included in the model. (DFA_{t}) is the treatment dummy that takes the value one from Dodd-Frank Act enactment date (21th July 2010), and zero prior for this date. (C_{t,c}) is a vector of county-level control variables. (delta_{c}) variable for the county fixed effect, and (gamma_{t}) is a variable for time fixed effect. (epsilon_{t,c}) is an error term. The county level variables are defined in Table 3.

Table 6 reports the Dodd-Frank Act’s effect on county small business lending activity. The first column shows the result for the aggregated small business loan activities county and columns 6 and 9 show the small business loan for businesses with gross revenues less than $1 million and for businesses with gross revenues of at least$1 million, respectively.

According to results, the coefficient of the interaction term, (Treated_{c}) x (DFA_{t}) is both negative and high in all predictions with county and time fixed effects. The results show that small business lending in treated counties decrease relative to control group counties after the Dodd-Frank Act in terms of aggregate small business loan and for businesses with gross revenues less than $1 million, respectively. There is no significant impact on for businesses with gross revenues of at least$1 million.

In order to check if alternative lenders increased their lending in counties where small business lending decreased due to the credit supply shock’s effect on small business loan applications, we use the following equation:

$$log left( {SBLoan_{t,c}^{P2P} } right) = a_{t,c} left( {Treated_{c} *DFA_{t} } right) + lambda DFA_{t} + rho treated_{c} + C_{t,c} + delta_{c} + gamma_{t} + epsilon_{t,c}$$

(3)

where (SBLoan_{t,c}^{P2P}) is small business loan origination volume of alternative lenders loan in county (c) in year (t). (Treated_{c}) is a dummy variable that identifies the treated counties and takes the value of 1 if there is a bank with $10 billion assets or over and affected by Dodd-Frank Act and there is low competition according to the C3 and HHI, which are in the top 75th. If the county has a bank asset below$10 billion exempts from the Dodd-Frank Act and there is high competition in the bottom 25th, it is defined as a control county and takes the value of 0. Counties other than the 75th and 25th percentile are not included in the model. ({DFA}_{t}) is the treatment dummy that other takes the value one from Dodd-Frank Act enactment date (21th July 2010), and zero prior for this date. ({C}_{t,c}) is a vector of county-level control variables. ({delta }_{c}) variable for the county fixed effect, and ({gamma }_{t}) is a variable for time fixed effect. (epsilon_{c,t}) is an error term. All variables are defined in Table 4 with loan-level variables. We acknowledge that it is not clear whether the DiD coefficient of this regression reports the effect of Dodd-Frank exposure (the main point of our paper) or the effect of bank concentration (unrelated to the paper). That said, we emphasize that the high concentrated counties with low competition are exposed to more regulatory impact and that this in turn should result in an advantage to P2P lenders in the less concentrated counties.

The main dependant variable measures lending volume of the alternative P2P lender data that we used the dollar amount of alternative P2P lender origination volumes from the loan book that is specified at the county level. Due to having limited county-level data, instead of using normalizedFootnote 11 variables similar as in Tang (2019) paper, the logarithm value of the small business loan origination is used in the analysis.

The results of Eq. (3) are presented in Table 7. It is proved that in regard of control counties, loan origination volume of alternative P2P lender enhanced remarkably in treated counties after the Dodd-Frank Act became law in July 2010, in terms of the total loan amount. According to our results, there was a notable difference, between control and treated counties, in alternative P2P lender loan volume after the enactment of Dodd-Frank. The trend after the Dodd-Frank Act proves that the growth in demand for alternative credit between control and treated markets is unlikely to be urged by observable differences.

In accordance with Table 4, we find that treated counties experienced an increase in alternative P2P lender mall business loan applications compared to control counties.

This result is coherent with FinTechs’ and banks being substitutes or complements with the findings of Tang (2019). However, this analysis is necessary for validating the Dodd-Frank Act as a negative shock to incumbents’ small business loan supply. We emphasise the limitation to our approach is the restricted data available on alternative lenders. To sum up, the results on the volume of alternative P2P lender loans reveal that, when incumbents cut lending in the small business credit market, some borrowers tend to move from incumbents to alternative P2P lenders.

To check the parallel-trends assumption, we present Fig. 2, which shows lending by banks overtime for the treated and control group.

The Fig. 2 shows that in treated states, new small business loan volume is similar to that in control states before the Dodd-Frank Act. This indicates that the parallel-trends assumption is valid. After the Dodd-Frank Act, the new small business loan volume decreased both for treated and control banks, but it decreased more and faster in treated counties than in control counties which are presented in Fig. 3.

Similarly, we check the parallel-trends assumption with an alternative P2P lender. Figure 4 shows an alternative P2P lender credit provision in treated and control counties. It shows that the volumes of new alternative P2P lender loans to small businesses in control and treated counties displayed parallel trends prior to the Dodd-Frank Act. After the Dodd-Frank Act, P2P small business lending increased in treated counties.

## Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.