Cheap Credit, Lending Operations, and International Politics: The Case of Global Microfinance

MARK J. GARMAISE and GABRIEL NATIVIDAD ∗

ABSTRACT

The provision of subsidized credit to financial institutions is an important and fre- quently used policy tool of governments and central banks. To assess its effectiveness, we exploit changes in international bilateral political relationships that generate shocks to the cost of financing for microfinance institutions (MFIs). MFIs that experience politically driven reductions in total borrowing costs hire more staff and increase administrative expenses. Cheap credit leads to greater profitability for MFIs and pro- motes a shift toward noncommercial loans but has no effect on total overall lending. Instead, the additional resources are either directed to promoting future growth or dissipated.

T HE PROVISION OF CREDIT to financial institutions at below-market interest rates is one of the primary tools used by governments and central banks to regu- late and stimulate the broad economy. When cheap financing is provided, the benefits of injecting capital must be balanced against the costs of distorting market incentives for financial institutions to operate efficiently. Despite the widespread use of low-priced credit, however, empirically analyzing its conse- quences is difficult. The allocation of subsidized financing may be endogenously related to countrywide economic conditions or to unobserved characteristics of the financial institutions receiving the influx of capital. In this paper, we investigate the effects of below-market credit on the operations, lending activities, and performance of microfinance institutions (MFIs) around the world. We em- ploy an empirical strategy based on changing bilateral international political relations between the host country of an MFI and the nations of its lenders, as measured by the similarity of their United Nations (U.N.) voting patterns. Com- paring MFIs in the same country, we show that those experiencing improved political relations between their home nation and the states of their lenders enjoy improvements in the terms of lending; these changes are plausibly unrelated to the characteristics of the MFIs. In this sense, shifting political

∗ Garmaise is at UCLA Anderson. Natividad is at NYU Stern. We thank Damian von Stauffenberg and MicroRate for access to the data, Cam Harvey (the Editor), two anonymous refer-

ees, Viral Acharya, Ray Fisman, Matthias Kahl, Justin Murfin, and audiences at NYU Stern, UCLA Anderson, the NBER, Drexel, the International Society for New Institutional Economics, the Pe- ruvian Banking Superintendency (SBS), the Peruvian Central Bank, and Universidad de Piura for useful comments. Natividad acknowledges the financial support of the Berkley Center at NYU Stern. DOI: 10.1111/jofi.12045

The Journal of Finance R

relationships may be viewed as an exogenous source of variation in the provision of cheap credit. The subsidized financing that is supplied is in many cases provided at rates below those of U.S. government bonds of equivalent maturity.

Given the international pervasiveness of below-market credit for financial institutions, which has been accentuated in the past few years, it is important to understand its effects. 1 In particular, it is useful to know to what extent the subsidy generates additional lending and to what degree it is dissipated or captured by managers and employees at the financial institutions. Our study analyzes MFIs, small lending institutions that resemble banks except that they typically finance themselves with loans, not deposits. They are a rising form of financial intermediary in many developing markets. We use political shocks to the supply of credit to gauge the impact of subsidized financing by studying three basic issues. First, we consider the effect of cheap financing on the hiring and operations of MFIs. We show that MFIs receiving low-cost funds increase the number of credit officers and especially augment their noncredit staff. They also have higher administrative expenses. Second, we assess the impact of below-market financing on the quantity and composition of lending by financial institutions (Bernanke and Blinder ( 1988 ), Gertler and Gilchrist ( 1993 ), Kashyap and Stein ( 1995 ), Den Haan, Sumner, and Yamashiro ( 2007 )). We show that, similar to its effects on banks in developed markets, cheap credit has little impact on the total amount of lending, but it does lead to a shift in favor of noncommercial loans (i.e., loans made to directly support the personal consumption of borrowers, as opposed to income-generating commercial loans to small enterprises). Third, we analyze the effect of the financial subsidy on the performance of lending institutions (Saunders, Strock, and Travlos ( 1990 ), Hovakimian and Kane ( 2000 ), Acharya and Yorulmazer ( 2008 )). We find that subsidized MFIs enjoy higher gross margins, but exhibit no change in their propensity to make nonperforming loans. We also find no evidence that the subsidy leads MFIs to charge lower rates to their customers.

The influence of politics on global capital flows is an increasingly important question, but it has received relatively little attention in empirical work. It is clear that for developing countries this is a first-order question as the supply of critical loans from multilateral organizations is likely affected by international political connections. Even in developed economies political proximity can facilitate bailouts (Faccio, Masulis, and McConnell ( 2006 )) and political connections can be valuable to individual firms (Faccio ( 2006 )). Given the role of globaliza- tion in explaining financial institutions’ lending activities (e.g., Cetorelli and Goldberg ( 2012 )), understanding the political economy of cross-country links is particularly important. Our results from a large sample of lenders from

47 countries and borrowers in 28 emerging markets allow us to quantify the significant impact of international political relationships on loan terms and to

1 Low-priced credit may take the form of reduced cost for government-guaranteed deposits (Gatev and Strahan ( 2006 ), Gatev, Schuermann, and Strahan ( 2009 )), central bank lending to

financial institutions (Flannery ( 1996 ), Artuc¸ and Demiralp ( 2010 )), direct support of government- owned banks (La Porta, Lopez-de-Silanes, and Shleifer ( 2002 ), Sapienza ( 2004 )), or emergency assistance (Gorton and Huang ( 2004 ), Duchin and Sosyura ( 2012 )).

1553 describe the eventual implications of this credit for the financial institutions

Cheap Credit, Lending Operations, and International Politics

that receive it. Our analysis makes use of a unique proprietary data source with both operating and loan-level financial information on MFIs. The global microfinance market has achieved considerable size—recent estimates place it as at least

$65 billion in loans to 93 million borrowers 2 —and it is fast growing. Moreover, the potential social impact of access to credit for the poor borrowers who make use of microfinance can be profound (Morduch ( 2000 ) and Pitt et al. ( 2003 )). Un- derstanding how and whether to subsidize microfinance is thus an important question with ramifications for millions of poor borrowers worldwide. As recent controversies and crises in microfinance have revealed, it is an issue on which there is little consensus. Microfinance has begun to receive sustained scholarly attention examining its impact (Karlan and Zinman ( 2008 , 2010 ), Gin´e et al. ( 2010 ), Kaboski and Townsend ( 2011 )). Most of this work focused on the effects of microfinance on the ultimate borrowers. Our study, by contrast, investigates the operations and financing of the MFIs themselves. While MFIs are increasingly important in their own right, in their basic lending activities they also share many features in common with banks. Our analysis therefore has policy implications for the subsidization of financial institutions more broadly.

We first establish that an increased similarity in the voting patterns of two countries in the U.N. General Assembly is associated with reduced interest rates and greater quantities of loans between lenders and MFIs in those countries in the following year. Specifically, we make use of a well-known bilateral measure from the political science literature (Signorino and Ritter ( 1999 )) that captures the “macro” affinity between countries in regressions explaining “micro” terms at the loan and MFI level. We find that the interest rate a lender charges an MFI decreases and the loan amount increases when the lender’s nation and the MFI’s host country become politically closer. In this test, we include MFI-lender pair fixed effects, and we use MFI-year fixed effects to control for all unobserved changes in the MFI and in economic conditions as well as the demand for credit more broadly. This finding indicates that an improvement in the affinity between the country of the lender and the MFI’s host country leads to the provision of more financing on better terms.

We next aggregate across all of an MFI’s current borrowing relationships and define its “average political shock” to be the weighted change it experi- ences in its political affinity across its full set of lenders, where the weights reflect the amounts borrowed. Controlling for country-year fixed effects, we find that MFIs with improved political affinities pay a lower average cost of debt funding. MFIs that suffer from declining affinities with their current lenders receive less subsidized financing from them, but they substitute market-rate financing with new lenders for the lost credit. As a result, an increase in political affinity is associated with an overall lower cost of credit, but has no significant effect on the total quantity of financing. A one-standard-deviation greater (i.e., more positive) political shock leads to a 88-basis-point decrease in

2 Source: mixmarket.org, accessed September 2011.

The Journal of Finance R

the average weighted cost of financing for an MFI. Along with the loan-level results, this finding suggests that our affinity measure is a suitable proxy for studying the effects of a positive financing supply shock on the hiring, lending, and performance of MFIs.

Using this empirical strategy, we show that MFIs hire more credit officers after they benefit from a positive political shock, and they particularly increase their noncredit staff. These MFIs also have higher administrative expenses and set aside greater amounts for loan loss provisions.

We find that a reduced cost of credit does not affect the total quantity of lending by an MFI, but it does lead to a shift toward noncommercial lending, as has been documented in the United States (Gertler and Gilchrist ( 1993 ), Kashyap and Stein ( 1995 ), Den Haan, Sumner, and Yamashiro ( 2007 )). Previ- ous research, however, has found it difficult to disentangle credit supply and demand effects, as subsidized financing is often provided in distressed macroe- conomic periods. Our empirical design, through its focus on the political shocks to specific MFIs in a given country, is able to control for any general nationwide economic factors. Our results therefore provide clear evidence that the supply of commercial loans is less sensitive to a financial institution’s cost of funds than the supply of noncommercial loans. This is consistent with the no- tion that commercial loans are relationship-based and thus less sensitive to a bank’s cost of funds.

MFIs receiving below-market funds do experience increased gross margins, as expected, but most other performance indicators are unchanged, including portfolio quality and the average rate charged to customers. There is no evidence of a trickle-down effect from the lower cost of credit received by MFIs.

Overall, these results provide a new perspective on both financial institution efficiency and the credit channel in a growing sector of the world economy that is largely understudied. While cheap credit is not associated with a significant increase in total loans, it does lead to more noncommercial lending. There is some evidence that subsidized MFIs use their additional resources to plan for the future by building their credit staff and increasing loan loss reserves. There is also evidence, however, that some of the proceeds are dissipated in the hiring of noncredit staff and through higher administrative expenses. The real effects linked to dissipation appear to occur more quickly in response to

a positive credit shock than those suggestive of planning for future growth. Overall, these findings help elucidate the functioning of the credit channel in emerging markets.

The role of finance in promoting economic development has been the subject of an influential and extensive literature that largely focuses on macro country-level variables. Our understanding of the mechanisms by which vi- brant capital markets encourage growth, however, is enhanced by micro-level research analyzing the actual funding and efficiency of financial institutions. We present evidence on this question, and our data from a wide set of emerging markets allow us to draw broad conclusions about the impact of subsidized credit on financial institutions themselves and on their borrowers. Our findings clearly indicate that a flow of inexpensive funding will not, by itself, result in

Cheap Credit, Lending Operations, and International Politics

1555 The rest of the paper is organized as follows. Section I describes the microfi-

nance setting and the data we use in the study. Our empirical specification is detailed in Section II . Section III discusses our results. Section IV concludes.

I. Empirical Setting

The global microfinance market consists of lending institutions that provide

small loans to poor borrowers. 3 In this section, we briefly review the evidence on the financing of MFIs and describe our data sources.

A. Financing MFIs Relatively little is known about how MFIs finance their lending activities

(Jansson ( 2003 )). While anecdotal evidence suggests that MFIs receive capital injections from social entrepreneurs, non-governmental organizations, governments, and donors, the economic terms on which MFIs receive capital to invest in a lending portfolio have remained largely unknown due to lack of data. The MFI industry relies directly on subsidized financing for its everyday activities such as lending (Morduch ( 2000 )). The stated social goals of MFIs facilitate their obtaining capital at below-market interest rates from many of their fund providers. MFIs, in turn, do not give away funds for free to their borrowers, as they incur various costs to select, serve, and monitor their clients. Although MFIs differ in the degree of their social orientation, access to finance and profitability are crucial to all MFIs, enabling them to accomplish their expansion goals in their untraditional, underserved segments of the financial services market.

B. Data Our main data source is a database of audited financial statements and

selected operating variables on MFIs provided to us by MicroRate, a leading microfinance rating agency. The data cover 133 MFIs over the period 1997 to 2009. MicroRate collects information directly from MFIs, visiting their head- quarters and branches as part of its evaluation services. The MFIs are drawn from 28 countries in Africa and Latin America. The MFIs that are evaluated by MicroRate likely represent a more successful and perhaps more commercially oriented sector of the overall microfinance market.

The MicroRate database provides audited information on both the financing

and lending activities of the MFIs. Table I shows some summary statistics. The median portfolio of loans granted by an MFI is $7 million, and the median amount of financing received by an MFI in a given year is $0.6 million per

3 Detailed descriptions of the microfinance industry are becoming more widely available. See Cull, Demirguc¸-K ¨unt, and Morduch ( 2007 ) for an industry description based on 124 MFIs drawn

from a different source, and Karlan and Zinman ( 2010 ) for a study of a specific segment in microcre- dit, the cash loan niche. Krauss and Walter ( 2008 ) assess the risk characteristics of microfinance,

and Garmaise and Natividad ( 2010 ) focus on the internal operations of MFIs and their similarity

The Journal of Finance R

Table I

Summary Statistics

The unit of observation is a microfinance institution (MFI) in a given year. Average interest rate is the quantity-weighted average of nominal interest rates on outstanding loans received by MFIs, translated into a dollar interest rate using forward exchange rates from Datastream. Total loans received is the sum over all outstanding loans expressed in millions of dollars. Continuation loans received sums over the loans of existing relationships of the MFI in millions of dollars. Weighted change in political affinity is constructed using U.N. voting data compiled by Voeten and Merdzanovic ( 2009 ) and available at http://dvn.iq.harvard.edu/dvn/dv/Voeten . Operational, credit, and performance variables are compiled by MicroRate from its client MFIs. Total staff include credit staff and noncredit staff. Administrative expenses are expressed as a fraction of the total loan portfolio of the MFI. Total loan portfolio is the sum in thousands of dollars of the loans an MFI makes to its clients. Share of commercial loans is the sum of all microcommerce loans and small business loans made by the MFI divided by the sum of commercial loans and noncommercial loans made by the MFI. Provision for loan loss is in thousands of dollars. Gross margin is interest and fee income minus interest and fee expense divided by the size of the loan portfolio in the previous year. Average rate charged to clients is defined as interest and fee income over total loan portfolio. Average loan size is the ratio of the total loan portfolio of the MFI divided by the number of loans. Portfolio quality is the fraction of the portfolio composed of loans with fewer than 30 days past due. Leverage is total liabilities over total equity.

Variable

Min. Max. # Obs. Average interest rate

Median

Mean

Std. Dev.

0.08 0.07 0.06 − 0.21 0.31 825 Total loans received (in logs)

8.47 8.33 1.87 0.88 12.51 825 Continuation loans received (in logs)

8.00 6.85 3.67 0.00 12.45 825 Weighted change in political affinity

0.00 − 0.01 0.08 − 0.54 0.68 825 Total staff (in logs)

4.76 4.77 1.18 1.39 8.63 812 Number of credit staff (in logs)

3.91 3.92 1.23 0.00 8.16 803 Number of noncredit staff (in logs)

4.19 4.16 1.20 0.69 7.66 803 Admin. expenses/portfolio

0.08 0.97 22.95 0.00 622.33 735 Total loan portfolio (in logs)

8.86 8.95 1.63 4.04 13.00 825 Share of commercial loans

0.81 0.67 0.36 0.00 1.00 735 No. of commercial loans (in logs)

8.41 6.87 4.01 0.00 12.44 824 Provision for loan loss/portfolio

0.03 0.08 1.18 − 0.01 32.00 735 Gross margin

0.24 0.29 0.17 0.03 1.13 736 Average rate charged to clients

0.28 0.32 0.15 0.07 1.25 825 Average loan size ($000, in logs)

− 0.48 − 0.62 0.88 − 3.11 1.34 795 Portfolio quality

0.95 0.84 0.28 0.00 1.00 811 Leverage

lender, with a median of five lenders per MFI. Forty-nine percent of the loan financing obtained by MFIs is from foreign lenders, and these lenders represent 54% of the distinct institutions lending to MFIs.

MFIs target microscale entrepreneurs and poor borrowers as their main customers. The median size of a loan originated by an MFI is $620, and the median number of clients served in a given year is 13,950. The overall quality of MFI investments is high, with a median of 0.95 for the fraction of loans with fewer than 30 days past due.

The U.N. voting data we use to construct a measure of political affinity are drawn from Voeten and Merdzanovic ( 2009 ). This data set contains the roll

1557 call votes of all countries in the U.N. General Assembly over the entire sample

Cheap Credit, Lending Operations, and International Politics

period.

II. Empirical Specification

A. Financing Terms and Political Affinity Many lenders to the microfinance industry choose to supply funds for non-

market motives, and these motives may also affect the terms (e.g., price or quantity) of a loan given to an MFI. In particular, we test whether an improvement in the political ties between the country of the lender and the country of the borrower leads to improved financing terms. The terms a lender offers an MFI in period t can be modeled as

(1) where S t− 1 measures the political affinity between the host nations of the MFI

LoanT erm t =a∗S t− 1 +b∗U+ǫ t ,

and the lender, U is a set of lender-specific characteristics (e.g., U may describe the lender’s propensity to supply finance for a philanthropic motive), and ǫ t is an error term.

We presume it is difficult to observe U , which may raise an issue from an econometric standpoint if the correlation between S t− 1 and U is nonzero. For example, consider an MFI in a developing nation that receives funding both from the national aid agency of a European government and from a bank in a neighboring country. The MFI’s host nation is likely to have a higher political affinity with the neighboring state than with the European country. On the other hand, the European aid agency may be extending the loan for charitable reasons (high U ), while the neighboring bank may be purely profit maximizing (low U ). In this case, the correlation between S and U is negative and estimating a version of (1) without controls will lead to inappropriate parameter estimates if U is unobservable. Our empirical specifications account for such an effect.

B. Measuring Political Affinity We adopt the popular Signorino and Ritter ( 1999 ) S variable as our mea-

sure of affinity between countries. This variable is a summary measure that describes the similarity between the voting patterns of two countries in the

U.N. General Assembly. For a given country I voting on resolution r, let P I r = 1 if the country votes “Yes,” P I r = 0 if it votes “Abstain,” and P I r =− 1 if it votes “No.” For countries I and J in year t, the affinity measure S is defined as

r= 1 |P I r −P r |

I,J,t

where R is the total number of resolutions in year t. (Resolutions for which at least one of the countries casts no vote are excluded.) The measure S therefore lies between −1 and +1, with higher values of S reflecting more similar voting

The Journal of Finance R

patterns between the two countries. Note that, S I,I,t =

1: the affinity between

a country and itself is always one.

Votes in the General Assembly rarely have direct political implications (unlike those in the Security Council), and thus may be viewed as a reasonable measure of the true preferences of states, since strategic considerations in Gen- eral Assembly voting are typically quite slight. This approach has been applied in studies of multilateral organizations (Andersen, Harr, and Tarp ( 2006 )). Fol- lowing the political science literature, we therefore view S as a time-varying measure of the true affinity between two countries.

C. The Impact of Political Shifts on Lender–Borrower Relationships To test if political affinity affects the loan contracts offered to MFIs, we ana-

lyze the financing terms of new loans provided by lender j (based in country J) to MFI i (based in country I). Following (1), we estimate

LoanT erm i, j,t =α+β∗ (S I,J,t− 1 ) + γ ∗ controls i,t +δ +λ i, j i,t +σ i, j,t , (3) where LoanT erm i, j,t is either the interest rate or the amount of the loan pro-

vided by lender j to MFI i in year t, S I,J,t− 1 measures the political affinity between countries I and J in the previous year, controls i,t is a vector of loan- level controls, δ is a fixed effect for the relationship between the MFI and i, j the lender, λ i,t is an MFI-year fixed effect, and σ i, j,t is an error term. We estimate robust standard errors double clustering at both the level of the MFI and the country of the lender (Petersen ( 2009 )). We are primarily interested in the coefficient β, which details the effect of S on financing terms. Loans are provided throughout the year, so, to avoid regressing loan characteristics on a political affinity measure calculated using subsequent votes, we always relate loan terms in the current year to voting patterns in the previous year.

Equation (3) makes use of MFI-lender relationship fixed effects, and therefore describes for a given relationship how changes in political affinity at the level of two countries affect interest rates (lender-specific unobservables such as U in (1) are absorbed in these fixed effects). The inclusion of MFI-year fixed effects (subsuming country-year fixed effects) ensures that the estimated impact of

S is unrelated to any general political or economic phenomena occurring in the MFI’s country. Moreover, these fixed effects also control for any changes to the MFI’s overall condition in a year. This specification therefore focuses only on differences among the various terms offered to a given MFI by its different lenders in a given year. We analyze how these terms change with the shifting political affiliations of the various lenders’ countries to the MFI home nation.

For example, consider a Bolivian MFI that borrows from both a Mexican lender and a Dutch lender. The specification (3) with loan rates as the dependent variable explores the extent to which higher Mexico–Bolivia affinity and lower Netherlands–Bolivia affinity in the previous year lead to relatively better rates this year to the MFI from its Mexican lender and relatively worse rates

1559 from its Dutch lender. The specification (3) with loan quantities as the depen-

Cheap Credit, Lending Operations, and International Politics

dent variable provides evidence on the impact of affinity on quantity. The fact that much of the motivation for lending to MFIs is not market driven suggests that it is reasonable to suppose that political considerations may play a role in determining both the interest rates charged and the quantity supplied.

D. Political Changes and Financing Terms for an MFI Aggregating the loan-level effects of S suggests that an MFI’s overall financ-

ing terms may depend on changes in the political relationships between its lenders’ countries and the MFI’s host nation. Changes in the composition of the set of lenders to an MFI require that care be taken in measuring an MFI’s average political affinity. For example, suppose that a Bolivian MFI borrows in period one from a Dutch aid agency with a relatively low affinity at a low interest rate (due to the philanthropic motives of the lender). If political relations between Bolivia and the Netherlands suffer, the subsidized loan will no longer

be made available in period two. Instead the MFI will receive financing in the second period from a Mexican bank (with a relatively high affinity). This loan is made on market terms and carries a higher rate.

In this case, a straight regression of rates on political affinity will mislead- ingly show a positive relationship because it ignores the changing characteristics of the MFI’s lenders (i.e., the switch from the Dutch aid agency to the Mexican bank). The loan-level specification (3) includes fixed effects for each MFI-lender relationship, but an MFI-level specification cannot include these fixed effects. Instead, we adopt a different strategy and implicitly control for unobserved lender characteristics by regressing the change in loan terms on the change in the political affinity experienced by the MFI with its current set of lenders. In this example, the negative change in Bolivia’s relationship with the Netherlands would thus predict a worsening in the MFI’s loan terms between periods one and two. This prediction is made without reference to the new relationship the Bolivian MFI forms with the Mexican bank in period two. Any future change in Bolivia’s relationship with Mexico, however, would be used to predict the subsequent change in the MFI’s financing terms.

Specifically, we calculate the average weighted change in political affinity j (S I,J,t −S I,J,t− 1 )l i, j,t

where i is an MFI identifier, t is the year, l i, j,t is the dollar value of the loan extended by lender j (in country J) to MFI i (in country I) in year t. The variable i,t describes the weighted average political shock experienced by MFI i in year t, where the weights are given by the loan amounts it receives from each current lender.

If variation in lagged S is associated with variation in credit terms, then i,t will have an impact on the change in the average financing terms on new loans received by MFI i in year t + 1. A first-differenced version of (3) at the

The Journal of Finance R

MFI-year level gives LoanT erms i,t+ 1 − LoanT erms i,t =η 1 +θ 1 ∗ i,t +κ 1 ∗ controls i,t+ 1

+µ i +ν I,t+ 1 +ξ i,t+ 1 , (5) where controls i,t+ 1 is a vector of MFI-level first-differenced controls, µ i is an

MFI-level fixed effect, ν I,t+ 1 is a country-year fixed effect at the level of the MFI’s country I, and ξ i,t+ 1 is an error term. Robust standard errors are clustered at the level of the MFI. This empirical method controls for the changing composition of the lender pool by holding the set of lenders fixed for the purposes of calculating the change in political affinity. A shift in lenders simply changes the reference point for calculating , but the change in political affinity is measured each period with respect to the current group of lenders. This allows for the appropriate aggregation of loan-level political shifts into a measure for the MFI as a whole.

This approach ignores variation in an MFI’s average political affinity that arises from its reoptimizing over its set of lenders. Instead, we take as fixed the current set of lender countries and measure how the relationships between the MFI host country and the countries of its lenders change over the year. We also do not account for the relationships between the MFI and the full set of potential lenders from which it does not currently borrow; we focus exclusively on its present group of lenders and implicitly argue that shocks to its relationships with these lenders matter more to the MFI than shocks to its relationships with other possible lenders. 4

As a variation on the previous example, if Bolivian MFI A receives most of its loans from a Mexican lender, and Bolivian MFI B receives most of its loans from a Dutch lender, then an improvement in Mexico–Bolivia affinity and a decline in Netherlands–Bolivia affinity would be predicted to lead to better financing terms for MFI A(which has experienced a positive political shock) and worse financing terms for MFI B (which has experienced a negative political shock).

We are interested in the implications that subsidized credit may have for the scale of operations, lending policies, and performance of MFIs. Unlike the loan data that are reported as flows through the year, the MFI characteristics are stock variables that are usually reported at year-end. Thus, political affinity in

a given year may have an impact on year-end MFI characteristics for the same year or in the subsequent year. For example, political affinity S t− 1 may increase both MFI characteristics Y t− 1 and Y t . As a result, the change in political affinity (S t −S t− 1 ) should have a positive effect on (Y t+ 1 −Y t− 1 ). We therefore estimate equations of the form

i,t+ 1 =ψ+χ∗ ( i,t ) + ρ ∗ controls i,t +τ i +υ I,t +φ i,t , (6) i,t+ 1 = MFIcharacteristic i,t+ 1 −MFIcharacter -istic i,t− 1 , τ i is an MFI fixed effect, ν I,t is a country-year fixed effect at the level of the

4 We thank an anonymous referee for these points.

1561 MFI’s country I, and φ i,t is an error term. For robustness, and to decompose

Cheap Credit, Lending Operations, and International Politics

the timing of the effects, we also consider results using only first-differenced MFI characteristics as dependent variables.

We view i,t as a proxy for the provision of subsidized credit in this reduced- form equation. This approach allows us to estimate the causal effect of subsidized financing on the operations and investment of an MFI. We are essentially contrasting MFIs in a given country that experienced a positive political shock (due to the nationalities of their lenders) in the previous year from those in the same country that experienced a negative political shock in the previous year. Given that MFIs are small organizations with a median loan portfolio of $7 million, they are unlikely to influence the diplomatic stances of their host nations. We therefore argue that variation in the international relations between states may be viewed as plausibly exogenous from the perspective of any given MFI.

E. Country-Year Fixed Effects All the equations (3) , (5) , and (6) include country-year fixed effects for each

MFI. (In the case of equation (3) , the country-year fixed effects are subsumed in the MFI-year fixed effects.) We therefore control for any unobserved changes occurring over time in the MFI’s home state. For example, the country-year fixed effects control for any nationwide impact of an economic crisis, changes in property rights, freedom of the press, general political character, macroeco- nomic condition of the country, etc. Identification in our empirical specifications arises solely from changes in the bilateral relationships between an MFI’s country and the nations of its lenders. In our running example, any broad impact of a national election on Bolivia’s economic performance and governance will be net- ted out by the country-year interaction fixed effects. Our approach essentially contrasts multiple Bolivian MFIs in the same year that have been differentially affected by changes in Bolivia’s relations with their lenders’ countries.

III. Results

A. Political Affinity, Lender-MFI Relationships, and Loan Terms Given the nonmarket motivation of many loans to MFIs, it is plausible that

the changing political affiliations between countries may affect loan terms. Approximately 15% of the loans in our sample are made at U.S. dollar interest rates below those of U.S. government securities of equivalent maturity. We refer to these debt contracts as “social loans.” In our first test, we relate the provision of social loans to changing affinity: we estimate equation (3) with

a binary indicator for social loans as the dependent variable. The regression uses the following loan-level controls: the number of semesters in which the MFI and its lender have had a loan relationship and MFI age. We also include MFI-lender relationship fixed effects and MFI-year fixed effects to control for all changes in the condition of the MFI (including, e.g., the performance of the

The Journal of Finance R

Table II

International Affinity and the Supply of Credit in Loan Relationships

The table reports regressions of loan-level terms on political affinity, following equation (3) . The dependent variables are all based on loan contract terms. Social loan is a dummy equal to one for whether the interest rate of the loan is lower than the United States’ risk-free interest rate (U.S. Treasury bonds matching the maturity of the MFI loans). Interest rate is expressed in U.S. dollars using forward exchange rates from Datastream and adjusted by maturity subtracting from it the U.S. Treasury bond rate matching the maturity of the MFI loans. The loan-level controls include the number of semesters in which the MFI and its lender have had a loan relationship, the age of the MFI expressed in years, as well as an unreported constant. Fixed effects for each MFI-lender pair (totaling 1,670 fixed effects) and for each MFI-year combination (totaling 816 fixed effects) are included. Standard errors are heteroskedasticity-robust and clustered separately by MFI and by country of lender. Robust t-statistics based on double-clustered standard errors are reported in parentheses.

Dependent Variable

Social Loan

Interest Rate Quantity

1=100% $ in logs

S t− 1 0.333 ∗∗∗

(2.94) (0.41) Age of MFI

(−11.53) (22.60) MFI×Lender pair fixed effects

Yes Yes MFI×Year fixed effects

Yes

Yes Yes R 2 0.64 0.61 0.69 Sample size

Yes

13,265 13,265 Number of clusters (MFI)

130 130 Number of clusters (country of lender)

47 47 47 ∗∗∗ p< 0.01, ∗∗ p< 0.05, ∗ p< 0.1.

MFI, the state of the economy, and the overall demand for credit). Due to the multiple dimensions of fixed effects, we estimate via ordinary least squares rather than using a binary model such as logit. The results, displayed in the

first column of Table II , show that, in a given relationship, an increase in the lagged political affinity of the home nations of the lender and MFI results in a significant (t-statistic = 3.19) increase in the probability that the loan provided is a social loan. Reported t-statistics are robust and double-clustered at the level of the MFI and the country of the lender. This accounts for all within-MFI and within-lender-country correlations. To measure the economic impact, we consider the variability of S within relationships over time. A one-standard- deviation increase in within-relationship S results in a 10.2% increase in the probability of a social loan relative to the mean. Social loans are substantially more likely to be provided when the affinity between the lender country and the MFI host nation improves.

Cheap Credit, Lending Operations, and International Politics

Changes in affinity also affect interest rates more broadly. We estimate equation (3) with the U.S. dollar interest rate premium as the dependent variable and present the results in the second column of Table II . (We use forward exchange rates from Datastream to convert rates from loans priced in other currencies. The rate premium is calculated by subtracting the maturity- matched U.S. Treasury bond rate from the U.S. dollar rate on the loan.) An increase in last year’s political affinity is associated with a significant decrease (t-statistic = −2.02) in the interest rate charged. A one-standard-deviation increase in within-relationship S leads to a 2.3% decrease in the interest rate

charged relative to the mean. Greater affinity between the lender’s country and the MFI home nation reduces rates, controlling for all MFI-year effects.

Political shocks also have an impact on the loan quantity. We estimate equation (3) with the log of the loan amount in U.S. dollars as the dependent variable. The results, described in the third column of Table II , show that an increase in last year’s political affinity is associated with a significant increase (t-statistic = 2.92) in the log of the loan quantity supplied. A one-standard-

deviation increase in within-relationship S leads to a 10.3% increase in the size of the loan. Taken together, these results suggest that positive political shocks lead lenders to supply financing on more favorable terms, at greater quantity, and with lower cost.

The basic intuition for the results in this section is that the supply of finance may be influenced by political factors. Earlier work shows that political ties affect government capital allocation (Faccio, Masulis, and McConnell ( 2006 )).

The funding of microfinance is often done for nonmarket reasons, so one might expect that political affinity plays a large role in determining loan rates and amounts. The results in this section clearly establish that this is so.

What drives political affinity? The S variable has become widely accepted in political science as an effective measure of shared national preferences, but there is relatively little work analyzing its determinants (Sweeney and Keshk ( 2005 )). To build some intuition for our findings, and to support the introduction

of S to the finance literature, in the Internet Appendix 5 we investigate three causes that affect bilateral political arrangements. We show that it is influenced by similarity in left–right political orientation, the participation of a country in an international conflict, and the extent of bilateral trade. Political affinity S is available for each country-pair observation every year, and in this sense we view it as a rich and objective measure that captures a wide variety of changes in policy.

B. MFI Cost of Capital and Affinity The findings in Section III.A describe how loan terms in a given MFI-lender

relationship are influenced by international political affinity, controlling for any MFI-year effects. This suggests that an MFI’s overall cost of capital may

5 The Internet Appendix may be found in the online version of this article.

The Journal of Finance R

Table III

International Affinity and the Supply of Credit for MFIs

This table reports regressions of financing variables on an MFI’s weighted change in political affinity, as detailed in equation (5) . All dependent and independent variables are expressed as differences, and the observations are at the MFI-year level. Average interest rate is the quantity- weighted average of nominal interest rates on outstanding loans received by MFIs. Total loans received is the sum over all outstanding loans, expressed in millions of dollars. Continuation loans received are those from existing relationships of the MFI. The controls include leverage, portfolio quality, as well as an unreported constant. Fixed effects for each MFI (totaling 118 fixed effects) and for each country-of-MFI×year combination (totaling 161 fixed effects) are included. Standard errors are heteroskedasticity-robust and clustered by MFI. Robust t-statistics based on clustered standard errors are in parentheses.

Dependent Variable (in Differences) Average

Total Loans Continuation

Interest Rate

Received Loans Received

$M, in logs $M, in logs

(2) (3) Weighted change in political affinity

(0.33) (−0.35) Portfolio quality

(−1.26) (1.59) Fixed effects: MFI

Yes Yes Country of MFI×Year

Yes

Yes Yes R 2 0.86 0.51 0.45

567 567 Number of clusters (MFIs)

Sample size

118 118 ∗∗∗ p< 0.01, ∗∗ p< 0.05, ∗ p< 0.1.

in part be determined by variations in political affiliations, which is the topic we consider in this section.

As we describe in Section II.A , an MFI’s average political affinity may be correlated with unobserved lender characteristics (the composition effect). To control for these unobservables, we therefore regress changes in an MFI’s average interest rate or total loans received on changes in its average political affinity.

To determine if improved political connections reduce rates at the MFI level, we estimate equation (5) regressing the first-differenced average interest rate paid by the MFI (weighted by loan amount) on the weighted change in political affinity, first-differenced leverage (the ratio of total liabilities to total equity), first-differenced portfolio quality, and both MFI-level and country-year fixed effects.

We find, as documented in the first column of Table III , that MFIs that experience positive shocks to their average political affinity subsequently

1565 enjoy significantly lower (t-statistic = −3.73) average interest rates. (Reported

Cheap Credit, Lending Operations, and International Politics

t -statistics are robust and clustered at the MFI level.) A one-standard-deviation greater (i.e., more positive) political shock is associated with an 88-basis-point decrease in the average rate paid.

This finding is consistent with the results described in Table II : MFIs whose host countries improve relations with their lenders’ countries receive lower rates and more financing from those lenders. An MFI that is subject to a negative weighted change in political affinity, however, need not experience a decline in total financing received; if capital markets are competitive, then this MFI may be able to replace the lost financing from its former lenders with new financing from new lenders, on market terms. The overall cost of financing for this MFI would increase (as we display in the first column) due to its decreased supply of subsidized financing, but the impact on its total loan supply is not clear.

As we show in the second column of Table III , the impact of an MFI’s weighted change in political affinity on its total log of loans received is insignificant (t-statistic = 0.79). This is indirect evidence that MFIs have effective access to funding on market terms and can apparently substitute market financing for subsidized financing if required. In the third column of Table III , we show that continuation loans (i.e., loans from last year’s lenders) do indeed significantly increase (t-statistic = 2.65) in an MFI’s weighted change in political affinity,

as we would expect from the results in Table II . The overall picture is quite clear: MFIs that experience a positive political shock enjoy increased financing on better terms from their existing lenders, and this reduces their average cost of capital. MFIs that experience a negative political shock receive reduced subsidized financing from their previous lenders but are able to acquire new financing on market terms that substitutes (at a higher cost) for their lost funding.

The coexistence of subsidized and market financing is apparent in the data. As we describe above, about 15% of the loans are social loans, but in 62% of the MFI-year observations the MFI receives at least one social loan. Political shocks thus serve mainly to change the subsidized/nonsubsidized composition of an MFI’s borrowing and its average interest rate, but not its total amount of financing.

MFIs are small and are very unlikely to influence their home countries’ foreign policies, so reverse causality is not a concern. Variation in its weighted change in political affinity is thus generated by credibly external political shocks that influence an MFI’s cost of capital. The political shock is unlikely to be correlated with MFI-level unobservables such as quality of investment opportunities, etc. In this sense, MFIs that benefit from a positive political shock have simply experienced good fortune that has given them access to less expensive financing.

We do acknowledge, however, that we cannot control for all potential variables that may influence an MFI’s cost of financing. For example, an MFI’s relationships with other potential lenders from which it is not currently choos- ing to borrow will likely affect its current credit terms, but this is not fully

The Journal of Finance R

captured in our analysis. It may be the case that MFIs experiencing positive political shocks from their current set of lenders also enjoy improved relations with a broader set of potential lenders, and that it is the latter and not the former that drives the improvement in their credit terms. Nonetheless, even with these reasonable concerns in mind, we argue that our empirical strategy exploits plausibly exogenous shocks to an MFI’s cost of credit.

C. Cheap Credit, Personnel, and Slack The results described in the previous subsection indicate that positive po-

litical shifts lead to an MFI’s being supplied with cheaper credit, though not with more financing overall. In this subsection we analyze the implications of this cheap credit for the operations of MFIs. In particular, MFIs with access to below-market financing may have the ability to expend resources to hire staff and expand administrative expenses.

We first consider the effect of cheap credit on the hiring policies of MFIs. We regress the change in the log of total staff on the weighted change in political affinity and we include as control variables the change in the MFI’s leverage and portfolio quality, as well as MFI and country-year fixed effects, as described in specification (6). All changes in real variables contrast the year before the political shock with the year after it to allow time for the financing effect to lead

to changes in firm policies. The results, reported in the first column of Table IV , show that a positive political shock results in a significant (t-statistic = 4.62) increase in the number of total staff, that is, MFIs make use of subsidized credit to expand their operations by hiring more staff. A one-standard-deviation greater political shock leads to a 11.8% increase in staff.

Does cheap financing lead to greater hiring of credit officers? To analyze this question, we regress the change in the log of the number of credit officers on the weighted change in political affinity and the standard controls. We report

the results in the second column of Table IV . A positive political shock leads to

a significant (t-statistic = 1.74) increase in the number of credit officers. A one- standard-deviation greater political shock is associated with a 8.9% increase in credit staff. As shown in the third column of Table IV , a positive political shock also leads to a significant (t-statistic = 3.16) increase in noncredit staff: a one- standard-deviation increase in the political shock generates a 13.7% increase in noncredit staff. The greater impact of low-cost financing on total staff than on credit staff suggests that MFIs may use this cheap money somewhat less productively. That is, below-market credit may offer MFIs some financial slack.

To gauge this effect, we regress the change in the ratio of administrative expenses to the total lagged loan portfolio on the weighted change in political affinity and the full set of controls. We find, as documented in the fourth col-

umn of Table IV , that administrative expenses are significantly increasing (t-statistic = 1.79) in the political shock. A one-standard-deviation increase in the political shock leads to a 4.4 percentage point increase in the administrative expenses ratio. Along with the evidence on the increase in staff (particularly

Cheap Credit, Lending Operations, and International Politics: The Case of Global Microfinance