Peer-to-Peer Lending as a Determinant of Federal Housing Administration-Insured Mortgages to Meet Sustainable Development Goals

Abstract

In this paper, we investigate the influential factors of Federal Housing Administration (FHA) mortgage loans, focusing our research interest on peer-to-peer (P2P) lending, the most successful FinTech lending model. We consider P2P lending an alternative source of financing that marginal borrowers use to pay the increased mortgage down payment, making them eligible to receive a mortgage from conventional banks. In other words, we examine whether and to what extent P2P lending has a positive impact on the FHA loans volume by providing the ability to circumvent the loan-to-value (LTV) cap policy. As a result, P2P lending can be seen as a means for ”rationed” borrowers to have access to the market by reducing inequalities and promoting financial inclusion, thus achieving Sustainable Development Goals (SDGs). We employ hand-collected data from FHA mortgages, P2P loans, and other economic factors from all 50 U.S. states during 2007–2017 and use panel data techniques for this purpose. Research shows that P2P lending, GDP per capita, population growth, broad money growth rate, interest rate, unemployment rate, new housing units, and consumer confidence Index produce effects on FHA loans. We show that P2P lending, a nonconventional determinant, is causally associated with a significant increase in the count and volume of FHA loans, implying that P2P lending has a positive impact on them. The ability of P2P to bypass mortgage supply constraints (tightened LTV caps) by providing small loans to borrowers to meet the increased down payment requirements is very important to policy-makers, as it shows that constraining the volume of mortgage loans may be not achieved. Macroprudential tools designed to control credit growth may prove ineffective, as the use of alternative forms of lending helps circumvent them and ultimately leads to excessive household leverage with all the risks that it poses to the financial system.

1. Introduction

After the housing crisis of 2007–2009 and the subsequent recession, the availability of mortgages decreased and lending standards tightened. This credit crunch drove borrowers to search for alternative sources of finance, such as peer-to-peer (P2P) lending, which is considered the most successful FinTech lending model.

The aim of our paper is to investigate whether P2P lending acted as a credit channel, bypassing the restrictions imposed by banks on mortgage lending, and can be considered a means to reduce inequalities and to facilitate the economic inclusion of borrowers in the mortgage market. In our paper, we show that the increase in volume of FHA mortgages, as it is fueled by P2P lending, among other factors, can be viewed as a path to financial inclusion for underserved borrowers, thus achieving the objective of equal access to financial resources, which is the main goal of sustainability. In that sense, P2P lending creates social impact by providing access to credit for individuals that are excluded or underserved by traditional financial institutions. In other words, this form of social lending aims to make mortgage financing and investing more accessible to all borrowers.

We focus on the Federal Housing Administration (FHA) mortgages because the FHA is considered the closest substitute for private subprime lending, thus representing the riskiest category of mortgages. Post crisis, subprime lending was almost eliminated, and riskier potential borrowers struggled to obtain a mortgage loan. FHA facilitated loan originations to low and moderate income borrowers, gaining market share, which peaked at 37% of overall home purchase mortgage market in 2009, while the corresponding percentage in 2007 was 7% [1].

While regulators did not explicitly impose mandatory loan-to-value (LTV) ratios on residential mortgages, the FHA increased its caps. In 2008, the Housing and Economic Recovery Act (HERA) increased the minimum cash down payment requirement on FHA loans to 3.5%. The down payment for FHA loans was as low as 3% for forty-three years (1965–2008). Almost half of FHA mortgage originations in the period from 2002–2006 had an LTV ratio above 97%. In the following years, the origination of risky loans gradually declined, reaching an all-time low of 1.23% in 2017. Since fiscal year 2010, over 50% of issued loans were within the LTV ratio of 95–97% [2].

Historically, the statutory cap on LTV was a commonly used borrower-based macroprudential policy to limit the risks, to curb the real estate boom, and to ensure safety in the economy. The effectiveness of the LTV policy was examined by studies such as [3,4,5]. More precisely, by performing a counterfactual policy analysis, Ref. [6] found that the dangerous expansion in the U.S. housing sector could have been effectively offset if authorities had followed a maximum LTV ratio policy.

One issue with implementing LTV limits is that there are ways to bypass the restrictions. In the U.S. during the housing boom, piggyback loans grew rapidly. They were used to help pay down payments on a property or to avoid paying private mortgage insurance (PMI). With piggyback mortgages, borrowers could obtain a secondary loan, usually from a different lender, to fund a fraction, or even the whole amount, of the down payment required for a home purchase. As a result, in many cases, the down payment was as low as zero [7]. In our paper, we evaluate the ability of P2P lending, a lightly regulated channel of FinTech credit, to provide funds to potential borrowers to meet the increased down payment for a house purchase. P2P lenders are generally regulated by the U.S. Consumer Financial Protection Bureau, the Federal Trade Commission, and state regulators rather than by the Federal Reserve. We investigate the increase in FHA loans within the framework of tight mortgage lending criteria attributed to, inter alia, the expansion of FinTech lending as a complementary channel to mortgage lending for marginal borrowers. Given that FHA loans mostly address subprime borrowers, who are borrowers with low credit scores and less available cash, it is important to examine how FHA loan originations were evaluated during a period of credit tightness and increased down payment requirements.

The rationale behind the regulatory change in 2008 was the belief that low down payment combined with falling house prices would lead to high rates of default. The maximum LTV ratio was reduced to 96.5% in order to limit the access of risky borrowers to mortgage credit and to prevent negative default externalities.

However, our results show that FHA-insured mortgages increased, since marketplace lending allowed subprime borrowers to ensure the necessary financing and bypass the imposed LTV-based policy.

Marketplace lending emerged after the global financial crisis and is now the leading alternative financing format globally. In the U.S., it is estimated that marketplace lenders currently supply 38% of all unsecured personal lending. According to Ref. [8], the P2P loan origination volume in the U.S. increased by 84%, on average, since 2007. In 2015, P2P consumer lending was equivalent to 12.5% of traditional consumer lending [9]. In 2017, consumer marketplace lending reached 46 billion USD in total, growing by 12.3 billion USD (36.5%) in 2017 alone.

Many researchers, such as [10,11,12,13], examined the risk related to marketplace lending. Ref. [14] mention six core problems associated with P2P lending, namely information asymmetry, determination of borrower credit scores, moral hazard, investment decisions, platform feasibility, and immature regulations.

Our results show that FHA mortgage originations rapidly increased in the post-crisis period and this event is closely related to the emergence of P2P lending in the U.S. Moreover, the recovery of FHA loans observed from 2015 to 2017, after the big loss of its market share from 2011 through 2014, coincides with the sharp increase in P2P loans. The total volume of FHA mortgages issued in 2015 was 72% higher than the previous year, and the corresponding increase in P2P lending was 83%.

Our analysis highlights the dynamic of crowdlending to circumvent tightened LTV caps (mortgage supply constraints) by providing small loans to borrowers, who can in turn meet the increased down payment requirements, thus increasing the total volume of FHA loans. To our knowledge, this paper is the first that examines to what extent the LTV cap policy can be bypassed by the availability of P2P lending. Beyond financial advantages such as faster approval, flexibility, and higher returns for investors, this form of lending can also encourage financial inclusion and social impact by providing access to credit for underserved segments of the population. In this sense, it supports social and environmental causes, thus promoting sustainability. Financial inclusion is essential for reducing poverty, promoting economic growth, and increasing the economic social and environmental impact of investments on people’s lives [15].

Several recent studies have examined the U.S. residential mortgage market in light of new technological achievements and show that traditional methods of lending may no longer be common practice. For example, Ref. [16] studies how the use of digital footprint is reshaping the mortgage market and shows that the number of lenders that use a borrower’s digital footprint has remarkably increased, and that these users bear a significantly lower risk compared to nonusers. Refs. [17,18] show that FinTech lenders dramatically increased their market share of U.S. mortgage lending, suggesting that technological innovation has improved the efficiency of financial intermediation in the U.S. mortgage market.

A paper closely related to ours is [19], which is the first to examine the growth of Chinese P2P lending after the increase in LTV limits. Authors conclude that policy intervention in the real estate mortgage market caused an increase in the demand for peer-to-peer lending in China, as it acts, to some degree, as a credit channel that circumvents city-level LTV caps and the housing market macroprudential policy in China.

In addition to using U.S. data instead of Chinese data in our analysis, there are two key differences between this study and our paper. First, in addition to crowdfunding, we study the effect of other factors on the development of FHA mortgages, such as the GDP per capita and population growth, which, as empirical results show, have a significant impact on mortgage origination. Second, while they use a difference-in-differences setting to study changes in P2P lending after the change in the LTV policy, we use a panel data model to analyze the correlation between FHA mortgages and crowdlending from 2007 to 2017. The panel data technique is common in applied research and it uses all the data information available. The use of panel data may offer several advantages, such as control for heterogeneity across individual units (states in this study), support of a great number of variables, and less multicollinearity among independent variables.

Clearly, the volume of FHA-endorsed mortgages is affected by many factors. The most important ones are related to the prevailing economic conditions over the eleven-year period under review. Some of the most significant economic and financial factors that are incorporated in our model to examine their effect on the U.S. mortgage originations are interest rate, GDP per capita, unemployment rate, house price index, new residential housing units, population growth, broad money growth rate, and consumer confidence index.

The main goal of this study is to examine the correlation between FHA-endorsed mortgages and marketplace lending, taking into account other macroeconomic and financial factors as well.

Technically, we use panel data analysis for the period 2007–2017. We employ a state-level data set for all 50 U.S. states, and we test the correlation with FHA loans.

This study suggests that the evolution of FHA loans in the sample period 2007–2017 was influenced by many factors, among which was the boom of P2P lending, our main variable of interest. The study finds that the emergence of crowdlending is causally associated with a significant rise in the volume of FHA loans during the study period, implying that P2P lending has a positive impact on mortgage loans.

Furthermore, there is evidence for seven more determinants of FHA mortgage endorsements: GDP per capita, population growth, broad money growth rate, interest rate, unemployment rate, new housing units, and consumer confidence Index.

The rest of the paper is organized as follows: Section 2 summarizes the growth in post-crisis FHA-insured mortgages and marketplace lending. Section 3 reviews the literature. Section 4 presents Holmstrom and Tirole’s 1997 model, which allows us to formulate the hypothesis that P2P lending has a positive effect on FHA mortgage origination. Section 5 describes the data used and the methodology. Section 6 presents and discusses the empirical results. Finally, Section 7 concludes the paper.

2. Loan Origination Behavior in the Post-Crisis Period

We conducted a data analysis of FHA mortgages and P2P loans issued from 2007 to 2017, as after the housing market downturn mortgage endorsements were severely curtailed while FHA mortgage volume significantly increased. Figure 1 presents the percentage change in total mortgage originations versus the percentage change in total FHA loan originations during the period under review. The 2007 financial year, which was the beginning of the financial crisis, is used as the base year for comparison purposes.

During the entire period under review and especially the first four years (2008–2011), aggregate mortgage endorsements were severely curtailed while FHA mortgages skyrocketed, despite the reduction in the LTV ratio. A characteristic example is the year 2008, during which the total mortgage originations decreased by 34% compared to the base year 2007, while FHA loans increased sharply by 390%.

This remarkable trend reversal provides evidence that economic factors that lead to an increase in FHA mortgages may exist, despite the declining trend in mortgages (in general).

P2P lending started in the U.S. in February 2006 with the launch of Prosper Marketplace, followed by Lending Club. We use a data set from Lending Club, the largest (based upon issued loan volume and revenue) online lender in the U.S. and worldwide. Lending Club matches borrowers’ demand and lenders’ supply for funding, without the intermediation of traditional banks. It does not originate loans itself but relies on WebBank, an online bank whose main activity is to finance peer-to-peer lending platforms.

Loan amounts requested on Lending Club vary from 1000 to 40,000 USD, and borrowers may ask for a loan for different reasons, such as debt consolidation, large purchases, or credit card refinancing.

Table 1. Lending Club dataset (P2P loan volumes, number of P2P loans, and number of states).

Lending Club	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016	2017
Volume (in million USD)	5	21	51	131	254	716	1918	3494	6401	6225	6559
No. of Loans	602	2286	5245	12,469	21,090	53,205	130,708	235,026	419,885	433,637	442,001
No. of States	37	46	42	43	44	45	48	48	48	49	49

shows the total volume of funded P2P loans, the total number of loans approved, and the number of states that participated in the P2P process from 2007 to 2017.

For the period from 2007–2017, Lending Club facilitated more than 25.7 billion USD in loans (total number of loan originations: 1.8 million). The loan originations progressively increased until 2012, and from 2013 onward, they experienced exponential growth. Figure 2 depicts the growth in P2P loans over the period analyzed.

Table A1. Yearly FHA mortgage origination (numbers and volumes) by U.S. state.

State Name	Loans	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016	2017
Alabama	numbers	10,007	24,895	29,109	22,763	19,269	18,771	16,719	11,333	17,255	20,429	18,371
	volumes	2,344,328	3,278,319	4,078,684	3,331,956	2,248,117	2,553,839	2,365,427	1,501,147	2,438,869	2,971,999	2,703,512
Alaska	numbers	1855	3792	5270	4201	3403	3870	2939	2178	3015	2713	2136
	volumes	374,738	799,994	1,116,827	918,931	705,289	838,690	644,072	482,236	735,486	673,518	530,991
Arizona	numbers	10,934	40,791	55,689	43,225	32,808	38,525	34,616	29,110	49,192	48,322	39,194
	volumes	1,946,794	7,341,106	9,201,438	6,787,491	4,557,582	6,219,334	5,769,816	4,861,489	8,874,732	8,912,700	7,470,663
Arkansas	numbers	6292	13,412	16,974	13,544	9967	10,998	9205	6431	9292	10,382	9554
	volumes	669,692	1,651,794	2,197,634	1,724,136	1,263,267	1,404,026	1,148,663	834,634	1,213,734	1,412,320	1,338,445
California	numbers	8547	88,874	179,181	163,952	129,192	163,841	114,116	79,979	142,264	154,403	112,138
	volumes	2,096,890	23,004,689	46,398,321	45,356,628	33,242,951	44,591,441	31,813,515	22,782,536	44,211,028	45,661,931	36,023,913
Colorado	numbers	14,167	37,062	57,421	41,685	27,686	39,050	31,985	22,376	35,725	36,215	32,334
	volumes	2,494,412	7,074,276	11,445,023	8,588,598	5,284,279	7,690,818	6,393,137	4,818,388	8,227,414	9,623,720	8,420,659
Connecticut	numbers	6266	13,540	20,920	15,899	11,228	14,137	12,220	7216	11,542	12,370	11,717
	volumes	1,271,939	2,891,313	4,621,540	3,597,370	2,388,390	2,996,856	2,522,750	1,431,516	2,413,272	2,613,220	2,430,901
Delaware	numbers	1760	5096	7086	5468	3852	4457	3980	2590	4754	5323	4729
	volumes	322,983	1,047,454	1,495,007	1,148,200	747,926	868,474	809,349	474,528	942,084	1,124,475	991,042
Florida	numbers	24,361	62,268	81,760	71,248	56,577	62,939	61,825	52,939	82,484	98,859	93,074
	volumes	3,906,447	10,513,056	12,928,652	10,869,158	8,315,867	9,848,490	10,286,832	8,883,540	14,737,306	18,692,357	18,077,196
Georgia	numbers	27,089	55,633	68,292	49,268	37,230	46,292	42,424	31,383	49,495	56,736	51,810
	volumes	3,870,162	8,590,113	10,783,089	7,459,137	5,251,401	6,791,406	6,277,571	4,741,262	8,047,389	9,613,499	9,065,960
Hawaii	numbers	617	1007	2544	2475	1490	1546	1219	701	1210	1335	1306
	volumes	143,171	343,488	878,766	886,945	510,367	523,002	391,786	260,634	560,372	532,514	547,348
Idaho	numbers	3849	9847	12,837	10,370	7102	7897	6543	5468	9503	9972	8087
	volumes	564,683	1,556,636	1,949,740	1,502,673	943,423	1,111,375	989,050	817,379	1,577,082	1,672,982	1,410,530
Illinois	numbers	18,881	51,948	76,684	49,550	37,370	45,811	40,550	28,421	42,716	46,462	40,172
	volumes	2,984,275	9,159,131	13,910,895	8,769,327	6,033,205	7,768,492	6,670,758	4,476,906	7,493,734	8,247,469	7,299,142
Indiana	numbers	17,377	38,159	50,990	36,401	27,912	34,468	31,100	21,599	31,641	33,972	31,207
	volumes	2,205,972	4,716,339	6,767,092	5,050,934	3,889,165	4,953,455	4,153,620	2,876,453	4,320,388	4,692,474	4,578,013
Iowa	numbers	4402	11,965	20,167	15,512	9954	10,417	7934	5632	8147	8182	6910
	volumes	466,854	1,432,551	2,569,146	2,119,922	1,239,110	1,377,196	1,050,977	697,105	1,107,619	1,090,418	959,193
Kansas	numbers	6063	13,892	20,556	14,985	10,615	11,924	9743	7246	10,284	10,276	9393
	volumes	673,618	1,773,184	2,829,220	2,033,753	1,443,412	1,646,724	1,287,942	918,814	1,486,443	1,443,830	1,348,832
Kentucky	numbers	8086	16,923	22,488	18,606	13,763	15,961	14,223	9717	15,062	16,848	15,226
	volumes	954,884	2,187,345	3,095,901	2,545,093	1,866,404	2,169,212	1,915,011	1,293,266	2,040,861	2,335,935	2,157,339
Louisiana	numbers	5577	14,102	19,880	17,648	13,639	15,482	13,203	10,155	14,893	16,300	14,325
	volumes	723,210	2,054,648	3,015,949	2,726,585	2,123,934	2,356,982	2,062,584	1,654,512	2,578,206	2,693,360	2,369,576
Maine	numbers	1236	3850	5780	4341	3238	3819	2600	1873	3132	3820	3573
	volumes	192,902	637,159	996,224	738,978	520,541	645,411	454,293	290,316	534,005	656,110	626,783
Maryland	numbers	11,563	36,328	51,721	38,156	26,121	32,589	26,649	17,623	33,298	35,511	29,794
	volumes	2,603,829	9,248,148	13,458,107	10,031,837	6,392,617	8,373,731	6,778,483	4,282,974	8,877,566	9,598,855	8,078,286
Massachusetts	numbers	3617	16,865	30,962	25,603	16,707	19,264	14,602	9885	17,896	18,991	16,457
	volumes	900,780	4,169,051	8,002,952	7,061,338	4,147,324	4,971,662	3,725,716	2,545,512	4,816,685	5,275,117	5,063,311
Michigan	numbers	17,888	41,440	51,370	36,655	31,204	39,579	37,551	28,717	40,067	40,978	35,454
	volumes	2,171,393	5,226,315	6,442,743	4,531,828	3,729,170	5,032,577	4,929,188	3,848,465	5,466,709	5,801,018	5,017,787
Minnesota	numbers	5880	22,525	38,662	29,674	22,651	29,610	22,883	14,357	23,730	23,068	19,634
	volumes	987,879	3,984,106	6,767,731	5,222,097	3,700,586	5,314,149	4,168,791	2,555,639	4,383,218	4,391,894	4,040,296
Mississippi	numbers	4264	8443	10,683	8899	6833	7519	6680	5118	7551	8853	8994
	volumes	506,032	1,124,047	1,463,573	1,194,164	906,319	1,019,604	941,551	682,164	1,046,971	1,265,512	1,340,750
Missouri	numbers	14,297	36,818	51,990	34,872	24,495	31,287	25,540	16,575	26,541	28,669	25,187
	volumes	1,749,334	4,961,102	7,302,330	4,867,403	3,378,388	4,339,579	3,383,973	2,148,321	3,782,090	4,132,811	3,664,685
Montana	numbers	2040	4330	6433	4609	2974	3222	2736	2024	3137	3152	2901
	volumes	273,170	695,638	1,077,553	754,250	473,648	523,820	454,858	339,860	575,303	616,668	564,877
Nebraska	numbers	2375	7881	15,496	11,829	7699	8904	7040	4848	7035	6970	5532
	volumes	280,866	1,009,326	2,043,662	1,584,861	1,012,352	1,182,901	965,005	614,329	1,012,762	1,009,005	816,019
Nevada	numbers	10,238	43,264	64,578	51,230	38,769	42,462	34,903	28,168	44,956	51,456	40,065
	volumes	2,206,748	9,070,546	11,358,680	8,440,528	5,993,555	7,041,613	9,136,285	5,108,178	8,573,734	10,066,453	9,019,218
New Hampshire	numbers	783	4049	7459	6138	4430	5421	4389	3042	5362	5770	4926
	volumes	165,996	873,522	1,609,585	1,311,418	947,365	1,097,900	896,906	612,220	1,127,913	1,233,116	1,098,559
New Jersey	numbers	11,879	31,724	54,668	36,258	26,384	32,900	27,317	17,597	32,262	35,270	31,029
	volumes	2,840,221	8,183,286	14,454,586	9,473,011	6,545,745	8,264,974	6,708,618	4,184,173	8,197,643	8,965,115	7,860,588
New Mexico	numbers	3949	9351	13,280	10,619	7371	9250	8155	5589	8112	8891	7736
	volumes	543,366	1,445,930	2,104,651	1,631,528	1,096,929	1,401,303	1,207,420	812,125	1,269,755	1,431,989	1,233,369
New York	numbers	16,152	32,759	51,435	44,357	38,881	39,900	33,357	24,167	34,580	37,398	32,860
	volumes	2,430,715	139,156,284	11,224,820	10,381,914	8,787,457	9,309,548	7,750,718	5,523,950	8,702,391	9,253,347	8,438,396
North Carolina	numbers	17,527	42,683	52,242	38,181	27,006	31,367	29,656	19,508	30,482	33,231	29,838
	volumes	2,342,839	6,241,933	8,024,001	6,057,219	4,125,521	5,014,615	4,533,340	2,922,464	4,684,489	5,499,047	4,976,273
North Dakota	numbers	1403	2717	3960	3861	2870	3019	2333	1854	2669	2458	2059
	volumes	150,747	334,448	528,299	534,891	421,737	469,529	377,435	324,304	493,710	465,823	384,654
Ohio	numbers	24,151	54,044	73,971	53,028	38,050	50,055	48,921	33,241	45,532	49,115	43,971
	volumes	3,011,455	6,907,777	10,104,589	7,291,898	4,919,937	6,736,083	6,337,953	4,011,337	5,973,757	6,670,530	6,264,297
Oklahoma	numbers	8330	19,837	27,613	21,068	16,420	17,921	14,674	12,222	14,728	15,896	14,566
	volumes	868,183	2,358,885	3,516,646	2,710,686	2,119,029	2,431,765	2,024,925	1,695,249	2,069,946	2,211,946	2,036,065
Oregon	numbers	4172	15,867	23,378	18,932	12,471	16,889	13,258	8945	16,359	16,598	14,381
	volumes	801,667	3,331,504	5,022,733	4,063,529	2,702,781	3,489,600	2,720,653	1,829,686	3,651,401	3,811,353	3,411,077
Pennsylvania	numbers	16,626	45,955	71,922	57,142	42,357	47,515	39,555	27,683	42,220	47,010	41,168
	volumes	2,178,041	7,110,253	11,906,541	9,343,193	6,637,483	7,761,038	6,167,154	4,017,724	6,660,292	7,634,675	6,665,558
Rhode Island	numbers	909	3846	6679	5105	3713	4711	4024	2610	4946	5381	5040
	volumes	202,989	824,225	1,392,988	1,077,098	732,214	934,122	774,643	459,674	951,572	1,111,925	1,104,641
South Carolina	numbers	6280	18,830	26,797	19,732	14,088	16,250	15,209	11,576	18,448	21,478	20,299
	volumes	857,619	2,742,942	4,057,878	2,973,121	2,081,416	2,556,390	2,379,175	1,755,881	2,874,950	3,454,094	3,298,975
South Dakota	numbers	938	2917	4491	3859	2723	3115	2437	1543	2581	2540	2306
	volumes	108,268	393,930	621,209	606,897	382,553	439,184	341,594	217,228	412,928	419,150	369,850
Tennessee	numbers	14,882	34,044	46,813	34,563	25,171	31,279	27,533	18,657	28,438	31,253	29,302
	volumes	1,910,059	4,758,518	6,890,125	5,164,718	3,788,566	4,799,829	4,246,658	2,905,476	4,796,969	5,376,985	5,131,370
Texas	numbers	50,594	93,218	128,474	107,189	87,982	108,223	107,479	78,739	101,604	110,847	95,743
	volumes	6,215,503	12,633,395	18,256,883	15,667,720	12,766,876	15,953,047	16,307,882	12,221,122	17,364,869	20,058,228	18,640,088
Utah	numbers	9598	25,108	38,222	29,816	16,786	27,452	17,970	12,369	23,500	23,254	18,697
	volumes	1,600,967	4,944,125	7,609,881	5,986,409	3,070,730	5,158,900	3,469,221	2,340,611	4,826,728	4,943,182	4,247,342
Vermont	numbers	310	1062	1978	1297	835	880	749	611	976	1131	965
	volumes	50,172	185,789	357,972	240,184	151,395	159,555	133,541	114,802	183,664	204,717	179,721
Virginia	numbers	11,897	42,731	62,249	45,204	32,510	39,633	32,630	20,581	35,604	37,239	31,103
	volumes	2,177,473	9,211,633	14,340,026	11,031,500	7,683,039	9,739,110	7,749,326	4,554,897	8,878,518	9,215,199	7,864,629
Washington	numbers	10,232	33,254	52,144	38,301	25,501	34,478	25,861	17,477	30,551	32,594	27,697
	volumes	2,097,795	7,697,463	12,337,924	9,073,911	5,807,555	7,702,214	5,854,336	3,826,055	7,337,332	8,141,704	7,304,159
West Virginia	numbers	1841	4947	6058	5010	3815	3939	3409	2511	3468	3967	3967
	volumes	231,850	662,400	894,560	665,530	492,623	539,852	484,882	315,222	489,607	556,082	544,627
Wisconsin	numbers	7255	17,836	28,322	20,540	13,221	17,664	14,187	9320	14,535	15,232	12,776
	volumes	989,129	2,686,490	4,337,405	3,267,261	1,997,317	2,754,257	2,153,180	1,364,289	2,152,077	2,364,673	2,030,557
Wyoming	numbers	1328	3477	4459	3171	2463	2682	2206	1838	2576	2474	2218
	volumes	194,231	590,737	772,598	538,761	422,701	459,150	396,217	329,262	511,632	448,442	404,393
total	numbers	470,564	1,265,206	1,862,137	1,446,039	1,078,796	1,309,184	1,111,037	795,342	1,245,350	1,349,594	1,161,921
	volumes	71,857,270	352,816,343	338,562,379	268,936,518	189,989,558	241,326,824	204,506,780	142,559,854	245,687,205	270,293,486	239,444,455

Source: Authors’ calculations based on historical Home Mortgage Disclosure Act (HMDA) data provided by the Consumer Financial Protection Bureau.

and

Table A2. Yearly P2P loan origination (numbers and volumes) by U.S. state.

State Name	Loans	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016	2017
Alabama	numbers	9	20	56	156	238	696	1602	3035	5347	5329	5142
	volumes	61,700	199,725	492,125	1,527,100	3,065,725	9,731,725	23,100,125	45,167,175	80,590,675	74,902,950	73,776,825
Alaska	numbers	0	5	9	21	49	180	375	599	954	1006	970
	volumes	0	61,500	112,325	224,000	667,775	2,778,500	6,272,625	10,285,300	16,361,125	17,297,625	16,697,225
Arizona	numbers	14	53	124	258	468	1196	2994	5509	9700	10,462	10,736
	volumes	126,900	462,100	1,238,075	2,713,050	5,461,700	14,889,250	42,835,850	78,694,450	142,326,550	149,029,450	153,743,125
Arkansas	numbers	3	8	24	80	141	424	972	1808	3152	3335	3284
	volumes	7600	72,150	251,500	791,375	1,525,550	5,343,300	13,234,625	25,833,125	45,878,500	45,923,750	44,037,125
California	numbers	7	456	894	2215	3740	9277	20,885	33,290	58,066	57,888	59,769
	volumes	73,225	4,401,175	9,217,625	23,760,550	44,970,150	123,410,375	302,006,300	488,123,300	891,314,075	880,078,725	917,950,625
Colorado	numbers	19	66	117	272	377	1002	2867	4931	9068	8800	9454
	volumes	146,225	503,800	1,106,600	2,975,400	4,794,375	13,436,675	42,526,475	74,817,350	139,447,350	128,861,575	139,209,475
Connecticut	numbers	12	42	99	260	393	864	1995	3485	6327	6794	7492
	volumes	87,950	413,700	960,575	2,632,600	5,022,675	11,837,625	30,273,300	51,439,700	99,382,550	102,855,450	112,296,900
Delaware	numbers	1	8	23	48	56	120	344	663	1239	1252	1243
	volumes	6000	74,125	232,800	470,300	601,500	1,656,575	5,081,775	9,981,125	18,608,750	17,587,675	18,663,300
Florida	numbers	85	162	396	857	1566	3987	8666	15,693	29,298	31,727	31,890
	volumes	763,850	1,349,125	3,661,725	8,682,375	17,974,550	50,401,425	119,450,875	220,137,275	423,257,500	444,092,250	456,796,250
Georgia	numbers	37	95	189	459	707	1639	4123	7602	14,110	14,460	14,242
	volumes	364,800	849,650	1,774,800	4,802,850	8,802,100	22,453,175	61,551,325	114,642,800	217,469,100	216,785,925	215,797,475
Hawaii	numbers	0	11	18	44	101	316	774	1198	2083	2019	1997
	volumes	0	66,800	159,325	460,450	1,225,100	4,372,575	11,590,150	18,382,550	32,817,575	31,891,950	31,575,525
Idaho	numbers	0	9	0	0	0	0	2	1	0	1225	1324
	volumes	0	72,350	0	0	0	0	9000	7000	0	17,485,225	18,416,525
Illinois	numbers	3	95	209	509	836	2048	5116	9630	16,864	17,715	18,292
	volumes	23,750	801,700	2,224,775	5,709,675	9,998,950	28,813,000	76,588,375	147,590,100	264,429,975	268,226,575	280,893,725
Indiana	numbers	19	0	0	0	0	27	2109	4208	7264	7658	7676
	volumes	144,075	0	0	0	0	358,575	32,896,050	61,663,850	106,329,475	108,950,700	109,450,125
Iowa	numbers	0	11	0	1	0	0	1	1	0	0	0
	volumes		90,550	0	9600	0	0	7200	6725	0	0	0
Kansas	numbers	5	26	48	28	187	503	1243	2105	3736	3567	3677
	volumes	11,600	183,125	458,350	296,400	2,174,600	6,978,450	18,042,500	30,912,975	56,531,625	51,428,800	53,287,650
Kentucky	numbers	1	14	37	102	201	460	1216	2381	4112	4225	4306
	volumes	7000	127,650	327,400	962,900	2,433,025	6,323,875	17,391,500	34,208,550	59,884,500	59,164,475	60,120,725
Louisiana	numbers	8	19	51	131	246	655	1570	2762	5098	4985	4981
	volumes	42,825	107,450	438,825	1,331,075	2,804,075	9,128,150	23,225,450	41,046,825	77,991,300	72,315,575	71,933,675
Maine	numbers	3	0	0	0	0	0	0	1	521	1410	1436
	volumes	9200	0	0	0	0	0	0	29,850	8,092,775	20,426,700	20,191,300
Maryland	numbers	23	62	142	348	538	1183	3008	5566	10,057	10,071	10,831
	volumes	214,525	514,500	1,455,075	3,707,825	6,745,775	16,848,275	46,615,425	87,121,425	160,386,200	158,010,875	169,454,050
Massachusetts	numbers	52	65	192	477	643	1373	2937	5233	9521	9830	10,161
	volumes	339,375	583,075	2,057,875	5,091,975	7,963,500	20,252,100	47,807,300	85,234,050	151,888,350	150,627,725	156,205,850
Michigan	numbers	0	53	123	251	363	1210	3238	6206	11,436	11,604	11,560
	volumes	0	452,775	1,131,475	2,697,500	4,393,100	15,922,075	46,245,500	87,800,700	166,060,275	160,990,375	162,605,725
Minnesota	numbers	0	38	80	195	335	885	2272	4277	7800	7700	7764
	volumes	0	342,725	725,250	2,004,300	3,665,200	11,634,450	32,412,725	60,713,375	116,009,600	109,975,875	111,637,425
Mississippi	numbers	0	12	13	0	1	1	2	1198	2593	2867	2855
	volumes	0	86,575	97,875	0	6000	4000	35,975	18,391,075	38,427,250	40,885,550	40,940,175
Missouri	numbers	22	38	73	234	392	845	2050	3789	6698	6951	7058
	volumes	142,250	266,250	660,275	2,444,200	4,378,025	11,123,800	29,600,025	55,070,575	100,114,525	99,016,450	99,619,525
Montana	numbers	3	12	7	25	49	164	395	730	1162	1172	1224
	volumes	52,400	71,400	52,350	189,100	620,075	2,072,350	5,366,050	10,197,975	16,467,725	16,221,625	16,016,725
Nebraska	numbers	8	3	0	0	0	1	2	0	1162	2116	2131
	volumes	50,400	38,550	0	0	0	22,250	17,200	0	17,027,650	28,934,650	29,054,575
Nevada	numbers	2	29	66	124	300	792	1907	3225	5924	6164	6420
	volumes	21,625	249,250	647,725	1,304,175	3,404,775	10,343,175	26,298,675	45,016,500	84,099,025	87,709,500	91,922,875
New Hampshire	numbers	6	11	25	56	89	249	623	1155	2062	2097	2279
	volumes	50,500	105,275	208,850	611,950	1,184,700	3,352,125	8,943,525	17,581,550	31,202,825	30,697,325	33,289,475
New Jersey	numbers	45	95	240	612	970	2100	4980	8863	15,195	15,891	16,164
	volumes	351,800	861,075	2,440,250	6,598,125	12,518,350	29,079,350	76,890,850	139,586,200	244,526,800	247,077,300	253,929,275
New Mexico	numbers	4	15	29	65	90	278	716	1367	2351	2286	2212
	volumes	47,400	153,775	277,575	671,800	948,100	4,007,625	10,911,650	20,619,600	35,603,000	33,214,525	31,749,250
New York	numbers	101	78	511	1246	1984	5102	10,848	19,923	33,844	35,505	37,153
	volumes	965,575	1,655,750	5,161,950	13,506,275	23,516,575	69,237,700	155,520,725	293,232,400	508,856,550	357,662,975	545,586,050
North Carolina	numbers	33	46	0	89	652	1587	3713	6442	12,037	12,279	12,081
	volumes	223,650	384,875	0	958,425	7,653,325	20,604,425	53,347,500	94,354,950	182,433,925	177,079,475	175,260,325
North Dakota	numbers	0	0	0	0	0	0	0	0	480	1051	1011
	volumes	0	0	0	0	0	0	0	0	7,447,300	15,927,050	14,728,100
Ohio	numbers	2	75	186	378	672	1552	4256	8011	14,394	15,000	14,436
	volumes	7500	692,075	1,699,950	3,705,400	7,958,400	20,812,825	60,551,275	114,606,675	211,130,650	208,673,075	201,676,300
Oklahoma	numbers	0	15	36	85	180	444	1196	2118	3768	4061	3937
	volumes	0	103,225	345,500	932,650	2,158,750	5,697,300	17,715,500	31,491,175	60,860,200	58,976,925	58,309,300
Oregon	numbers	0	26	56	143	233	665	1850	2912	4944	4669	5184
	volumes	0	181,650	508,575	1,444,050	2,727,850	8,183,725	25,149,100	40,318,600	69,877,175	64,481,375	73,306,975
Pennsylvania	numbers	0	83	245	534	772	1747	4475	8426	14,969	14,502	15,026
	volumes	0	584,900	2,314,400	5,412,950	8,836,325	22,906,825	63,729,200	123,262,525	226,444,700	210,137,900	216,211,750
Rhode Island	numbers	0	16	29	59	100	245	536	1043	1843	1881	1964
	volumes	0	135,025	243,950	583,800	970,375	3,082,000	7,646,300	14,890,425	27,194,800	25,689,700	27,588,775
South Carolina	numbers	6	21	64	133	261	631	1448	2971	5056	5492	5444
	volumes	52,150	151,375	640,125	1,395,025	3,058,425	8,467,375	21,542,850	44,817,100	76,763,600	79,521,350	79,821,625
South Dakota	numbers	1	4	6	16	39	120	280	506	835	879	866
	volumes	25,000	10,575	57,100	150,750	401,600	1,431,725	3,747,325	7,122,525	11,814,875	12,552,925	11,938,250
Tennessee	numbers	1	30	0	0	1	0	2046	3907	6902	7272	7112
	volumes	2825	222,125	0	0	28,000	0	29,895,425	58,688,950	103,132,650	103,680,150	99,931,725
Texas	numbers	5	170	381	887	1436	4265	10,004	18,967	34,698	37,036	37,218
	volumes	25,350	1,606,000	3,661,125	9,566,550	18,404,175	59,458,575	154,894,925	298,467,650	557,706,125	568,130,525	574,759,225
Utah	numbers	9	19	39	77	132	383	1064	1669	2836	2796	2770
	volumes	71,925	180,325	323,300	817,950	1,669,775	5,006,175	15,584,050	24,890,575	42,769,125	41,870,950	41,447,750
Vermont	numbers	1	4	5	14	32	131	169	542	892	942	1041
	volumes	2500	48,000	37,700	123,550	292,450	1,643,175	2,290,225	7,669,800	12,974,450	12,670,600	13,707,400
Virginia[E]	numbers	20	91	188	478	694	1674	3985	6965	12,032	11,687	12,027
	volumes	153,900	878,250	2,009,175	4,878,850	8,962,450	23,817,750	63,034,900	110,375,675	194,941,075	185,924,000	193,311,525
Washington	numbers	12	49	113	241	456	1229	3200	5118	8917	8301	9031
	volumes	121,725	442,275	1,100,150	2,502,325	5,262,375	16,127,900	47,348,325	76,440,825	138,185,650	125,420,725	137,568,900
West Virginia	numbers	0	2	26	67	89	221	681	1311	1975	770	369
	volumes	0	9000	221,900	588,225	1,157,025	2,936,475	10,018,675	19,135,575	29,911,400	11,095,800	5,734,450
Wisconsin	numbers	18	24	61	165	241	609	1650	3101	5663	5973	5939
	volumes	150,400	160,925	603,100	1,649,525	3,042,775	8,302,350	23,553,125	44,642,175	82,637,750	80,786,825	83,999,850
Wyoming	numbers	2	0	15	29	40	125	323	583	900	935	822
	volumes	20,000	0	156,425	284,650	482,175	1,726,725	5,211,075	9,148,875	13,874,375	14,339,650	12,769,625
total	numbers	602	2286	5245	12,469	21,090	53,205	130,708	235,026	419,885	433,637	442,001
	volumes	4,969,475	21,008,250	51,495,825	131,171,600	253,932,275	716,017,850	1,918,008,900	3,493,859,525	6,401,483,000	6,225,289,125	6,558,920,400

Source: Authors’ calculations based on the Lending Club dataset available from https://www.lendingclub.com/info/download-data.action (accessed on 29 June 2021).

in Appendix A show the loan volume (USD) and the number of mortgages and P2P loans, respectively, issued by each state for each year under study.

Examining the volume of FHA mortgage and the P2P loan origination in each of the U.S. states, we observe that there are ten states with the highest origination of P2P loans and FHA mortgages. More specifically, 50% of total endorsements of U.S. FHA mortgages are issued in these ten states with more than 60% of total P2P endorsements originated in the same states.

Table A3. Yearly percentage of FHA mortgage origination volume (comparing to the total FHA mortgage volume) for each of the 10 states with the highest concentration of FHA mortgages.

State	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016	2017
California	2.9%	6.5%	13.7%	16.9%	17.5%	18.5%	15.6%	16.0%	18.0%	16.9%	15.0%
Texas	2.4%	3.6%	5.4%	5.8%	6.7%	6.6%	8.0%	8.6%	7.1%	7.4%	7.8%
Florida	5.4%	3.0%	3.8%	4.0%	4.4%	4.1%	5.0%	6.2%	6.0%	6.9%	7.5%
Georgia	5.4%	2.4%	3.2%	2.8%	2.8%	2.8%	3.1%	3.3%	3.3%	3.6%	3.8%
New York	5.4%	13.4%	3.3%	3.9%	4.6%	3.9%	3.8%	3.9%	3.5%	3.4%	3.5%
Pennsylvania	1.0%	2.0%	3.4%	3.2%	2.8%	3.2%	3.1%	3.4%	3.3%	3.6%	3.5%
Maryland	3.6%	2.6%	4.0%	3.7%	3.4%	3.5%	3.3%	3.0%	3.6%	3.6%	3.4%
New Jersey	4.0%	2.3%	4.3%	3.5%	3.4%	3.4%	3.3%	2.9%	3.3%	3.3%	3.3%
Virginia	3.0%	3.5%	4.2%	4.1%	4.0%	4.0%	3.8%	3.2%	3.6%	3.4%	3.3%
Illinois	2.0%	2.6%	4.1%	3.3%	3.2%	3.2%	3.3%	3.1%	3.1%	3.1%	3.0%
Total	46.1%	47.1%	49.0%	51.1%	53.1%	53.2%	53.4%	56.7%	57.2%	58.6%	58.3%

Source: Authors’ calculations based on historical Home Mortgage Disclosure Act (HMDA) data provided by the Consumer Financial Protection Bureau.

in Appendix A presents the percentages of the total FHA mortgage volume issued in each of these ten states during the eleven-year period in question.

Table A4. Yearly percentage of P2P loan volume (comparing to the total P2P loans) for each of the ten states with the highest concentration of P2P loans.

State	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016	2017
California	1.5%	20.9%	17.9%	18.1%	17.7%	17.2%	15.7%	14.0%	13.9%	14.1%	14.0%
Texas	0.5%	7.6%	7.1%	7.3%	7.2%	8.3%	8.1%	8.5%	8.7%	9.1%	8.8%
New York	4.0%	7.9%	10.0%	10.3%	9.3%	9.7%	8.1%	8.4%	7.9%	5.7%	8.3%
Florida	15.4%	6.4%	7.1%	6.6%	7.1%	7.0%	6.2%	6.3%	6.6%	7.1%	7.0%
Illinois	0.5%	3.8%	4.3%	4.4%	3.9%	4.0%	4.0%	4.2%	4.1%	4.3%	4.3%
New Jersey	7.1%	4.1%	4.7%	5.0%	4.9%	4.1%	4.0%	4.0%	3.8%	4.0%	3.9%
Pennsylvania	0.0%	2.8%	4.5%	4.1%	3.5%	3.2%	3.3%	3.5%	3.5%	3.4%	3.3%
Georgia	7.3%	4.0%	3.4%	3.7%	3.5%	3.1%	3.2%	3.3%	3.4%	3.5%	3.3%
Virginia	3.1%	4.2%	3.9%	3.7%	3.5%	3.3%	3.3%	3.2%	3.0%	3.0%	2.9%
Maryland	4.3%	2.4%	2.8%	2.8%	2.7%	2.4%	2.4%	2.5%	2.5%	2.5%	2.6%
Total	59.2%	67.6%	69.2%	68.9%	66.4%	65.3%	61.6%	61.2%	60.9%	60.2%	61.4%

Source: Authors’ calculations based on the Lending Club dataset available from. https://www.lendingclub.com/info/download-data.action (accessed on 29 June 2021).

shows the corresponding percentages of the total P2P loan volumes issued in each of these ten states. Finally, Figure A1 illustrates the evolution of P2P loans and FHA mortgages in each of these ten states over time (2007–2017).

The patterns followed by both types of loans in each of these states over time are remarkably similar, confirming our prediction that there is a positive correlation between the increase in P2P credit and the increase in FHA mortgages.

The diagrams illustrate the housing crisis of 2007–2009. The greatest fluctuations are observed in the first three years of the period analyzed (housing crisis period). Following the first three years, the patterns of the two types of loans become quite similar and smoother without remarkable deviations during the entire period in question.

3. Literature Review

The emerging literature on P2P lending is diverse and rapidly growing. Among the different studies investigating the significant impact of FinTech lending on the housing market, this is the first study, to our knowledge, to examine P2P as a source of credit to which borrowers resort to bypass tightened LTV caps and obtain mortgages from conventional banks, thus ensuring equal opportunities to everybody. In this sense, our research contributes to three aspects of financial literature: residential mortgage lending, LTV policy, and P2P lending.

3.1. Changes in Mortgage Lending after Crisis

In a broad sense, this study is related to the literature investigating the structural changes in residential mortgage lending in the U.S. in the aftermath of the financial crisis; for example, see Refs. [20,21,22,23]. Ref. [24] measures the mortgage originations in order to quantify the tightness of mortgage credit post-crisis. The author finds that credit tightness resulted in more than one million fewer housing purchase loans per year than would have been originated with reasonable lending standards. Moreover, he assesses the policy actions taken by Federal Housing Finance Agency (FHFA) and FHA, concluding that the FHA still has some important actions to undertake.

Ref. [25] investigates the increased demand for FHA loans as a result of the expansion of loan limits in 2008. They find evidence that the boom in FHA led to a riskier pool of borrowers than would have existed in the absence of FHA loans.

Another study [26], using a seemingly unrelated regression model, examines the impact of FHA availability on the overall mortgage market and conventional lending. It shows that FHA endorsements in 2008 were associated with an increase in overall volume originations without substantially displacing conventional loans. In 2014, the FHA impact on overall lending was smaller, but the degree of substitution between FHA and conventional was greater (one-for-one).

3.2. Effectiveness of LTV Cap Policy

This paper is also related to a large body of research, such as [3,27,28,29], which examines the effectiveness of LTV ratio as a macroprudential tool to constrain loan creation.

Using panel data methods from 46 countries, Ref. [3] finds that the imposition of an LTV cap reduces credit by 5.9% one year after its implementation, whereas Ref. [30] finds that macroprudential policies are much more effective in booms than in bust periods.

Based on data from 57 countries over three decades, Ref. [31] investigates the effect of LTV ceiling imposition on house prices and concludes that there are no significant effects of such a policy.

Ref. [4] examines the effects of a maximum LTV policy on the behavior of borrowers and finds that the imposition of a more conservative ceiling reduces both probability of arrears and the mean value of loans. Similarly, Ref. [5] finds the LTV policy to be effective in reducing both household mortgage leverage and systematic risk associated with boom-and-bust cycles in property markets by reducing mortgage delinquency ratios.

Ref. [32] shows that the effectiveness of macroprudential measures in Europe has been limited by circumvention, particularly through nonbanks when nonbanks are subject to less stringent regulation. Ref. [33], similarly, shows evidence of leakages to the shadow banking sector. Using cross-country data, their results suggest that macroprudential policies lead to substitution from bank-based financial intermediation to nonbank intermediation.

3.3. Peer-to-Peer Lending in the U.S.

Finally, this study falls into the rapidly growing literature on P2P lending marketplaces in the U.S.; for example, see Refs. [34,35,36].

Ref. [37] finds that the Lending Club platform penetrates areas where traditional financial institutions are underserved (small number of bank branches per capita), and the portion of the loans increases in areas where the local economy is not performing well.

There are several studies that examine whether crowdlending platforms and traditional banks are substitutes or complements in the credit market. For instance, Ref. [38] shows that a substantial fraction of the P2P loan volume substitutes the small commercial bank personal loan volume. Thus, small banks appear to lower their borrower quality threshold to curtail volume loss.

Ref. [39] shows that the P2P lending market both substitutes and complements the traditional banking system. Using a regulatory change to accounting standards (i.e., the implementation of FAS 166/167 in 2011), as an exogenous shock to bank credit supply, Ref. [39] finds that P2P lending is a substitute for bank lending in terms of serving inframarginal bank borrowers yet complements bank lending with respect to small loans.

Although Ref. [40] examines the effect of P2P lending on the overall volume of small business loans issued by traditional lenders, in our study we investigate the link between marketplace lending and FHA-insured mortgages. Our results are consistent with the hypothesis that P2P platforms complement banks.

A segment of the existing emergent literature that reviews financial technology explores the rise of FinTech lending in the U.S. mortgage market.

Our paper is close to [17], which provides a detailed analysis of the growth of shadow banks, particularly FinTech shadow banks, in the residential lending market. It finds a shift in mortgage lending from banks to shadow banks in response to improved technology and post-crisis regulatory burdens on traditional banks.

Similarly, Ref. [18] shows that FinTech lenders increased their market share of U.S. mortgage lending, as they process mortgage applications faster than other lenders, alleviating capacity constraints associated with traditional mortgage lending.

4. Holmstrom and Tirole’s Model in Our Analysis

The analysis is based on the framework of the Holmstrom and Tirole (1997) model [41], which allows us to formulate our key empirical prediction: the impact of P2P lending on the effects of changes in loan-to-value caps and mortgage down payment borrower requirements, as in Braggion et al. (2019) [19]. The rise in down payment requirements from traditional lenders is analogous to a “collateral squeeze”, which restrains credit in Holmstrom and Tirole’s model. We show that the availability of P2P lending allows borrowers to bypass the important LTV cap, neutralizing its effects such that the levels of new credit are not reduced. These results allow us to formulate the empirical prediction for our test.

We construct a principal–agent equilibrium model, based on the abovementioned model, which explains how loan provision and investments are affected by changes in the supply of capital. The fixed investment scale model has three types of agents: firms, intermediaries (banks), and individual (uninformed) investors. This paper examines the effect of firms’ capital tightening (due to a collateral squeeze) on investment behavior and interest rates in the economy. The increase in down payment requirements in FHA-insured mortgages corresponds to this collateral squeeze.

Firstly, we analyze the possibility for financing a project (home purchase) by using indirect financing from banks (“the two-party contract”).

Each firm (borrower) has an asset (down payment) that is pledged as collateral. If the project (I) costs more than the asset (A) available, the borrower must find external funds (I-A) to undertake the investment. The project yields R (success) or 0 (failure). The probability of success depends on whether the borrower exerts effort after the project has received funding. A borrower needs to exert effort to raise the probability of success to p_H. The borrower’s private benefit (opportunity cost from managing the project diligently) from not exerting effort is B, which lead to p_L, with p_H − p_L = Δp > 0.

If the project fails, neither party obtains anything. If the project succeeds, the borrower is paid R_b > 0 and the bank is paid R_m > 0. If the borrower does not default, the distribution in returns is R = R_b + R_m.

The open market rate of return-on-investment capital is γ.

The participation constraint for the bank is

p_HR_m ≥ γ (I − A)

(1)

and for the borrower is p_HR_b ≥ γA. The incentive constraint for the borrower is p_HR_f ≥ p_LR_b + B, where R_F is the risk-free rate, i.e., the borrower must prefer to be diligent, so that R_b ≥ $\frac{B}{\Delta p}$. The maximum expected income for banks must be

$${\mathrm{R}}_{\mathrm{m}}=\mathrm{R}-\frac{B}{\Delta p}$$

(2)

Combining (1) and (2) constraints, we have

$${\mathrm{p}}_{\mathrm{H}}(\mathrm{R}-\frac{B}{\Delta p})\ge \gamma (\mathrm{I}-\mathrm{A})$$

(3)

We rearrange (3) and have the following condition for the banks to originate a loan:

$$\mathrm{A}\ge \overline{A\left(\gamma \right)}=\mathrm{I}-\frac{pH}{\gamma }(\mathrm{R}-\frac{B}{\Delta p})$$

(4)

So, only borrowers with assets (down payment) A ≥ $\overline{A\left(\gamma \right)}$ can purchase a house using a residential mortgage issued by traditional banks.

The rise in down payment requirements that was imposed post-crisis on FHA-insured mortgages equals an increase in minimum required assets ($\overline{A}$ + δ).

Holmstrom and Tirole in 1997 proved that when aggregate firm capital is reduced, investment will decrease and the minimum initial assets will increase. Consequently, poorly capitalized firms will be the first to lose their financing. Similarly, in our study, subprime borrowers, with less available cash, cannot afford the increased minimum down payment and will be excluded from the mortgage market, leading to a decrease in the FHA loan volume.

When we introduce P2P lending in the model, the outcome on the housing market is reversed. When a borrower fails to obtain a mortgage from a bank due to insufficient assets (down payment), they use P2P lending to cover the difference δ and the minimum required down payment (‘the three-party contract”).

The participation constraint for the P2P lender is

$${\mathrm{p}}_{\mathrm{H}}{\mathrm{R}}_{\mathrm{p}2\mathrm{p}}\ge (\overline{A}+\delta )-\mathrm{A}$$

(5)

The participation constraint for the banks becomes

$${\mathrm{P}}_{\mathrm{H}}{\mathrm{R}}_{\mathrm{m}}\ge \mathrm{I}-(\overline{A}+\delta )-\mathrm{A}$$

(6)

The incentive constraint for the borrower remains

$${\mathrm{R}}_{\mathrm{b}}\ge \frac{B}{\Delta p}$$

(7)

The distribution of returns becomes

R = R_m + R_p2p + R_b.

(8)

Therefore, this analysis allows us to formulate the hypothesis that P2P lending has a positive effect on FHA mortgage origination, which is tested in Section 5.

5. Data and Description of Variables and Methodology

To explain the increase in FHA mortgages in the years during and after the financial crisis, we must consider both macroeconomic and financial factors. One of the most important factors is expected to be crowdlending.

We employ annual data over the period 2007–2017 for all 50 U.S. states. Thus, we consider a panel with 50 cross-sectional units and 11 time-series observations per unit. We assemble a hand-collected database containing FHA loan originations and examine nine factors that are considered as important determinants of the changes in the FHA mortgage market. The volume of FHA mortgage loans is the dependent variable in our model and the variables P2P lending, interest rate, GDP per capita, unemployment rate, house price Index, new residential housing units, population growth, broad money, and consumer confidence index are the independent variables.

The following regression model is used for the analysis:

FHA_it = β₀ + β₁P2P_it + β₂IR_it + β₃GDP_it + β₄UR_it + β₅HPI_it + β₆BM_it + β₇NHU_it + β₈POPUL_it + β₉CCI_it + u_it

(9)

Among them, _i identifies a particular state, _t denotes time (year), and FHA is the FHA-insured mortgage volume of state _i in time _t. β₀ is a constant, β₁ to β₉ are the coefficients of the independent variables, and u_it is the random error term. All the variables are summarized in

Table 2. Selected variables.

Variables	Notation	Source	Value
FHA mortgage loans volume (USD)	FHA	Federal Financial Institution Examination Council	USD, in logs
Peer-to-Peer loan volume (USD)	P2P	Lending Club	USD, in logs
Interest rate	IR	The Federal Reserve System	logs
GDP per capita (USD)	GDP	Bureau of Economic Analysis	USD, in logs
Unemployment rate	UR	U.S. Bureau of Labor Statistics	percentage points
House price index	HPI	U.S. Federal Housing Finance Agency	Index
Broad money (USD)	BM	The World Bank	USD, annual growth rate
New residential housing units	NHU	U.S. Census Bureau	logs
Population	POPUL	U.S. Census Bureau	logs
Consumer confidence Index	CCI	Organization for Economic Cooperation and Development (OECD)	Index

.

5.1. Dependent Variable

The volume of FHA mortgage loans is the dependent variable in our model. Information on the majority of U.S. mortgage loans is available in the Home Mortgage Disclosure Act of 1975 (HMDA). The Federal Financial Institution Examination Council (FFIEC) requires financial institutions to maintain, report, and publicly disclose information about home-lending activity every year. Data include, among others, the year, state, loan amount, loan type, and the borrower’s income, race, and ethnicity.

5.2. Explanatory Variables

In the literature, various economic indicators have been proposed as factors that affect the endorsements of mortgages. In this study, we examine the economic and financial factors that may affect the FHA mortgage market. These are in accordance with previous studies, such as [1,42,43], but also enhanced with variables that emerge because of unusual economic conditions (financial crisis).

The P2P is the main variable of interest. Other factors equal, a positive association between FHA-insured mortgage loans and P2P lending is expected, since P2P loans may be used as a source of financing the increased mortgage down payment. This study uses data from Lending Club, the largest P2P platform in the U.S. The original dataset contains 1,756,154 borrowing records.

Many studies consider the interest rate and the GDP per capita as important economic factors that affect the mortgage market.

The interest rate reflects the conditions in the credit market. Higher interest rates increase the cost of borrowing, lowering the demand for houses [44,45]. In this study, we use the effective federal funds rates (EFFR).

GDP per capita is an indicator of economic development and standard of living. Ref. [46] noted that GDP growth has a positive impact on the housing market. High GDP per capita indicates high purchasing power and high demand for houses, increasing the mortgage uptake.

In the literature, the unemployment rate has been assessed as a factor that also affects mortgage endorsements. Ref. [47] argues that the unemployment rate has an impact on the ability of potential borrowers to take up mortgages. Ref. [48] claims that lack of job security affects the ability to access housing loans. Thus, we expect rising unemployment to reduce the demand for housing.

Many studies suggest that population growth is another factor that affects housing demand and explains the fluctuations in the housing market. For instance, Refs. [49,50] study the relationship between demographics and the housing market. We expect that an increase in population increases the demand for residential mortgage loans.

Another variable that is introduced in our model is the house price index. Based on this index, an increase in housing prices is expected to be negatively related to the volume of mortgages.

Two additional factors are also included in our study: broad money and residential housing units. Broad money represents the money supply in an economy. According to [51], excessive money supply may lead to an inflationary environment and might affect investments and mortgage uptake. So, we expect that an increase in broad money has a negative effect on mortgage originations. Additionally, an excess supply of houses leads to a decrease in prices. So, new housing units are expected to be positively correlated to the demand for housing loans.

Finally, the consumer confidence index is another important economic indicator for investors and measures how optimistic consumers are regarding both their financial situation and the overall economic outlook. Confidence increases the willingness for house purchases and leads to an increase in new granted mortgages.

To determine the relevant factors that affect the FHA mortgages, this study employs the panel data model. Before proceeding to estimate the model, a multicollinearity test and correlation analysis was carried out.

6. Presentation and Discussion of Empirical Results

Table 3. Descriptive statistics of all variables.

Variables	Obs.	Mean	Median	Std. Dev.	Min.	Max.	Coefficient of Variation (%)	Skewness	Kurtosis
FHA	550	9.372	9.405	0.524	7.700	11.144	5.60	−0.162	2.733
P2P	550	6.115	6.799	2.280	0.000	8.963	3.72	−1.635	5.042
IR	550	−0.525	−0.584	0.584	−1.046	0.720	−111.21	0.976	2.534
GDP per capita	550	4.693	4.688	0.088	4.493	4.915	1.88	0.258	2.519
UR	550	0.062	0.061	0.022	0.024	0.356	35.58	0.621	2.850
HPI	550	347.14	315.62	103.21	185.37	756.98	0.297	1.036	3.769
BM	550	5.064	4.902	3.352	−2.752	11.713	66.19	−0.383	4.193
NHU	550	7.040	7.084	0.469	5.845	8.247	6.67	−0.085	2.924
POPUL	550	6.586	6.647	0.439	5.728	7.595	6.67	−0.049	2.405
CCI	550	99.09	98.936	1.429	96.742	101.344	0.01	0.034	1.799

presents the descriptive statistics of the variables. Apart from the variables UR, HPI, CCI, and BM, all the other independent variables and the dependent variable are log-transformed to reduce data variability.

The higher the coefficient of variation (CV) is, the greater the level of dispersion around the mean. CCI, HPI, GDP per capita, POPUL, NHU, P2P loans, and FHA loans have a lower CV value. However, this is not the case for UR, IR, and BM growth rate.

Table A5. Median of the chosen variables by U.S. state.

State Name	FHA (in 000s)	P2P (in 000s)	GDP per Capita	UR	HPI	NHU Growth	POPUL Growth
Alabama	2,553,839	9732	38,740	6.80	287.52	−1.35	0.30
Alaska	705,289	2779	74,166	6.95	288.93	1.66	0.91
Arizona	6,787,491	14,889	41,462	6.75	307.15	7.09	1.47
Arkansas	1,338,445	5343	36,488	5.95	251.58	1.71	0.40
California	36,000,000	123,410	56,500	7.45	455.71	4.24	0.83
Colorado	7,690,818	13,437	52,689	4.90	361.25	11.06	1.54
Connecticut	2,522,750	11,838	67,911	7.75	390.20	−9.43	0.17
Delaware	942,084	1657	67,030	6.55	433.92	0.52	0.89
Florida	10,500,000	50,401	40,975	6.25	327.48	5.75	1.28
Georgia	7,459,137	22,453	44,798	7.25	294.12	7.11	1.05
Hawaii	523,002	4373	52,822	4.35	500.81	−7.41	0.96
Idaho	1,410,530	0	36,227	5.60	298.99	8.57	1.24
Illinois	7,493,734	28,813	55,958	7.00	318.08	10.57	0.12
Indiana	4,578,013	359	45,554	5.95	249.49	1.24	0.40
Iowa	1,107,619	0	51,310	4.25	248.65	−1.58	0.53
Kansas	1,443,830	6978	48,787	4.55	238.55	−9.87	0.39
Kentucky	2,157,339	6324	40,198	6.40	291.11	−0.66	0.42
Louisiana	2,356,982	9128	49,683	6.40	246.06	4.87	0.55
Maine	626,783	0	39,817	5.65	467.18	−2.79	0.05
Maryland	8,373,731	16,848	56,830	5.85	433.18	−0.08	0.58
Massachusetts	4,816,685	20,252	66,708	5.75	642.31	−0.70	0.72
Michigan	5,017,787	15,922	42,309	7.90	266.82	11.97	0.05
Minnesota	4,168,791	11,634	55,104	4.95	325.54	2.35	0.69
Mississippi	1,046,971	87	33,386	7.55	248.53	0.60	0.17
Missouri	3,782,090	11,124	44,629	6.10	280.19	−0.98	0.27
Montana	564,877	2072	41,883	5.00	374.56	−0.93	0.84
Nebraska	1,009,326	22	55,363	3.25	252.92	1.95	0.69
Nevada	8,573,734	10,343	47,742	7.85	258.69	8.20	1.50
New Hampshire	1,097,900	3352	51,075	4.25	399.00	−2.13	0.26
New Jersey	8,183,286	29,079	58,742	6.65	483.00	6.37	0.19
New Mexico	1,269,755	4008	41,989	6.70	297.03	−4.17	0.33
New York	8,787,457	69,238	67,614	6.30	579.76	10.17	0.26
North Carolina	4,976,273	20,604	45,083	6.35	321.73	0.26	1.12
North Dakota	421,737	0	67,325	3.10	285.07	1.86	1.47
Ohio	6,337,953	20,813	46,834	6.40	246.87	0.41	0.13
Oklahoma	2,119,029	5697	45,287	4.80	203.62	−7.00	0.81
Oregon	3,411,077	8184	44,736	6.75	406.13	5.17	1.06
Pennsylvania	6,665,558	22,907	50,181	5.85	375.83	−3.09	0.17
Rhode Island	934,122	3082	48,935	7.75	473.31	−2.81	0.01
South Carolina	2,742,942	8,467	37,167	6.55	325.91	10.43	1.28
South Dakota	393,930	1432	52,246	3.40	297.80	−4.91	1.05
Tennessee	4,796,969	222	43,924	6.60	298.44	9.80	0.83
Texas	16,000,000	59,459	54,072	5.15	226.04	4.76	1.77
Utah	4,826,728	5006	45,127	3.80	353.02	8.85	1.91
Vermont	179,721	1643	46,149	4.45	443.02	−1.52	0.09
Virginia	8,878,518	23,818	54,382	5.25	411.49	−0.78	0.97
Washington	7,337,332	16,128	58,086	6.10	435.40	3.90	1.43
West Virginia	539,852	2936	37,340	6.75	213.46	−5.26	0.03
Wisconsin	2,153,180	8302	47,997	5.35	306.65	1.41	0.29
Wyoming	448,442	1727	66,697	4.65	278.39	−1.76	0.83
Total	2,545,303	6298	48,791	6	315.62	1.535	0.649

in Appendix A lists the U.S. states along with the varying median across state variables (FHA mortgages, P2P loans, GDP per capita, unemployment rate, house price index, new residential housing units growth rate, and population growth) during the sample period. As can be observed, there are considerable differences in the values of these parameters across the states.

Although the panel data model can better solve the collinearity problem compared with cross-sectional and time-series models, there is also the possibility of intercorrelations and interassociations among the independent variables. A high degree of multicollinearity makes the model estimation of the coefficients unstable and significantly inflated the standard errors of the coefficients.

Thus, we tested multicollinearity using the variance inflation factor (VIF) method. Based on the results shown in

Table 4. Multicollinearity test.

Variable	VIF	1/VIF
IR	1.98	0.506
UR	2.72	0.367
CCI	1.77	0.564
P2P	1.63	0.612
BM	1.61	0.623
NHU	6.78	0.148
GDP per capita	1.53	0.653
POPUL	7.82	0.128
HPI	1.40	0.715
Mean VIF	3.03

, we can see that the VIF of all variables is below 10; therefore, there is no collinearity problem as there is no linear association among predictor variables. Moreover, the tolerance value (1/VIF) is greater than 0.10, indicating that variables are not correlated.

Correlation analysis is also carried out to measure the strength of the relationship between the variables considered in the study. Pearson’s moment correlation coefficient test is used to measure the degree of correlation between two variables with the correlation coefficient varying between +1 and −1. A value of ±1 indicates a perfect correlation between the two variables.

Table 5. Correlation matrix of all variables.

	FHA	P2P	IR	GDP per Capita	UR	HPI	NHU	BM Growth	POPUL	CCI
FHA	1
P2P	0.3947 ***	1
IR	−0.1304 ***	− 0.266 ***	1
GDP per capita	0.1163 ***	0.2812 ***	−0.0693	1
UR	0.3718 ***	0.0215	−0.4646 ***	−0.2687 ***	1
HPI	0.0844 **	0.1748 ***	0.1643 ***	0.4670 ***	−0.1727 ***	1
NHU	0.7559 ***	0.2778 ***	0.1290 ***	0.0219	−0.0240	−0.0377	1
BM	−0.194 ***	− 0.2541 ***	0.5703 ***	−0.0855 **	−0.3201 ***	0.0631	0.0985 **	1
POPUL	0.8704 *	0.3511 ***	−0.0092	0.0014	0.2729 ***	0.0687 *	0.3755 ***	−0.0094	1
CCI	−0.0950 **	0.3480 ***	0.1526 ***	0.2782 ***	−0.5179 ***	0.1720 ***	0.1813 ***	−0.0469	0.0144	1

Note: The level of significance is noted by * for 10%, ** for 5%, and *** for 1%.

shows the correlation among all the chosen variables.

Our results show that there is a highly negative relationship between interest rate (IR) and unemployment rate (UR) and a highly positive correlation between IR and the broad money growth (BM). The inflationary pressure that the BM creates justifies this highly positive correlation with the IR variable; i.e., a higher inflation rate explains a higher interest rate. On the other hand, job loss and the inability of people to save or borrow money may also justify this highly negative correlation with the interest rate, i.e., the higher the UR the lower the IR.

Afterward, we observe a highly negative correlation between UR and the consumer confidence index (CCI) (−0.5179). A possible explanation is that when the unemployment rate increases, the optimism that the CCI captures decreases. Concerning the positive correlation between GDP per capita and house price index (HPI) (0.4670), it can be explained by the following: as GDP per capita increases, personal income increases as well, which means that the demand in the housing market may also rise, thus increasing house prices.

The panel data estimation is employed to capture the impact of the explanatory variables (P2P lending, interest rate, GDP per capita, unemployment rate, house price index, new residential housing units, population growth, broad money growth and consumer confidence index) on the explained variable (FHA mortgages). For the estimation, two techniques of panel data analysis are used: fixed effects model (FE) and random effects model (RE).

The difference between the FE and RE models is that in the RE model, unlike the FE model, the variation across entities is assumed to be random and not correlated with the predictor or independent variables of the model.

The selection between these two models is made using the Hausman test, with the preferred model being the RE model under the null hypothesis and the preferred model being the FE model under the alternative hypothesis.

The Hausman test gives a ch2 = 61.10 and p-value > chi2 = 0.000, p-value is less than 0.05, demonstrating that the hypothesis test is statistically significant, the null hypothesis is rejected, and the FE model is preferred.

Table 6. Summary of regression results using the FE model.

Dependent Variable: FHA Mortgages	FEM Regression Results
Independent variables	Coefficient	Standard Error
P2P	0.0092 *	0.0049
IR	0.0525 **	0.0203
GDP per capita	0.8357 **	0.3259
UR	−0.7008	0.8029
HPI	0.0003	0.0003
NHU	−0.2633 ***	0.0862
BM	−0.0265 ***	0.0024
POPUL	4.6083 ***	0.8894
CCI	−0.0752 ***	0.0070
Constant	−15.562	5.4837
Observations	550
Number of states	50
R-squared	0.7831
Adjusted R-squared	0.4256
F-statistic	51.64
p-value	0.0000
Hausman test (p-value)	0.0000
Modified Walt test (p-value)	0.0000
Born and Breitung HR-test (p-value)	0.0005

Note: The level of significance is noted by * for 10%, ** for 5%, and *** for 1%.

below presents the estimation results from Equation (9) using the FE model.

The modified Walt test is used for groupwise heteroscedasticity in the residuals of the FE regression model. The results (p-value < 0.05) indicate that we must reject the null hypothesis of homoscedasticity. The heteroscedasticity robust HR-test introduced in Ref. [52] is used to identify serial correlation in the error term in our panel data model. The null hypothesis is no first order autocorrelation. The HR-test confirms a statistically significant correlation in the error term in the model (p-value < 0.05).

Because of the problems of heteroscedasticity and serial correlation, the Driscoll and Kraay standard errors (D-K Std. Err.) estimation model is used to manage these issues. This model proposes a nonparametric covariance matrix estimator that is spatially and temporally independent.

Table 7. Summary of regression results using the Driscoll and Kraay Std. Err. model.

Dependent Variable: FHA Mortgages	Regression with Driscoll–Kraay Std. Err.
Independent variables	Coefficient	Standard Error
P2P	0.0778 ***	0.0078
IR	−0.0772 *	0.0746
GDP per capita	0.9695 ***	0.0984
UR	−0.7180 *	0.0823
HPI	−4.4232	0.0001
NHU	0.3752 ***	0.0306
BM	−0.0279 *	0.0147
POPUL	0.5811 ***	0.0297
CCI	−0.0659 **	0.0284
Constant	27.9458	2.5676
Observations	550
Number of states	50
R-squared	0.8509
Adjusted R-squared	0.6753
Residuals	Normally distributed (Shapiro–Wilk test, p = 0.156)
F-statistic	83,076.22
p-value	0.0000

Note: The level of significance is noted by * for 10%, ** for 5%, and *** for 1%.

below presents the estimation results from Equation (9) using the D-K Std. Err. model.

The regression results with the D-K Std. Err. model show that the value of the adjusted R² has increased in comparison with the adjusted R² with the FE model (which was 0.4256) and is now significant. This adjusted R² of the D-K Std. Err. model (0.67) shows that the independent variables included in the model can explain most of the variation in the dependent variable.

Furthermore, we checked the normality of residuals since it helps to ensure that the model’s assumptions are met and that the data fits well to our model. The assumption is that the error terms are normally distributed; thus, the estimators are efficient and have desirable statistical properties such as minimum variance. To assess normality, we use the statistical test of Shapiro–Wilk. The results of the test indicate that the residuals are normally distributed because the p-value is greater than the conventional significance level of 0.05.

The estimated coefficients of the variables included in the model confirm our hypotheses about the signs of the coefficients. For example, the sign of the interest rate coefficient is now negative, as expected, (while with the FE model the sign is positive) and the sign of the new house units coefficient has a positive and statistically significant effect on FHA mortgages (while with the FE model this correlation was significantly negative). Moreover, the sign of the unemployment rate variable remains negative (as expected), but in the regression with the D-K Std. Err. model, it becomes statistically significant at the 10% level. The P2P, GDP per capita, and POPUL coefficients carry the expected sign and are significant at the 1% confidence interval. The coefficient of P2P lending is positive and highly significant. This finding indicates that as more P2P loans are issued, the higher the volume of FHA mortgage origination. Therefore, the main hypothesis of our analysis is fulfilled (that an increase in P2P loans is strictly related to an increase in FHA mortgages).

The variable GDP per capita has a positive and statistically significant (at 1% level) effect on real FHA mortgages, as expected. Specifically, a 1% increase in GDP per capita is associated with about a 0.97% increase in FHA loans. The same holds between FHA mortgages and population with a 1% increase in the population leading to a 0.58% increase in FHA mortgages.

Finally, BM growth has to have a statistically significant (at 10% level) and negative impact on FHA mortgages, with a 1% increase in the variable leading to a 0.028% decrease in FHA mortgages.

7. Discussion of Results

The main aim of our study is to examine the relationship between the volume of FHA mortgages and P2P lending, considering various economic and financial factors, and the possibility of financial inclusion of underserved borrowers.

Indeed, based on our results, the initial hypothesis of our study, that there is a positive relationship between crowdlending and FHA mortgages, is verified. P2P lending has a positive and statistically significant (at the 1% level) impact on the volume of FHA mortgages, which is in line with the existing literature supporting that FinTech lending has a significant impact on the housing market. However, among the studies that investigate FinTech lending as a substitute for traditional mortgage lending (banks), this study is the first to examine P2P as a source of financing for subprime borrowers to obtain FHA mortgages from conventional banks. Based on our analysis, P2P lending can be a strong alternative for house buyers who do not fulfill the typical criteria required to obtain a FHA mortgage for their purchase. As a result, P2P reinforces the housing market and the economy as a whole, thus contributing to its sustainability.

Another result, also strictly related to sustainability, is the highly positive and statistically significant impact (coefficient 0.969) of the GDP per capita variable on FHA mortgages. This finding is in line with various studies in the literature, such as Refs. [42,46], and shows that a unit increase in the GDP per capita drives a significant increase in the volume of FHA mortgage loans. “GDP per capita” is a crucial variable in our analysis and its increase is directly related to the sustainability of the economy concerning economic growth, higher income, and improvement in people’s welfare that may also drive an increase in the housing demand.

Population growth is another important factor in our study that also has a positive and significant impact (coefficient +0.5811) on FHA originations and promotes sustainability in terms of economic growth, housing demand, and improvement in living conditions. This result is straightforward since the greater the population, the greater the housing needs, and is also shown in the existing literature such as Ref. [49], the results of which are directly comparable to ours (coefficient +0.337).

A positive relationship is also observed between the volume of FHA mortgages and the new residential housing units. This result is strictly related to the law of supply and demand. If the market functions well, then an excess supply of houses lowers house prices, and boosts the housing market and the economy growth. So, we can safely conclude that if the market functions well, a positive relationship between FHA mortgages and new residential housing units contributes to the sustainability of the economy in terms of economic growth and welfare.

Other important variables in our analysis that also significantly affect the volume of FHA mortgage loans in a negative way are broad money growth, unemployment rate, and interest rate. Studies such as Refs. [51,53] concerning money supply, Ref. [54] concerning interest rate, and Refs. [47,48] concerning the unemployment rate, among others, indicate similar results to our study and show this negative relation between those variables and the volume of mortgages. An increase in these variables drives a significant decrease in the volume of FHA mortgages, which actually harms the sustainability of an economy. More specifically, an increase in the interest rate implies an increase in the cost of borrowing: an increase in the unemployment rate reinforces a sense of uncertainty, lowering people’s willing to spend or to borrow, and the broad money growth increases inflationary pressures. An increase in these variables lowers the demand in the housing market, thus increasing poverty and the income discrepancies among segments of the population, and harming sustainability.

8. Conclusions

In this paper, we set under examination the determinants of the FHA mortgage volume in the years following the financial crisis, where mortgage credit was tight and borrowers’ creditworthiness particularly low. We examine all 50 U.S. states in the period from 2007 to 2017 and run a correlation analysis for various variables as determinants of FHA mortgages.

The models used for the relationship analysis between FHA mortgage loans and crowdlending are the fixed effects and the Driscoll and Kraay standard errors models, which are used to address the problems of heteroscedasticity and serial correlation. The variables used in the analysis are interest rate, GDP per capita, unemployment rate, house price index, new residential housing units, population, broad money growth rate, and consumer confidence index.

Based on our results from the panel data analysis, we conclude that P2P lending, the GDP per capita, unemployment rate, interest rate, new residential housing units, population growth, broad money growth rate, and the consumer confidence index are significant determinants for the volume of FHA mortgages. More specifically, we show that there is a positive and significant relationship between the volume of FHA mortgages and P2P lending, GDP per capita, population growth, and new residential housing units, and a significant negative relationship with unemployment rate, interest rate, consumer confidence index and broad money growth rate.

An interesting result of our analysis is that we considered P2P lending as a factor of FHA loan volume, since P2P loans are used as a source for financing the increased down payment requirements. The empirical evidence of this study supports the hypothesis that an increased volume in P2P loans had a positive and significant impact on FHA mortgages for the period 2007–2017, in line with the SDGs for sustainable finance and economic inclusion, adopted by the United Nations.

P2P lending has various financial advantages such as flexibility, simplified process, speed, lower lending standards, and higher returns for investors. Beyond the financial advantages, P2P lending also creates social impact. In this context, crowdlending can provide financial resources to underserved people who have limited or no access to the traditional banking system due to their lack of credit history, lack of collateral, or other barriers. In this way, poverty, economic inequalities, and regional economic disparities may be reduced, and sustainable development with social and economic equity may be promoted.

In our paper, we show that P2P lending can contribute to a country’s sustainable development goals, mainly through the financial inclusion. Thus, this analysis concerning the relationship between the volume of FHA mortgages and various economic and financial factors, with an emphasis on P2P lending, is directly related to the notion of sustainability.

Nonetheless, it is crucial to note that while crowdlending has the potential to contribute to sustainable growth, there are also associated risks, such as borrowers defaulting and regulatory complexities. Thus, it is recommended that policy decision-makers establish a regulatory framework for the crowdlending market, analogous to that of traditional banking system, and an effective mechanism for assessing and managing risks.

Based on our results, the policy intervention that imposes limits on LTV ratios as a tool to constrain mortgage lending has proved ineffective, since it is eluded via the P2P credit channel. Even if P2P loans concern small amounts, they drive an increase in household debt and mortgage lending. In light of this research, one potential approach to contain household leverage would be for macroprudential regulation to design additional tools, to monitor not only LTV but other indicators as well. With appropriate regulation, crowdlending can contribute to a more inclusive and resilient economy.