Abstract

How are neighboring firms affected when a bank learns more about a given firm? We analyze exchange-rate-induced movements of Peruvian firms across a threshold that governs their regulatory treatment by banks. Firms that cross the threshold supply more information to their banks and experience a substantial increase in financing. We find positive spillover effects: the neighbors of the above-threshold firms also experience increased financing. These spillovers are confined to neighbors sharing a bank, and the performance of new loans to these neighbors improves, suggesting that the bank has become better informed about other local firms.

Received October 15, 2015; accepted May 16, 2016 by Editor Efraim Benmelech.

When a firm receives improved access to bank loans, to what extent do other local businesses benefit or suffer? We consider this question in the context of the emerging market of Peru and study the effect of a regulatory information and credit shock to a firm on the firms’ neighboring businesses. Credit constraints are likely to be especially important in developing economies ( McMillan and Woodruff 1999 ; Khwaja and Mian 2005 ). We show that a shock generating both increased information about one firm and enhanced access to credit for the firm leads to increased funding for its neighbors too. Does a neighbor’s increased credit arise from new information about the firm and the neighborhood, or is the increased credit purely a consequence of the new loan to the firm? We attempt to untangle these information and financial aspects of the shock and provide evidence that the greater provision of loans to neighbors is driven by information spillovers, rather than by a financial effect.

Compelling evidence suggests that banks obtain information about their borrowing firms (e.g., Petersen and Rajan 2002 ; Dass and Massa 2011 ). Our results suggest that in the course of learning about one company, banks can also acquire knowledge about their neighbors and lend more to them as well. This provides evidence in support of the importance of information externalities in finance, as emphasized by Benveniste, Busaba, and Wilhelm (2002) . Our findings are also directly related to the arguments of Almazan (2002) and Hauswald and Marquez (2006) that reductions in information costs should allow banks to expand their lending. In essence, we provide evidence that a bank’s information acquisition process can cross firm boundaries and generate deeper insights into hard-to-observe neighborhood characteristics (Kurlat and Stroebel 2015).

This result has implications for bank lending strategies. The presence of information spillovers indicates that banks do not make decisions simply about whether to learn about individual firms; when a bank researches a company, it is also choosing to allocate resources to understanding the broader business matrix in which the firm operates. This suggests that banks are likely to lend to clusters of neighboring firms and provides an argument for banks considering geographic expansion to first focus on regions contiguous to locations in which they are already active, as lenders are likely to know more about areas close to their current clients. Our findings may also interest regulators who offer financing subsidies to small- and medium-sized firms (SMEs) by acting in concert with private financial institutions. 1 These loans may help to spur additional lending to neighboring companies and, as a consequence, should perhaps be first directed at firms located in areas with limited access to credit.

Our empirical approach makes use of Peruvian banking regulations that apply a set of different rules to loans to firms with total debt that exceeds a certain threshold. Firms above this threshold (set at $20,000 to 2003 and $30,000 thereafter) are designated as “Commercial” (COM), and those below it are referred to as “Micro-enterprises” (MES). A crucial distinction between firms in the two categories is that banks lending to COM firms, unlike those lending to MES firms, are required to collect quantitative information about them in the form of financial statements. A shift in status for a firm from MES to COM thus results in a move to significantly more intense information collection and a different regulatory regime for its bank.

The decision to grant a firm COM status should be generally seen as negotiated by the firm and its bank, a result of an endogenous bargaining process. It is the case, however, that some exogenous factors influence this negotiation. In particular, most Peruvian firms borrow in the local currency (the Nuevo Sol, called Sol hereafter), while, for historical reasons, the thresholds have been defined in U.S. dollars during the sample period. As a result, among the class of MES firms with Sol loan balances below but close to the threshold, a group may be pushed above the threshold by Sol-U.S. dollar exchange rate movements in the subsequent months. These exchange rate movements cause some firms to be forced across the threshold and into COM status, while other, very similar firms fall just short of the threshold and remain MES. We make use of this regulatory threshold and of currency movements that are clearly exogenous from the perspective of any firm to implement a regression discontinuity design contrasting future outcomes for firms that are just above and just below the threshold.

We first show that firms with exchange-rate-adjusted balances above the threshold are indeed more likely to be assigned COM status. The relationship is subject to some noise (due, for example, to shifts in the loan balance over the course of the month), but we find a large discontinuous jump in the probability of COM status for firms with exchange-rate-adjusted balances just above this threshold, allowing us to implement a “fuzzy” regression discontinuity design. We also provide evidence in support of the argument that the assignment to COM status is quasi random.

We label the firms close to the threshold “focal firms.” We show that focal firms with exchange-rate-adjusted balances just above the threshold subsequently receive significantly more new financing from their banks in the subsequent twelve months. We thus view exchange-rate-generated transitions of firms to COM status as a regulatory shock that leads them to both supply more information to their banks and to receive more financing ( Stiglitz and Weiss 1981 ).

To uncover the effect of focal firms’ shocks on nearby businesses, we turn our attention toward their “financial neighbors,” a term we define as all businesses within 500 meters of a focal firm and that share a bank with the focal firm. 2 What impact would we expect on the financing of these neighbors? Two arguments suggest that financial neighbor firms should enjoy more financing after a focal firm’s transition to COM. First, the focal firm should be expected to grow more quickly now, and this growth may generate economic benefits for its financial neighbors, leading the financial neighbors to borrow more. Second, a focal firm’s shift to COM status leads its bank to be provided with far more detailed financial information about the focal firm and its neighborhood. From an information perspective, we may expect this reduction in asymmetric information to lead to more lending to financial neighbor firms ( Stiglitz and Weiss 1981 ). An effect of this type illustrates the diffusion of information across firm borders within a banking network and would add a new dimension to our understanding of how asymmetric information operates in influencing financing outcomes.

Conversely, there are two reasons to expect the financial neighbors of focal firms that transition to COM to borrow less. First, it may be that local bank managers have relatively fixed loan allocations ( Zhang 1997 ). This may be due to diversification considerations or to agency problems within large banks that can result in local investment allocations that mainly depend on bank-wide performance and do not fully reflect local opportunities ( Scharfstein and Stein 2000 ; Ozbas and Scharfstein 2010 ). Second, the increased flow of funding to the focal firm may lead its financial neighbors to scale back on investment and financing for competitive reasons.

We estimate the impact of the shock by comparing outcomes for financial neighbors of focal firms just above and just below the threshold. These neighbor firms should be quite alike, except that some experience an increase in information supplied to and financing provided by their bank to the local focal firm and others do not. We find that the financial neighbors of focal firms just above the cutoff subsequently receive approximately 1.4% more new financing in the subsequent twelve months. Financial neighbors even closer to the focal firm (within 50 meters) experience a 6.6% increase in new financing. In other words, financial neighbor firms receive more financing when focal firms provide financial statements to their lenders and receive more loans. This result is consistent with either the broader economic benefits of a loan to a focal firm or with improved bank information after the focal firm’s transition to COM.

To help disentangle these two mechanisms, we also consider the impact of a focal firm’s transition on the financing of its geographic neighbors, defined as those firms within 500 meters of the focal firm, regardless of whether they share a bank. If lending to the focal firm generates pure financial spillovers (e.g., by promoting local economic growth), these should affect all nearby firms, regardless of whether they share a lender. We show that geographic neighbors do not experience an increase in new financing after the focal firm becomes COM; only financial neighbors receive more new loans. This finding suggests that financial spillovers are not driving the increased credit to financial neighbors, leaving the information hypothesis as the theory most consistent with the data. Under the information hypothesis, banks unaffiliated with the focal firm would not receive any additional information after its transition, so these banks’ lending should not be affected. Indeed, we do show this.

To provide more direct evidence for the information effects, we consider whether the performance of new loans to financial neighbors improves after the focal firm crosses the COM threshold. Specifically, we analyze the probability of financial neighbor firm receiving a new loan and subsequently ceasing business operations. If banks are better informed after a focal firm transition, then we would expect to see them making fewer loans to firms that fail. This is, indeed, what we find: the correlation between making a new loan and subsequent firm failure is more negative for financial neighbors of focal firms that traverse the COM eligibility border. Banks are better able to distinguish between the high- and low-quality neighbors of focal firms pushed across the COM boundary by exchange rate movements.

The informational spillovers to financial neighbors and associated improved loan performance that we find raise the question of why banks do not collect financial data from MES firms on their own without any regulatory impetus. Part of the answer may lie in the costs of eliciting this information. The MES firms are quite small, and producing audited financial statements may be very unattractive to them. Indeed, we find that MES firms have borrowing levels clustered just below the COM threshold, consistent with the argument that they are careful to not breach it by small amounts and thereby incur the costs of COM status, including the provision of formal financial data. Firms in developed markets typically would have to be much larger than Peruvian MES firms before being forced to prepare audited statements. Might banks benefit from subsidizing these costs for small borrowers to generate more information? While the transition to COM does lead to more lending to neighbors, some banks may not consider the overall magnitude of the increased lending, or their ability to benefit from it, necessarily larger than the costs they would incur in subsidizing the transition. Our results do suggest, however, that such an approach would be most appealing when a large number of small firms is close to the focal firm.

We study a large sample of heterogeneous small businesses in an emerging market, an empirical context in which shocks to information and financing may be expected to be of first-order importance. Our main finding is that when a firm provides additional information to its lenders and benefits from enhanced financial access, its neighbors also receive more financing. These positive spillovers appear to be driven by an enhanced information flow to the lending bank. Our analysis thus suggests that policies of strengthening anchor firms in a bid to promote their financing, and thereby the financing of other local firms, may meet with some success, if as part of this process lenders are also encouraged to collect more information about the anchor firms. Our findings indicate the need for a more complex understanding of the factors influencing local credit flows.

1. Data

We analyze monthly business bank loan data from Peru for the period 2001-2010. The Peruvian economy is dependent on millions of small firms ( de Soto 2002 ), and in Peru, as in developed economies, banks play a key role in financing growth. The data are supplied by the Peruvian banking regulator Superintendencia de Banca, Seguros, y AFPs (SBS) and are labeled the RCD ( Reporte Crediticio de Deudores ) database. 3 For each Peruvian financial institution, the data describe the monthly loan balances of every business borrower. We draw from two data sets. Firms are assigned to a category and an associated data set, based on the amount of their borrowing. The first is the Micro-enterprise (MES) data set, which is designed to report loan balances for all firms with a total borrowing across the entire financial system of less than $20,000 (changed to $30,000 in 2003). We describe this cutoff as the COM threshold. The second is the COM data set for firms with a total loan balance above the MES threshold. In Section 1.1, we discuss in some detail the rules for assigning firms to either category and the implications of this assignment for firms and banks. We are primarily interested in firms in the MES data set, with a particular focus on those that transition to COM status. Eighteen million firm-bank-month observations are in this joint MES and COM database.

In addition to supplying loan balances, the data specify the currency in which each loan is denominated (either Peruvian Soles, denoted S/., or U.S. dollars). Over the term of the sample period, 77% of the loan balances of MES firms are in Soles, with this fraction increasing over time. By 2010, 90% of the loan balances of MES firms are in Soles. Much of our analysis will consider the amount of new financing received by a firm. The RCD database provides information on loan balances and does not identify new loans. We therefore adopt the classification rule that any exchange-rate-adjusted increase in the loan balance of more than 5% is treated as a new loan. For a firm that receives a new loan, we view the entire new balance as a new loan. As we show in the Online Appendix , our results are robust to using cutoffs of 20% and 50% to define new lending.

We also have geocoded location information for firms in 136 out of 195 provinces in Peru, including all major cities. Locations are provided in the form of eight-digit longitudes and latitudes and are precise to an accuracy of +/- 7.5 meters. This precision enables us to undertake a microgeographical analysis of the effects of a firm’s financing on the funding of its neighbors.

1.1 MES versus COM

The central distinction between the banking regulations applicable to MES and COM firms is that once firms enter the COM category, their lenders are required to collect formal financial statements from them. 4 That is, the transition to COM results in the provision of quantitative information to lenders. Why do banks not require financial statements from MES borrowers as well? Generating these documents is costly for small borrowers, and collecting and evaluating them is costly for banks.

Also, along with the formal requirement for the provision of financial statements, banks will often send representatives to meet with firms that transition to COM. These representatives are typically assigned responsibility for specific geographic areas. In some cases, these bank representatives will also canvass neighboring firms for their views of the company. In general, a shift to COM status leads to significantly more information gathering by the bank, a process that includes an intensified consideration of the firm’s local market conditions.

COM and MES firms also differ in certain other aspects of their regulatory regimes. Most importantly, delinquency is assessed differently for COM and MES firms, and banks are required to make different loss provisions for delinquent COM and MES firms. Our analysis will also consider the possible impact of these regulatory differences.

2. Empirical Specification

We are interested in the effect of an exogenous information and financing shock on the provision of loans to a firm and to its neighbor firms. Following the formal rules of the Peruvian banking regulation, firms with a total loan balance above the threshold should be subject to COM regulations. Banks and firms likely agree to the COM transition for unobserved reasons, so a regression of financing characteristics on an indicator for COM status likely would be subject to endogeneity concerns. The formal eligibility threshold can, however, be used in a regression discontinuity design to measure the causal impact of a transition to COM status. Banks are required to assign firms with total balances above the threshold to COM. The threshold level in Soles is unknown until the end of the period, at which point the official exchange rate is announced. The firm’s end of month balance in Soles also may be unknown, particularly if the firm has U.S. dollar debt. This suggests the presence of some noise in the assignment of firms to COM status. Nonetheless, a bank may observe a firm’s previous period balance and the current Sol per Dollar exchange rate R t to assess whether the firm is likely to exceed the eligibility threshold. Consider the set of firms with MES status in period t – 1. A firm i in month t with an exchange-rate-adjusted month t – 1 balance that exceeds the month t exchange-rate-adjusted cutoff should be likely assigned to COM:  

(1)
CommercialStatusi,t=α+β(ExchangerateadjustedBalancei,t>Cutofft)+ϵi,t=α+β(USDbalancei,t1*Rt+Solesbalancei,t1>(DollarCutofft)Rt)+ϵi,t,
where CommercialStatusi,t is an indicator variable for whether firm i is assigned to the COM database for the first time, Cutoff t is the month t Commercial cutoff measured in Soles, and ε εi,t is an error term. Equation (1) can be estimated by a local linear regression (Hahn, Todd, and Van der Klaauw 2001).

The bank may also use other unobserved variables to assign a firm to COM status, but we will not exploit this potentially endogenous information in our design. The bank may also have information about within-month balances that we cannot exploit given our end-of-month database. In this sense, Equation (1) describes a “fuzzy” regression discontinuity design, in which we are testing for a discontinuous jump in the probability of COM status assignment, but this jump need not be equal to one. In essence, we are testing if firms pushed across the eligibility threshold by exchange rate movements are substantially more likely given COM status.

We first consider whether a focal firm’s entry into COM has an impact on its financing. This suggests the following specification:  

(2)
FinancingOutcomei,t+12=γ+δ(ExchangerateadjustedBalancei,t>Cutofft)+νi,t,
where νi,t is an error term. We estimate Equation (2) using local linear regression techniques.

As described in Section 1.1, focal firms that enter COM status receive both an information and a regulatory shock that should lead to more financing. What effect might this have on neighbor firms? Benveniste, Busaba, and Wilhelm (2002) argue, in an investment banking context, that information externalities between firms in the same industry may be significant. In our setting, we are considering information externalities between firms in the same local area. More information about the focal firm transitioning to COM may supply the lending bank with more information about its neighbors as well. Almazan (2002) shows that as a bank’s costs of monitoring distant firms decline, the bank is able to finance projects further afield. In the context of these Peruvian firms, a focal firm’s switch to COM status may reduce the bank’s cost of monitoring its neighbors, and this should lead to more lending to these neighbors. Consistent with this, Hauswald and Marquez (2006) argue that lowering the cost of information should lead to a broader lending market for a bank. All these theories suggest the information shock to the focal firm should lead to more lending to its neighbors.

To test this hypothesis, we match each focal firm that potentially may be subject to a transition to the set of neighbor firms within 500 meters of its location that share a bank with the focal firm. The focal firm is designated the “local focal firm” for each of its neighbors. We consider a focal firm to be potentially subject to a transition if its exchange-rate-adjusted balance is within some window of the cutoff in period t . We then contrast financing outcomes T months in the future for neighbors of focal firms that cross the threshold with the corresponding outcomes for neighbors of focal firms that do not cross the threshold. Specifically, we estimate  

(3)
NeighborFirmFinancingOutcomei,t+T=ζ+λ(LocalFocalFirmExchangerateadjustedBalancei,t>Cutofft)+controlsi,t+ui,t,
where controlsi,t is a set of neighbor firm controls include location, time and industry fixed effects and ui,t is an error term. The coefficient of interest is λ , which, according to the theories described above, should be positive, reflecting the positive information spillover of the focal firm’s transition to COM on lending to its neighbor firms. The smallest administrative subdivision of Peru is a district, of which there are 1,834. Given Peru’s population of 29.4 million, this gives an average size of just over 16,000 people per district, or roughly twice the population of a typical U.S. ZIP code. We include fixed effects at the district-year-month interaction level to control for a rich set of time and location unobservables. We estimate Equation (3) using ordinary least squares (OLS), analyzing the differences between the neighbors of focal firms on opposing sides of the threshold, for samples in which the focal firms are all within tight windows of the cutoff.

In our main specification, we consider focal firms with exchange-rate-adjusted balances within 2,500 Soles of the threshold (during our sample period the Sol traded at an average of 3.3 per US dollar). Sample statistics are given in Table 1 for the set of 8,674 firm-month observations that fall within this window.

Table 1

Summary statistics

Variable Mean Median SD 1st pctile. 99th pctile. 
Existing debt 81,853 83,582 11,112 35,307 98,043 
New financing 122,895 92,745 128,829 538,844 
Firm classification 0.16 0.62 
Share of troubled debt 0.03 0.14 
Fraction of new financing prior 12 months 225 2.90 19,260 37 
Variable Mean Median SD 1st pctile. 99th pctile. 
Existing debt 81,853 83,582 11,112 35,307 98,043 
New financing 122,895 92,745 128,829 538,844 
Firm classification 0.16 0.62 
Share of troubled debt 0.03 0.14 
Fraction of new financing prior 12 months 225 2.90 19,260 37 

Summary statistics are based on 8,674 firm-month observations on focal firms, defined as firms with a monthly foreign exchange adjusted existing debt at t – 1 within 2,500 Soles of the threshold of month t . The foreign exchange adjustment is USD (US dollar) balance t-1 * (exchange rate in Soles/USD) t + Soles balance t-1 . Existing debt is in Soles and for month t . New financing, also expressed in Soles, sums over all new loans between months t + 1 and t + 12 for the focal firm. Firm classification is the worst classification (i.e., delinquency status, with more delinquent loans receiving a higher classification) of a loan received by the firm in the financial system by any bank in that month. The share of troubled debt over total debt is a ratio, where troubled debt is all past due, refinanced, restructured, and in judicial collection debt. The fraction of new financing in the prior twelve months is the ratio of all new financing received by the firm between month t – 11 and month t , divided by total debt as of month t – 12.

3. Results

3.1 Transition to commercial status

As described in Section 1.1.1 above, Peruvian firms are assigned to either MES or COM status, and this categorization is formally governed by the total outstanding loan balance, expressed in dollars, held by the firm in the financial system. We begin by analyzing the relationship between loan balances and COM/MES status. Does the loan balance threshold determine the firm classification in practice?

We adopt the approach described in Section 2 and analyze the effect on a firm’s COM status of exchange rate shocks to both the threshold and the firm’s balance in the previous month. This approach does not make use of within-month balance changes (which we do not observe) or other endogenous variables that may govern the bank’s decision to grant a firm COM designation, so we do not expect to perfectly predict this outcome. By exploiting the impact of currency shifts, however, we can contrast firms that fall on either side of the threshold for exogenous reasons.

3.1.1 Sample selection.

For firms to be included in our sample, they must meet two criteria. First, we only consider MES firms. Transitions over the COM threshold will have no impact on firms already classified as COM. Transition to COM is unidirectional; once a firm has become COM, it remains COM and its movements through the COM threshold have no effect on its status. Second, we only include MES firms with a t – 1 balance below the threshold. That is, we exclude MES firms already above the threshold, but that have not been classified as COM, for whatever reason. We exclude these MES firms because they have possibly not been classified as COM due to some unobserved characteristic, and this may lead to endogenous selection. By excluding this second group of firms, we are able to focus on MES firms that transition over the threshold only due to exchange-rate movements and for no other reason.

The distribution of the prior month ( t – 1) loan balances for firms in the sample is depicted in Figure 1 . The figure illustrates two features of the data. First, as described, all the firms in the sample have prior-month loan balances below the COM threshold. Second, a clear peak in the data is observed just below the threshold, and a significant mass of firms is quite close to the threshold. While these firms may be somewhat distinctive in this regard, the key point of our empirical approach is to always make comparisons across firms all within this general grouping; we compare firms close to the threshold that are pushed over the threshold by the exchange rate with other firms close to the threshold, but for which exchange rate movements leave them just short of the threshold. That is, we exploit quasi-random variation within the general group of firms with balances close the threshold.

Figure 1

Density of the prior-month loan balance distance to threshold This figure shows the density of the prior-month loan balances close to the COM threshold, which is denoted zero.

Figure 1

Density of the prior-month loan balance distance to threshold This figure shows the density of the prior-month loan balances close to the COM threshold, which is denoted zero.

3.1.2 Are differences between above- and below-threshold firms quasi-random?

We exploit the exchange rate variation that pushes some of the firms above the threshold and leaves others below it. Are differences between above- and below-threshold firms indeed quasi random? This cannot be proven incontrovertibly, but three arguments suggest quasi randomness likely would occur. The first argument is that exchange rate changes are hard to forecast and exogenous from the perspective of any given firm, so it seems quite likely that this introduces an element of random noise into the exchange-rate-adjusted loan balance (Lee 2008). The second point is to consider the distribution of exchange-rate-adjusted loan balances. A significant discontinuity in this distribution at the threshold showing, for example, substantially more firms below the threshold than above, might indicate that the exchange-rated-adjusted loan balances are being manipulated. Figure 2 shows the exchange-rate-adjusted loan balance (with the threshold normalized to zero) for loans within 2,500 Soles of the boundary. As the figure makes clear, no significant discontinuity exists at zero. A formal McCrary test comparing the relative log heights of the estimated probability densities at zero yields a coefficient of -0.024 and a t -statistic of -0.35. No evidence of a jump in the frequency of firms with exchange-rate-adjusted balances just above or below the cutoffs exists. Banks and firms may purposely choose initial loan balances above or below the threshold (as suggested by Figure 1 ), but Figure 2 demonstrates that exchange rate movements generate enough local noise to ensure that exchange-rate-adjusted balances are quasi randomly distributed around the cutoff. For this reason, our analysis focuses on exchange-rate-adjusted balances rather than on the balances themselves.

Figure 2

Density of the exchange-rate-adjusted loan balance distance to threshold This figure shows the estimated density (and accompanying confidence interval) for the exchange-rate-adjusted loan balances, where zero denotes the COM threshold.

Figure 2

Density of the exchange-rate-adjusted loan balance distance to threshold This figure shows the estimated density (and accompanying confidence interval) for the exchange-rate-adjusted loan balances, where zero denotes the COM threshold.

As a third test, we analyze the distribution of observable firm characteristics around the threshold. We present results for three variables. The first is the worst classification of any loan held by the firm; Peruvian banking regulations mandate that all financial institutions report on the delinquency status of each loan, on a five-point scale from normal (a score of 0) to loss (a score of 4). The second characteristic is the share of troubled debt, defined to be any debt with a classification of below normal, over total debt. The third is the ratio of all new financing received by the firm between month t – 11 and month t , divided by total debt as of month t – 12. As shown in Figure 3 , none of these variables exhibits a discontinuity at the cutoff. That is, above- and below-threshold firms have quite similar loan classifications, fractions of troubled loans, and ratios of new financing.

Figure 3

Comparing observable characteristics around the threshold

Each panel displays the kernel densities of below-threshold firms (solid line) and above-threshold firms (dashed line) for a given observable characteristic contemporaneous or prior to the month in which these focal firms enter the range within 1,500 Soles of the MES threshold. Only observations within the narrow window of 1,500 Soles around the MES threshold are used for the comparison. The observable characteristics are the firm’s loan classification, defined as the worst classification in the financial system by any bank in that month ( A ), the share of troubled debt over total debt, where troubled debt is all past due, refinanced, restructured, and in judicial collection debt ( B ), and the ratio of all new financing received by the firm between month t – 11 and month t , divided by total debt as of month t – 12 ( C ). The p -values of Kolmogorov-Smirnov tests of equality of densities are reported at the bottom of each graph.

Figure 3

Comparing observable characteristics around the threshold

Each panel displays the kernel densities of below-threshold firms (solid line) and above-threshold firms (dashed line) for a given observable characteristic contemporaneous or prior to the month in which these focal firms enter the range within 1,500 Soles of the MES threshold. Only observations within the narrow window of 1,500 Soles around the MES threshold are used for the comparison. The observable characteristics are the firm’s loan classification, defined as the worst classification in the financial system by any bank in that month ( A ), the share of troubled debt over total debt, where troubled debt is all past due, refinanced, restructured, and in judicial collection debt ( B ), and the ratio of all new financing received by the firm between month t – 11 and month t , divided by total debt as of month t – 12 ( C ). The p -values of Kolmogorov-Smirnov tests of equality of densities are reported at the bottom of each graph.

3.1.3 Does transition across the COM threshold lead to COM classification?

Next, we consider whether exchange rate shocks that push MES firms with Sol loans across the COM eligibility threshold actually cause these firms to be given COM status. We address this issue by estimating Equation (1) regressing COM status on the exchange-rate-adjusted balance. The results from this regression are shown in the first column of Table 2 . The local linear estimator shows a discrete jump of 11.7 percentage points ( t -statistic=18.4) in the probability of COM status precisely at the formal eligibility threshold. This result demonstrates that even when a firm is pushed across the threshold by an exogenous exchange rate shock, the formal cutoffs continue to have a substantial effect on classification. This is a fuzzy regression discontinuity design in which we do not observe all the information the banks uses to classify firms. Nonetheless, it is clear that exchange-rate-driven shocks have a strong impact in pushing some firms across the threshold and leading them to enter COM status, while other, very similar firms remain below the threshold and maintain their MES classification.

Table 2

Exchange-rate-adjusted distance to threshold and transitioning to commercial status

Dependent variable: Transitioned to commercial status (1/0)  
Estimation: RD OLS 
Focal window width (Soles):   2.5 K  2.0 K  1.5 K  
 (2.1) (2.2) (2.3) (2.4) (2.5) 
Above threshold 0.117*** 0.102*** 0.097*** 0.082*** 0.111*** 
 (18.43) (15.82) (13.41) (9.65) (11.21) 
Polynomials No No No No 
R 2  0.03 0.02 0.02 0.01 
Sample size 23M 8,674 7,031 5,261 23M 
Number of clusters (firms)     775,105 
Dependent variable: Transitioned to commercial status (1/0)  
Estimation: RD OLS 
Focal window width (Soles):   2.5 K  2.0 K  1.5 K  
 (2.1) (2.2) (2.3) (2.4) (2.5) 
Above threshold 0.117*** 0.102*** 0.097*** 0.082*** 0.111*** 
 (18.43) (15.82) (13.41) (9.65) (11.21) 
Polynomials No No No No 
R 2  0.03 0.02 0.02 0.01 
Sample size 23M 8,674 7,031 5,261 23M 
Number of clusters (firms)     775,105 

Using different estimation techniques and different samples, this table reports the discontinuous impact of total exchange-rate-adjusted debt on whether a firm receives its first commercial loan in month t . The baseline sample is all MES firms that have no commercial loans in month t – 1 and for which total debt balance is below the COM threshold in month t – 1. The assignment variable for all models is defined using the following foreign exchange adjustment: USD (US Dollar) balance t-1 * (exchange rate in Soles/USD) t + Soles balance t-1 . This assignment variable and the threshold are scaled by 100,000 (i.e., 0.025 units = 2,500 Soles). The threshold is for month t and is expressed in Soles. The second, third, and fourth models are estimated using a narrow window of the assignment variable. Above threshold is a dummy for when the assignment variable is greater than or equal to the threshold. *, **, and *** indicate significance at the 10%, 5% and 1% level, respectively. Standard errors are heteroscedasticity robust and clustered when indicated t -statistics shown in parentheses.

Columns 2-4 of Table 2 display results from estimating Equation (1) in which we use OLS to regress COM status on an indicator for an above-threshold exchange-rate-adjusted balance in varying narrow windows around zero. The results are shown for windows of +/-2,500, 2,000 and 1,500 Soles, in Columns 2-4, respectively. To give a sense of magnitudes, the mean of the above-threshold indicator during the sample period was 0.04 for the whole database; in the narrow window of 2,500 Soles, the mean of the above-threshold indicator was 0.38. These results only make use of firms very close to the threshold to estimate the discontinuity. The results are somewhat smaller than for the local linear estimator, with coefficients ranging from 8.2 to 10.2 percentage points, and the estimated coefficients are significant (the t -statistics range from 9.65 to 15.82). Column 5 of Table 2 shows the similar result from an OLS polynomial specification, with the polynomial of order seven. This result is graphically depicted in Figure 4 .

Figure 4

Exchange-rate-adjusted loan balance distance to threshold and probability of transition to commercial status

Each scattered dot represents a bin of the distance to debt threshold variable. The dashed lines represent 95% confidence intervals of seventh-degree polynomial fits of whether the firm received a commercial loan on the exchange-rate-adjusted distance to debt threshold values, as in the fifth model of Table 2 .

Figure 4

Exchange-rate-adjusted loan balance distance to threshold and probability of transition to commercial status

Each scattered dot represents a bin of the distance to debt threshold variable. The dashed lines represent 95% confidence intervals of seventh-degree polynomial fits of whether the firm received a commercial loan on the exchange-rate-adjusted distance to debt threshold values, as in the fifth model of Table 2 .

3.2 Focal firm financing

The analysis in Table 2 establishes that firms pushed across the COM threshold by exchange rate movements are indeed significantly more likely to be granted COM status. We now consider the impact of COM status on the amount of financing received by a firm. Specifically, we estimate Equation (2) and regress the log of the new financing received by the firm in the next year on an indicator for whether a firm has an exchange-rate-adjusted balance above the classification threshold. The local linear estimator using the optimal bandwidth of Imbens and Kalyanaraman (2012) yields a coefficient of 0.57 ( t -statistic=4.83) on the above threshold indicator, as displayed in the first row of the first column Table 3 . This indicates that firms that achieve COM status due to exchange rate movements receive significantly more new financing in the following year than otherwise very similar firms for which exchange-rate-adjusted balances fall just below the COM threshold. The results displayed in the second and third columns show that this positive effect is robust to the choice of other bandwidths. In other words, COM status appears to have a large causal effect on subsequent financing. In Columns 4-6 of Table 3 we show the results from regressing an indictor for new loan financing on whether the firm’s exchange-rate-adjusted balance exceeded the COM limit. These results consistently establish that above-threshold firms are more likely to receive new financing. The estimated coefficients on the above-threshold indicator range from 0.039 to 0.058 and the t -statistics range from 3.57 to 8.19.

Table 3

Financing shocks and focal firm financing in local regression models

 Dependent variable: 
Log of 12-month new financing New loan indicator 
(3.1) (3.2) (3.3) (3.4) (3.5) (3.6) 
Above threshold 0.572*** 0.323** 0.902*** 0.058*** 0.039*** 0.057*** 
 (4.83) (2.06) (9.9) (7.00) (3.57) (8.19) 
Multiple of optimal bandwidth 100% 50% 200% 100% 50% 200% 
 Dependent variable: 
Log of 12-month new financing New loan indicator 
(3.1) (3.2) (3.3) (3.4) (3.5) (3.6) 
Above threshold 0.572*** 0.323** 0.902*** 0.058*** 0.039*** 0.057*** 
 (4.83) (2.06) (9.9) (7.00) (3.57) (8.19) 
Multiple of optimal bandwidth 100% 50% 200% 100% 50% 200% 

Each entry is from a different model analyzing the impact of an above-threshold exchange-rate-adjusted loan balance on focal firm financing. Each local regression model fits nonparametric local linear regressions using the optimal bandwidth of Imbens and Kalyanaraman (2012) with the rectangular kernel, calculating the model at different multiples (i.e., 100%, 50%, and 200%) of this optimal bandwidth. Observations for each regression are at the firm-month level for all firm-month combinations that have no history of commercial loans up to month t – 1 and for which total debt balance does not cross the MES threshold in month t . Amounts of new financing are in logs of one plus the new financing amount.Twelve-month values include all months between +1 and +12. The new loan indicator is a dummy for whether the amount of new financing in the months between +1 and +12 is greater than zero. *, **, and *** indicate significance at the 10%, 5% and 1% level, respectively. t -statistics shown in parentheses.

Why do focal firms that transition to COM status receive more financing? One argument is that firms exogenously pushed by exchange-rate movements across the threshold into COM status receive an information shock as their banks shift to the intensive information gathering processes required for COM firms. As a result, information asymmetries are reduced, and more financing is supplied. However, a shift to COM status results in more than a pure information shock. Presumably, significant costs are absorbed by both firms and banks in establishing COM review procedures, and this is why small firms are much more likely to shift to COM status only when required to do so by SBS regulations.

Suppose, for example, that COM status did not result in any additional information for the bank, but simply led to greater compliance costs. Even if this were true, we still might expect COM firms to receive greater future financing because they have already crossed the regulatory threshold, so the compliance costs already have been paid. For MES firms, an increase in future financing may be unattractive to the bank because it may lead to a costly transition and increase in compliance costs. For COM firms, these costs are already sunk. In other words, MES firms may face a bank reluctant to extend credit that would lead to the firms breaching the COM threshold. This may place a ceiling on future lending to MES firms not present for COM firms. 5

It is thus possible that any observed differences in future financing between focal firms that cross the cutoff for exchange rate reasons and those that do not may be driven by regulatory or informational considerations, or combination of both. The evidence is clear, however, that exchange-rate-driven transitions to COM generate a financial shock, resulting in more lending to focal firms.

3.3 Financial Neighbor Firm Financing

What should we expect to be the impact of a focal firm’s targeted regulatory information and financing shock on other local firms? Two arguments suggest that the focal firm’s shock will lead to more lending to the firm’s neighbors. First, due to the provision of financing, the focal firm may experience greater growth, which can generate positive economic spillovers for neighbors that may respond to these opportunities by borrowing more. Second, neighbors that borrow from the same bank may also benefit as the bank learns more about the focal firm and its local area. This improved information for the bank may lead to greater lending to all local firms ( Stiglitz and Weiss 1981 ). Conversely, two mechanisms may lead to less borrowing by neighbor firms. First, if a bank imposes local capital budgeting limits ( Zhang 1997 ), perhaps for diversification reasons or due to agency considerations ( Scharfstein and Stein 2000 ; Ozbas and Scharfstein 2010 ), then its increased lending to the focal firm may siphon away financing from neighbors. Second, if the focal firm competes with some neighbors, this may also lead them to receive less financing as the focal firm is strengthened by its financing shock. Understanding the local impact of this shock is thus clearly a question for empirical analysis.

To examine the impact of the shock on neighboring firms, we identify for each focal firm all the MES firms sharing the same bank within 500 meters of its location. We label these firms the financial neighbors of the focal firm. We then contrast the outcomes for financial neighbors of focal firms for which exchange-rate-adjusted balances fall just above and just below the COM cutoff. Neighbors of above threshold focal firms are significantly more likely to be subject to a local financial shock, as above threshold focal firms are more likely to achieve COM status.

To assess the impact of the financial shock on the neighboring firms, we estimate Equation (3) , using a window of [2,500S/.,+2,500S/.] . For each financial neighbor firm, we regress the log of new financing in the subsequent twelve months on an indicator for whether its associated focal firm has an exchange-rate-adjusted balance above the cutoff, age fixed effects, industry fixed effects, bank fixed effects, a control for the number of local neighbor firms (quartile fixed effects), and district-year-month interaction fixed effects. We report robust standard errors clustered by province to allow for local correlations. The result, detailed in the first column of Table 4 , is that the neighbors of above-threshold firms receive 1.4% more financing ( t -statistic=4.70). A local information and financial shock due to the transition of a focal firm to COM status results in more financing for its financial neighbor firms. This is consistent with the presence of information spillovers as described by Benveniste, Busaba, and Wilhelm (2002) and the positive effects of information on lending detailed by Almazan (2002) and Hauswald and Marquez (2006) .

Table 4

Financing shocks and financial neighbor firms’ financing

 Dependent variables: 
Log of 12-month new financing New loan indicator 
Focal window width (Soles):  2.5 K  2.0 K  1.5 K  2.5 K  2.0 K  1.5 K 
 (4.1) (4.2) (4.3) (4.4) (4.5) (4.6) 
Above threshold 0.014*** 0.017** 0.021*** 0.001*** 0.002*** 0.002*** 
 (4.70) (2.04) (3.23) (6.72) (3.04) (4.52) 
Controlling for existing debt Yes Yes Yes Yes Yes Yes 
Age fixed effects Yes Yes Yes Yes Yes Yes 
Industry fixed effects Yes Yes Yes Yes Yes Yes 
Bank fixed effects Yes Yes Yes Yes Yes Yes 
Neighbor firms quartile fixed effects Yes Yes Yes Yes Yes Yes 
District × Year-month fixed effects Yes Yes Yes Yes Yes Yes 
R2 0.05 0.05 0.05 0.03 0.03 0.03 
Sample size 2.6M 2.1M 1.6M 2.6M 2.1M 1.6M 
Number of clusters (provinces) 120 116 113 120 116 113 
 Dependent variables: 
Log of 12-month new financing New loan indicator 
Focal window width (Soles):  2.5 K  2.0 K  1.5 K  2.5 K  2.0 K  1.5 K 
 (4.1) (4.2) (4.3) (4.4) (4.5) (4.6) 
Above threshold 0.014*** 0.017** 0.021*** 0.001*** 0.002*** 0.002*** 
 (4.70) (2.04) (3.23) (6.72) (3.04) (4.52) 
Controlling for existing debt Yes Yes Yes Yes Yes Yes 
Age fixed effects Yes Yes Yes Yes Yes Yes 
Industry fixed effects Yes Yes Yes Yes Yes Yes 
Bank fixed effects Yes Yes Yes Yes Yes Yes 
Neighbor firms quartile fixed effects Yes Yes Yes Yes Yes Yes 
District × Year-month fixed effects Yes Yes Yes Yes Yes Yes 
R2 0.05 0.05 0.05 0.03 0.03 0.03 
Sample size 2.6M 2.1M 1.6M 2.6M 2.1M 1.6M 
Number of clusters (provinces) 120 116 113 120 116 113 

Observations are at the focal firm / neighbor firm / bank / month level for all financial neighbors of all focal firms within 0.5 kilometers of the focal firm and that share a bank with the focal firm. Above threshold is defined for focal firms, as in Table 2 , and its value is imputed to neighboring firms to explain these neighboring firms’ financing in the twelve months following the month in which the focal firm’s exchange-rate adjusted balance is within 2,500 Soles (2.5 K ) or 2,000 Soles (2.0 K ) or 1,500 Soles (1.5 K ) of the MES threshold. Amounts of new financing are in logs of one plus the new financing amount. Twelve-month values include all months between +1 and +12. The new loan indicator is a dummy for whether the amount of new financing in the months between +1 and +12 is greater than zero. *, **, and *** indicate significance at the 10%, 5% and 1% level, respectively. t -statistics based on robust standard errors clustered by province are shown.

The results are not dependent on the specific window used with respect to the threshold. As shown in the second and third columns of Table 4 , the finding that a shock for the focal firm leads to more financing for its neighbors is robust across a number of specifications. In Columns 4-6 of Table 4 we show that the indicator for new loan financing is also significantly higher for neighbors of focal firms that transition to COM, regardless of the particular window: estimated coefficients on the above threshold indicator range from 0.001 to 0.002, and the associated t -statistics have a minimum of 3.04 and a maximum of 6.72. Given this general robustness, much of our subsequent analysis will focus on the neighbors of focal firms with exchange-rate-adjusted balances in the window of [2,500S/.,+2,500S/.] As we show in the Online Appendix Tables A1 and A2, the results are robust to using new loan definition cutoffs of 20% or 50%, rather than 5% as in our main approach. The results also hold in specifications that include focal firm fixed effects and cluster by focal firm, as well as in regressions that pool all the financings of neighbor firms for each focal firm ( Online Appendix Table A3).

3.3.1 Financing shocks and geographic distance: Very close versus farther locations.

The magnitude and significance of spillovers may vary depending on the geographic distance between the focal firm and its financial neighbors. So far, the analysis has been conducted by employing data on all financial neighbors within 500 meters of the focal firm. While this boundary appears natural in our empirical context (e.g., in the case of Lima, it amounts to about four blocks), using other geographic distances may help illustrate the informational nature of the financing shock.

The first column of Table 5 reduces the boundary for distance by 90% to focus only on financial neighbors within 50 meters of the focal firm. The coefficient on the above threshold indicator is 0.066 ( t -statistic=4.50), consistent with an economically and statistically large effect of the shock on close financial neighbors. Moreover, the second column of Table 5 further cuts this short distance by an additional 80%, to keep in the sample only those financial neighbors within 10 meters of the focal firm, and we again find a positive and significant coefficient of 0.063 ( t -statistic=3.95) for very close neighbors. The shock therefore appears to be stronger at closer locations, consistent with the hypothesis that information diffuses more strongly and rapidly within short distances.

Table 5

Financing of neighbors at very close versus farther geographic locations

 Dependent variable: Log of 12-month new financing  
Neighbor location:  50 meters   10 meters  500m<x1000m 
 (5.1) (5.2) (5.3) 
Above threshold 0.066*** 0.063*** 0.012** 
 (4.50) (3.95) (1.99) 
Controlling for existing debt Yes Yes Yes 
Age fixed effects Yes Yes Yes 
Industry fixed effects Yes Yes Yes 
Bank fixed effects Yes Yes Yes 
Neighbor firms quartile fixed effects Yes Yes Yes 
District × Year-month fixed effects Yes Yes Yes 
R2 0.07 0.08 0.07 
Sample size 190,241 146,439 2.3M 
Number of clusters (provinces) 120 115 64 
 Dependent variable: Log of 12-month new financing  
Neighbor location:  50 meters   10 meters  500m<x1000m 
 (5.1) (5.2) (5.3) 
Above threshold 0.066*** 0.063*** 0.012** 
 (4.50) (3.95) (1.99) 
Controlling for existing debt Yes Yes Yes 
Age fixed effects Yes Yes Yes 
Industry fixed effects Yes Yes Yes 
Bank fixed effects Yes Yes Yes 
Neighbor firms quartile fixed effects Yes Yes Yes 
District × Year-month fixed effects Yes Yes Yes 
R2 0.07 0.08 0.07 
Sample size 190,241 146,439 2.3M 
Number of clusters (provinces) 120 115 64 

Observations are at the focal firm / neighbor firm / bank / month level for all financial neighbors of the focal firm that share a bank with the focal firm and are located within specified distances from the focal firm: within 50 meters, within 10 meters, or between 500 and 1,000 meters. Above threshold is defined for focal firms, as in Table 2 , and its value is imputed to neighboring firms to explain these neighboring firms’ financing in the twelve months following the month in which the focal firm’s exchange-rate adjusted balance is within 2,500 Soles of the MES threshold. *, **, and *** indicate significance at the 10%, 5% and 1% level, respectively. t -statistics based on robust standard errors clustered by province are shown.

By contrast, the third column of Table 5 reports a regression that goes beyond the 500-meter mark to focus exclusively on those financial neighbors that are relatively far from the focal firm. Specifically, only financial neighbors located between 500 meters and 1,000 meters from the focal firm are considered for the test. The coefficient of interest is -0.012 and has a t -statistic of -1.99. This suggests that the nature of the positive spillovers we are analyzing is very local and that there even may be negative spillovers at a greater distance, perhaps due to the capital budgeting constraints or competition rationales discussed earlier.

The findings in Tables 4 and 5 are consistent with both the argument that financing generates positive economic spillovers in the narrow local area and with the claim that increased information about one firm leads to more information about its near neighbors. In our subsequent analysis, we provide further evidence on these two possible mechanisms.

3.3.2 Financing shocks, financial neighbors’ financing, and geographic neighbors’ financing.

The analysis in Table 4 focuses on the impact of a focal firm’s exchange-rate-induced transition to COM on the financing of neighbor firms located within 500 meters that share a bank with the focal firm. If the additional financing received by a transitioning focal firm leads to more financing for its neighbors because of general economic spillovers, then we should expect to see all local firms benefitting from these spillovers, not just those that share a bank. In this section we consider whether the spillovers documented in Table 4 are confined solely to neighboring firms with which it shares a bank (financial neighbors) or whether all local firms within 500 meters (geographic neighbors) benefit.

For each geographic neighbor firm, we regress the log of the new financing received by the neighbor in the next twelve months on an indicator for whether the associated focal firm was pushed across the COM boundary, an indicator for whether the neighbor and focal firm share a bank, the interaction between these two variables and the standard controls. The result, shown in the first column of Table 6 , is that the coefficient on the interaction is 0.023 ( t -statistic=4.83). This shows that financial neighbors that share a bank receive more financing. The coefficient on the above threshold-indicator alone is -0.009 ( t -statistic=-1.87), which indicates that geographic neighbors of transitioning focal firms do not receive any additional financing. This conclusion is consistent across other narrow windows of exchange-rate-adjusted balances: the evidence shows that geographic neighbors do not receive more new loans. This suggests that a pure financial shock to a focal firm does not generate a broad economic spillover benefit for all local firms. The findings in Tables 4 and 6 show that only local firms that share a bank with the focal firm are affected by the spillover.

Table 6

Financing shocks, financial neighbors’ financing, and geographic neighbors’ financing

 Dependent variable: Log of 12-month new financing  
Focal window width (Soles):  2.5 K  2.0 K  1.5 K 
 (6.1) (6.2) (6.3) 
Above threshold × Shares a bank 0.023*** 0.016*** 0.013* 
 (4.83) (2.71) (1.65) 
Above threshold 0.009* 0.002 0.004 
 (1.87) (0.50) (0.70) 
Shares a bank 0.032* 0.032** 0.023 
 (1.83) (2.10) (1.52) 
Controlling for existing debt Yes Yes Yes 
Age fixed effects Yes Yes Yes 
Industry fixed effects Yes Yes Yes 
Bank fixed effects Yes Yes Yes 
Neighbor firms quartile fixed effects Yes Yes Yes 
District × Year-month fixed effects Yes Yes Yes 
R2 0.08 0.08 0.08 
Sample size 6.1M 5M 3.8M 
Number of clusters (provinces) 120 117 114 
 Dependent variable: Log of 12-month new financing  
Focal window width (Soles):  2.5 K  2.0 K  1.5 K 
 (6.1) (6.2) (6.3) 
Above threshold × Shares a bank 0.023*** 0.016*** 0.013* 
 (4.83) (2.71) (1.65) 
Above threshold 0.009* 0.002 0.004 
 (1.87) (0.50) (0.70) 
Shares a bank 0.032* 0.032** 0.023 
 (1.83) (2.10) (1.52) 
Controlling for existing debt Yes Yes Yes 
Age fixed effects Yes Yes Yes 
Industry fixed effects Yes Yes Yes 
Bank fixed effects Yes Yes Yes 
Neighbor firms quartile fixed effects Yes Yes Yes 
District × Year-month fixed effects Yes Yes Yes 
R2 0.08 0.08 0.08 
Sample size 6.1M 5M 3.8M 
Number of clusters (provinces) 120 117 114 

Observations are at the focal firm / neighbor firm / bank / month level for all neighbors of all focal firms within 0.5 kilometers of the focal firm, regardless of whether or not they share a bank with the focal firm. Above threshold is defined for focal firms, as in Table 2 , and its value is imputed to neighboring firms to explain these neighboring firms’ financing in the twelve months following the month when the focal firm’s exchange-rate adjusted balance is within 2,500 Soles (2.5 K ) or 2,000 Soles (2.0 K ) or 1,500 Soles (1.5 K ) of the MES threshold. Shares a bank is a dummy equal to one when the neighbor shares a bank with the focal firm. Above threshold × Shares a bank is an interaction of these variables. *, **, and *** indicate significance at the 10%, 5% and 1% level, respectively. t -statistics based on robust standard errors clustered by province are shown.

Overall, these findings are consistent with the hypothesis that after a firm becomes COM, its bank receives more information about both the firm and the local area and is therefore able to expand its financing to neighboring companies. The evidence presented to this point, however, has mainly served to weigh against alterative mechanisms. In the next section we consider a more direct test of the information hypothesis.

3.3.3 Financing shocks, financial neighbors’ new loans, and removal from tax system.

One way to assess whether banks are more informed about the neighbors of firms that experience an exchange-rate-generated transition to COM is to evaluate the success of their lending to these neighbors. Specifically, we consider whether banks are more or less likely to make loans to neighbors that subsequently experience business failure. Presumably, more informed banks should be less likely to make such loans. For this analysis, we are only interested in loans to financial neighbors, other firms in the local area that share a bank with the focal firm. Other banks unrelated to the focal firm are unlikely to become more informed by its transition to COM.

Our measure for failure is whether the neighbor firm was removed from the database of the Peruvian tax authority SUNAT. Firms removed from the tax system are presumed to be no longer actively engaged in any business activity (e.g., they are not issuing receipts). Are the neighbors of above-threshold focal firms that receive new loans less likely to go out of business? To evaluate this question, we regress an indicator for whether a financial neighbor firm is removed from the tax system over the next twelve months on an indicator for an above-threshold associated focal firm, an indicator for whether the neighbor received a new loan, the interaction between these variables and the usual controls. The coefficient of interest is the one on the interaction term. It describes whether the new loans made to the neighbors of above-threshold focal firms are more or less likely to subsequently exit the tax system.

As shown in the first column of Table 7 , we estimate this coefficient to be -0.008 ( t -statistic=-6.03). This indicates that financial neighbors of above-threshold firms that receive new loans are subsequently significantly less likely to end operations. This is evidence in favor of the argument that banks have more information about these neighbors and are able to identify better prospects for financing.

Table 7

Financing shocks, financial neighbors’ new loans, and removal from tax system

 Dependent variable: Firm is removed from the tax system (1/0) over the next...  
 12 months 24 months 36 months 
 (7.1) (7.2) (7.3) 
Above threshold × Received new loan 0.008*** 0.015*** 0.011*** 
 (6.03) (7.07) (4.55) 
Above threshold 0.006*** 0.012*** 0.011*** 
 (3.72) (5.82) (5.26) 
Received new loan 0.030*** 0.044*** 0.044*** 
 (6.24) (5.74) (6.09) 
Controlling for existing debt Yes Yes Yes 
Age fixed effects Yes Yes Yes 
Industry fixed effects Yes Yes Yes 
Neighbor firms quartile fixed effects Yes Yes Yes 
District × Year-month fixed effects Yes Yes Yes 
R2 0.02 0.03 0.04 
Sample size 1M 1M 1M 
Number of clusters (provinces) 120 120 120 
 Dependent variable: Firm is removed from the tax system (1/0) over the next...  
 12 months 24 months 36 months 
 (7.1) (7.2) (7.3) 
Above threshold × Received new loan 0.008*** 0.015*** 0.011*** 
 (6.03) (7.07) (4.55) 
Above threshold 0.006*** 0.012*** 0.011*** 
 (3.72) (5.82) (5.26) 
Received new loan 0.030*** 0.044*** 0.044*** 
 (6.24) (5.74) (6.09) 
Controlling for existing debt Yes Yes Yes 
Age fixed effects Yes Yes Yes 
Industry fixed effects Yes Yes Yes 
Neighbor firms quartile fixed effects Yes Yes Yes 
District × Year-month fixed effects Yes Yes Yes 
R2 0.02 0.03 0.04 
Sample size 1M 1M 1M 
Number of clusters (provinces) 120 120 120 

Observations are at the focal firm / neighbor firm / month level for all financial neighbors of all focal firms within 0.5 kilometers of the focal firm and that share a bank with the focal firm. The dependent variable is a dummy for whether the neighbor firm was removed from Peru’s tax system within the next 12 months, 24 months, or 36 months. Above threshold is defined for focal firms, as in Table 2 , and its value is imputed to neighboring firms to explain these neighboring firms’ reception of a new loan in the 12 months following the month when the focal firm’s exchange-rate adjusted balance is within 2,500 Soles of the MES threshold. This new loan reception dummy is introduced both in levels and as an interaction with the above threshold variable. *, **, and *** indicate significance at the 10%, 5% and 1% level, respectively. t -statistics based on robust standard errors clustered by province are shown.

This regression is not meant to be interpreted in a causal manner. Indeed, we are making use of the fact that granting a new loan is an endogenous decision by the bank. We are interested in whether the bank exhibits an enhanced ability to selectively make loans to better financial neighbor firms after the transition of the focal firm to COM status. The negative coefficient we estimate on the interaction shows that the correlation between making a new loan to a neighbor and the neighbor’s firm performance is indeed tighter (more negative) after an associated focal firm becomes COM. A similar pattern can be observed in Table 7 Columns 2 and 3, which consider exit within longer horizons of 24 and 36 months. The results are consistent with the argument that the bank knows more about financial neighbor firms after the focal firm is pushed across the COM threshold by exchange rate movements. Banks can better distinguish between the high- and low-quality neighbors of above-threshold focal firms, indicating that the bank does indeed have more information about these neighbors.

3.4 Externalities and Regulatory Implications

The results we find in Tables 4, 6, and 7 suggest that the transition of a firm from MES to COM results in significant information spillovers that lead to more lending to financial neighbors of focal firms. Given these benefits, a natural question is whether it makes sense for regulators to require formal financial statements from more firms. In fact, given the additional neighborhood lending that banks are able to initiate after a transition, one might wonder why banks do not choose themselves to require this information from firms, even without any regulatory impetus. 6

We have argued in Section 1.1 that generating formal financial statements is likely costly for borrowers and that it also may be expensive for banks to evaluate these statements. From the small borrower’s perspective, especially, transforming from an informal accounting system to a formal audit is likely cumbersome. In recognition of these costs, regulators in many countries offer audit exemptions for small firms. In the United Kingdom in 2004, for example, the financial thresholds for audit exemption were a turnover of £5.6m and a balance sheet total of £2.8m. The MES firms in our sample with total borrowing of less than $30,000 are very small relative to these limits. Moreover, given a choice, smaller firms in the United Kingdom are much less likely to voluntarily choose to be audited ( Collis, 2010 ). The average audit cost in the United Kingdom for small private businesses was estimated to be around £1,000. Even if this might be quite a bit lower in Peru, it would still be a very large cost for small Peruvian firms.

If the cost to MES firms of preparing formal financial statements is substantial, then one would expect to observe two phenomena. First, there should be an explosion in new lending for firms that breach the COM threshold. Once these firms are above the threshold, the cost of preparing the formal statements is now sunk, and there is no need to restrain borrowing to avoid breaching the COM limit. Indeed, this is precisely what we find in Table 3 . Second, one would expect to see MES firms clustering relatively close to the COM threshold, as these firms borrow as much as they can without going above the threshold. This is clearly documented in Figure 1 , which shows a large number of MES firms with borrowing just below the limit. These two findings are consistent with the argument that MES firms are reluctant to incur the costs of preparing formal statements.

With respect to the question of optimal regulatory requirements for formal financial statements, the Peruvian COM threshold already seems quite low from an international perspective. Lowering the limit and requiring even more firms to produce these statements could discourage borrowing by small firms and remove them from the banking system entirely. The neighboring firms that we show to benefit from the information spillover could choose to simply cease borrowing if the COM limit was lowered. We do not have enough information on the costs of producing formal financial statements to make regulatory recommendations, but reducing the COM threshold clearly could have significant negative effects on credit market participation that may well counterbalance the information benefits we document.

Should banks require more firms to submit formal financial statements? A similar argument applies here that small borrowers may dislike this requirement and may choose different lenders if their banks initiate new demands for audited statements. If the benefits to banks are big enough, could they not offer incentives to small firms to provide formal statements, thereby sharing their gains? While the transition to COM does lead to more lending to neighbors, the effect is not enormous in magnitude, and we cannot estimate the extent of the bank’s ability to capture the gains. Nonetheless, it is consistent with our findings to suggest that these benefits to banks are likely greatest when a large number of small firms in the near vicinity to a focal firm are close to the COM threshold. In cases of this type bank incentives to encourage formal financial disclosures by the focal firm should be most carefully considered.

4. Conclusion

We have studied the impact of information and financing shocks to firms on the supply of bank loans to their neighbor companies in a sample of Peruvian businesses in the period 2001-2010. Banks of firms with total loan balances above a certain U.S. dollar threshold were required by banking regulation to collect formal financial statements from their clients; these firms were designated to have a Commercial status and were governed by different regulations. Exploiting currency movements and implementing a regression discontinuity analysis, we have contrasted outcomes for firms with exchange-rate-adjusted balances just above and below the threshold. We have labeled the companies close to the threshold as focal firms, and we have found that focal firms pushed into Commercial status by exchange rate movements subsequently receive substantially more financing.

Further, we have found that the very close financial neighbors that share a bank with focal firms that transition receive significantly more financing. Geographic neighbors in the same local area that do not share a bank are unaffected. New loans made to the financial neighbors exhibit improved performance. This evidence generally supports the argument that the increased information flow to the bank lending to the focal firm enables the bank to lend more and with greater success to other local clients. Economic spillovers arising purely from the provision of a loan to the focal firm do not appear to have much effect on neighboring firms. Our results therefore suggest that in this large sample of small firms in an emerging market, improved information about a specific company has a clear positive spillover effect on the funding of its neighbors. A better understanding of the complex financial interrelationships between neighboring firms likely would have substantial welfare implications.

This document has been officially screened by Superintendencia de Banca, Seguros, y AFP del Peru (SBS) to ensure that no confidential information is revealed. The authors are grateful to SBS for access to the banking data, Analytics and Angel de Las Casas for geocoded data, and Kenneth Ahern and Daniel Paravisini and audiences at Columbia, the Dartmouth strategy camp, ESMT in Berlin, the European Business School in Oestrich-Winkel, the Frankfurt School of Finance and Management, MIT, the NBER, NYU, the Peruvian Central Bank, the Peruvian Economics Association meeting, Pomona College, SBS, Universidad Autonoma de Madrid, Universidad de Chile, Universidad de los Andes (Chile), Universidad de los Andes (Colombia), Universidad de Piura, Universidad Diego Portales, the University of Alicante, and UC Berkeley for useful comments. Javier Romero Haaker provided valuable research assistance. Garmaise gratefully acknowledges support from the Ziman Center for Real Estate and the Center for Global Management at UCLA. Supplementary data can be found on The Review of Corporate Finance Studies web site.

1 This policy has been adopted, for example, in the United States, Canada, France, and Germany ( Fuest and Tillessen 2005) , Ecuador ( USAID 2011) , Hong Kong ( Mok 2005) , India ( Office of the Development Commissioner 2006 ), Italy ( Becchetti and Trovato 2002) , Japan ( Japan Finance Corporation 2012) , Sweden ( Winborg and Landström 2001) , and Tanzania ( Kuzilwa 2005) .
2Degryse and Ongena (2005) study the microgeographical effects of distance on loan pricing, and Iyer and Puri (2012) analyze bank run spillovers between neighboring financial institutions.
3Berger, Frame, and Ioannidou (2011) make use of a similar registry in Bolivia.
4Source : SBS Resolution 808-2003, among others.
5 In support of this argument we find, in an unreported descriptive regression, that even when including fixed effects for current loan balance, firms that currently have COM status are dramatically more likely than MES firms to have a loan balance exceeding the cutoff in the next month.
6 We thank an anonymous referee for pointing out these issues.

References

Almazan
A.
2002
.
A model of competition in banking: Bank capital vs expertise
.
Journal of Financial Intermediation
 
11
:
87
121
.
Becchetti
L.
Trovato
G.
.
2002
.
The determinants of growth for small and medium sized firms. The role of the availability of external finance
.
Small Business Economics
 
19
:
291
306
.
Benveniste
L.
Busaba
W.
Wilhelm
W.
.
2002
.
Information externalities and the role of underwriters in primary equity markets
.
Journal of Financial Intermediation
 
11
:
61
86
.
Berger
A.
Frame
W.
Ioannidou
V.
.
2011
.
Tests of ex ante versus ex post theories of collateral using private and public information
.
Journal of Financial Economics
 
100
:
85
97
.
Collis
J.
2010
.
Audit exemption and the demand for voluntary audit: A comparative study of the UK and Denmark
.
International Journal of Auditing
 
14
:
211
31
.
Dass
N.
Massa
M.
.
2011
.
The impact of a strong bank-firm relationship on the borrowing firm
.
Review of Financial Studies
 
24
:
1204
60
.
de Soto
H.
2002
.
The other path: The economic answer to terrorism
  .
New York
:
Basic Books
.
Degryse
H.
Ongena
S.
.
2005
.
Distance, lending relationships, and competition
.
Journal of Finance
 
60
:
231
66
.
Fuest
C.
Tillessen
P.
.
2005
.
Why do governments use closed ended subsidies to support entrepreneurial investment?
Economics Letters
 
89
:
24
30
.
Hauswald
R.
Marquez
R.
.
2006
.
Competition and strategic information acquisition in credit markets
.
Review of Financial Studies
 
19
:
967
1000
.
Imbens
G.
Kalyanaraman
K.
.
2012
.
Optimal bandwidth choice for the regression discontinuity estimator
.
Review of Economic Studies
 
79
:
933
59
.
Iyer
R.
Puri
M.
.
2012
.
Understanding bank runs: The importance of depositor-bank relationships and networks
.
American Economic Review
 
102
:
1414
45
.
Japan Finance Corporation
.
2012
.
Guide to the operations of the small and medium enterprise (SME) Unit
  .
Khwaja
A.
Mian
A.
.
2005
.
Do lenders favor politically connected firms? Rent provision in an emerging financial market
.
Quarterly Journal of Economics
 
120
:
1371
411
.
Kurlat
P.
Stroebel
J.
.
2014
.
Testing for information asymmetries in real estate markets
.
Review of Financial Studies
 
28
:
2429
61
.
Kuzilwa
J.
2005
.
The role of credit for small business success: A study of the national entrepreneurship development fund in Tanzania
.
Journal of Entrepreneurship
 
14
:
131
61
.
McMillan
J.
Woodruff
C.
.
1999
.
Interfirm relationships and informal credit in Vietnam
.
Quarterly Journal of Economics
 
114
:
1285
320
.
Mok
K.
2005
.
Fostering entrepreneurship: Changing role of government and higher education governance in Hong Kong
.
Research Policy
 
34
:
537
54
.
Office of the Development Commissioner, Government of India
.
2006
.
Revised guidelines on credit linked capital subsidy scheme for technology upgradation of small scale Industries
  .
Ozbas
O.
Scharfstein
D.
.
2010
.
Evidence on the dark side of internal capital markets
.
Review of Financial Studies
 
23
:
581
99
.
Petersen
M.
Rajan
R.
.
2002
.
Does distance still matter? The information revolution in small business lending
.
Journal of Finance
 
57
:
2533
70
.
Scharfstein
D.
Stein
J.
.
2000
.
The dark side of internal capital markets: Divisional rent-seeking and inefficient investment
.
Journal of Finance
 
55
:
2537
64
.
Stiglitz
J.
Weiss
A.
.
1981
.
Credit rationing in markets with imperfect information
.
American Economic Review
 
71
:
393
410
.
USAID
.
2011
. An anchor firm approach to strengthening value chain competitiveness. Field Brief No. 13.
Winborg
J.
Landström
H.
.
2001
.
Financial bootstrapping in small businesses: Examining small business managers’ resource acquisition behaviors
.
Journal of Business Venturing
 
16
:
235
54
.
Zhang
G.
1997
.
Moral hazard in corporate investment and the disciplinary role of voluntary capital rationing
.
Management Science
 
43
:
737
50