cumulative distribution functions decreases rather sharply towards the end of the auction, which can be accounted for by the sniping effect that has been observed in other online auctions with a similar structure. There is generally a good amount of bidding activity at the end of the auction, since people are bidding to capture extra benefit awhile trying to avoid bidding wars and bid chasing from the other lenders. End-of-auction bidding is a common practice for both local and nonlocal lenders on Prosper as well. One might be concerned that it is not just timing that matters in determining if a bid is actually early or not; therefore, Table 12 displays the summary statistics for the number of bids placed and money pledged in a listing immediately before a lender bids by local status. As can be clearly seen, local lenders bid earlier in an auction not just hronologically. Local lenders have demonstrated that they are more comfortable bidding in periods where limited information is being revealed by other lenders I have documented that local lenders tend to bid earlier and larger amounts, on average, than nonlocal lenders. additionally local lenders bid different interest rates from nonlocal lenders depending on the ex-post outcome of the loan. This empirical evidence strongly suggest that while altruism might play a part, the dominant channel explaining this observed behavior is that local lenders are better informed than other lenders which fits well with theory. This asymmetric information based on physical geography creates a situation where social learning can occur as local lenders, through their bidding behavior, eveal their private information to nonlocal lenders. As a listing accumulates bids, a general sense of quality is signaled that may lead to rational herding behavior in crowdfunding markets,(Agrawal et al., 2011; Zhang and Lui, 2012; Burtch, Ghose, and Wattal, 2013) Given these results, the following section constructs a theoretic framework grounded in the previously established facts to better understand and explain how the differential behavior of local lenders will affect P2P lending auctions. I develop a simple information-based social-learning model to explain and predict how better informed local lenders reveal their private information to nonlocal lenders through their actions. The model produces the testable hypothesis that listings with more early local lenders will attract more lenders to the listing 5. Motivating model The central proposition of this paper is that local lenders are more informed about local listings than nonlocal lenders. Thus, local lenders are better able to correctly evaluate the underlying risk in the listing when they bid. Being better informed allows local lenders 2 Bajari and Hortacsu(2003); Ariely and Simonson(2003); Ariely, Ockenfels, and Roth(2005);Ockenfels nd roth(2006)
cumulative distribution function’s decreases rather sharply towards the end of the auction, which can be accounted for by the sniping effect that has been observed in other online auctions with a similar structure.20 There is generally a good amount of bidding activity at the end of the auction, since people are bidding to capture extra benefit awhile trying to avoid bidding wars and bid chasing from the other lenders. End-of-auction bidding is a common practice for both local and nonlocal lenders on Prosper as well. One might be concerned that it is not just timing that matters in determining if a bid is actually early or not; therefore, Table 12 displays the summary statistics for the number of bids placed and money pledged in a listing immediately before a lender bids by local status. As can be clearly seen, local lenders bid earlier in an auction not just chronologically. Local lenders have demonstrated that they are more comfortable bidding in periods where limited information is being revealed by other lenders. I have documented that local lenders tend to bid earlier and larger amounts, on average, than nonlocal lenders. Additionally, local lenders bid different interest rates from nonlocal lenders depending on the ex-post outcome of the loan. This empirical evidence strongly suggest that while altruism might play a part, the dominant channel explaining this observed behavior is that local lenders are better informed than other lenders, which fits well with theory. This asymmetric information based on physical geography creates a situation where social learning can occur as local lenders, through their bidding behavior, reveal their private information to nonlocal lenders. As a listing accumulates bids, a general sense of quality is signaled that may lead to rational herding behavior in crowdfunding markets, (Agrawal et al., 2011; Zhang and Lui, 2012; Burtch, Ghose, and Wattal, 2013). Given these results, the following section constructs a theoretic framework grounded in the previously established facts to better understand and explain how the differential behavior of local lenders will affect P2P lending auctions. I develop a simple information-based social-learning model to explain and predict how better informed local lenders reveal their private information to nonlocal lenders through their actions. The model produces the testable hypothesis that listings with more early local lenders will attract more lenders to the listing. 5. Motivating Model The central proposition of this paper is that local lenders are more informed about local listings than nonlocal lenders. Thus, local lenders are better able to correctly evaluate the underlying risk in the listing when they bid. Being better informed allows local lenders 20Bajari and Hortaçsu (2003); Ariely and Simonson (2003); Ariely, Ockenfels, and Roth (2005); Ockenfels and Roth (2006). 16
to be more comfortable bidding earlier, acting mostly on their own private information. Similar to Sorensen(2006)and Smith and Sorensen(2008), local lenders actions serve as signals of their private information. As pointed out by Devenow and Welch(1996), social learning involves an informational externality where lenders may gain useful informatior from observing previous lenders decisions. Due to the existing informational asymmet nonlocal lenders can learn from observing the actions of others, particularly local lenders This behavior leads to listings that have more early local bidding to have more revealed private information available and thus attract more bidding activity To provide some theoretical context and motivation for my analysis, I abstract away from the auction environment on Prosper and develop a simple social learning model with heterogeneous agents in the spirit of Banerjee(1992)and Bikhchandani, Hirshleifer and Welch(1992). There is a population of N identical risk neutral lenders, with VNM utility, who are maximizing their expected utility of monetary payoff from investing. 21 The heterogeneity comes from the existence of two different types of lenders: local and nonlocal. An L share of the lenders are local, who are better informed, and a(1-L)share are nonlocal. Each lender is presented with the same collection of listings indexed by iE 0, 1. The monetary return from investing in the ith listings is z()ER, which is the same for all lenders. Assume that only one unique listing, i', will yield strictly positive returns, while all other listings produce returns of zero, z(i)=0 for all jfi*.This assumption can be interpreted as there existing a single listing which has excess returns that are strictly greater than those of all other listings. Therefore, the optimal ex-post outcome for all lenders is to invest in that listing. Although in Prosper lenders decide how much money to pledge in a particular listing, for simplicity, each lender's decision restricted to simply choosing which listing to invest in At start of the game, all lenders have the same ex-ante uniform priors about which listing will pay positive returns. However, some lenders have an idea of which listing is the likely candidate for i*. Formally, with probability a E(0, 1), lender i will receive a noisy signal, Si E[ 0, 1, about which listing to invest in. The signal need not be true, and will be false with some positive probability. Local lenders are assumed to be more informed than nonlocal lenders, so their signals will be more informative. The signal of a local lender is correct with probability B E(0.5, 1), and with probability(1-B)the signal is strictly noise drawn randomly from u(o, 1. The signal structure for nonlocal lenders 2 If lenders'risk tolerances are identical, risk preference has no effect on the outcome of the game Smith and Sorensen(2000) also study herding with heterogeneous agents, but the types vary along preferences, not the information quality. 2It is possible that a lender's demand level for a particular listing may be a transmission channel for information about his or her private signal. The median bid, regardless of local status, is $50; thus most bids contain no additional information 17
to be more comfortable bidding earlier, acting mostly on their own private information. Similar to Sorensen (2006) and Smith and Sørensen (2008), local lenders’ actions serve as signals of their private information. As pointed out by Devenow and Welch (1996), social learning involves an informational externality where lenders may gain useful information from observing previous lenders’ decisions. Due to the existing informational asymmetry, nonlocal lenders can learn from observing the actions of others, particularly local lenders. This behavior leads to listings that have more early local bidding to have more revealed private information available and thus attract more bidding activity. To provide some theoretical context and motivation for my analysis, I abstract away from the auction environment on Prosper and develop a simple social learning model with heterogeneous agents in the spirit of Banerjee (1992) and Bikhchandani, Hirshleifer, and Welch (1992). There is a population of N identical risk neutral lenders, with vNM utility, who are maximizing their expected utility of monetary payoff from investing.21 The heterogeneity comes from the existence of two different types of lenders: local and nonlocal. An L share of the lenders are local, who are better informed, and a (1 − L) share are nonlocal.22 Each lender is presented with the same collection of listings indexed by i ∈ [0, 1]. The monetary return from investing in the ith listings is z(i) ∈ R, which is the same for all lenders. Assume that only one unique listing, i ∗ , will yield strictly positive returns, while all other listings produce returns of zero, z(j) = 0 for all j 6= i ∗ . This assumption can be interpreted as there existing a single listing which has excess returns that are strictly greater than those of all other listings. Therefore, the optimal ex-post outcome for all lenders is to invest in that listing. Although in Prosper lenders decide how much money to pledge in a particular listing, for simplicity, each lender’s decision is restricted to simply choosing which listing to invest in.23 At start of the game, all lenders have the same ex-ante uniform priors about which listing will pay positive returns. However, some lenders have an idea of which listing is the likely candidate for i ∗ . Formally, with probability α ∈ (0, 1), lender i will receive a noisy signal, Si ∈ [0, 1], about which listing to invest in. The signal need not be true, and will be false with some positive probability. Local lenders are assumed to be more informed than nonlocal lenders, so their signals will be more informative. The signal of a local lender is correct with probability β ∈ (0.5, 1), and with probability (1 − β) the signal is strictly noise drawn randomly from U[0, 1]. The signal structure for nonlocal lenders is 21If lenders’ risk tolerances are identical, risk preference has no effect on the outcome of the game 22Smith and Sørensen (2000) also study herding with heterogeneous agents, but the types vary along preferences, not the information quality. 23It is possible that a lender’s demand level for a particular listing may be a transmission channel for information about his or her private signal. The median bid, regardless of local status, is $50; thus most bids contain no additional information. 17