WPS7577 Policy Research Working Paper 7577 The Political Economy of Bank Lending Evidence from an Emerging Market Sumit Agarwal Bernardo Morais Claudia Ruiz Jian Zhang Development Research Group Finance and Private Sector Development Team February 2016 Policy Research Working Paper 7577 Abstract This study investigates the existence of political rents in political connection, varies gradually along the political bank lending, using a comprehensive loan-level data set cycle, and is mainly offered by large and domestic banks. of the universe of commercial loans in Mexico from 2003 Consistent with the quid pro quo hypothesis, the study to 2012. Identification relies on changes in the state of finds that banks that extend political loans receive sig- origin of a senate committee chairman as a source of exog- nificantly more government borrowings with better credit enous variation in firms’ political relationship. The study quality. The study also shows that the greater credit supply finds that banks offer favorable loan terms to politically due to political connection leads to a large and significant connected firms with larger loan quantities, lower loan increase in firm-level employment and assets. The study spreads, longer maturities, and lower collateral require- provides estimates of the total social cost of political lend- ments. Furthermore, political loans exhibit higher default ing and net revenue for banks that are engaged in rent rates. To isolate the bank supply channel, a rich set of provision activity. Finally, a series of robustness tests are per- fixed-effects is included with various specifications. The formed to rule out alternative mechanisms and explanations. favorable lending increases with the strength of a firm’s This paper is a product of the Finance and Private Sector Development Team, Development Research Group. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at cruizortega@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team The Political Economy of Bank Lending: Evidence from an Emerging Market* Sumit Agarwal†, Bernardo Morais‡, Claudia Ruiz§, Jian Zhang** JEL classification: D72, DG21, L33, P16 Keywords: Bank Lending, Political Economy, Privatization, Rent Seeking * We benefited from the comments of Gene Amromin, Bo Becker, Zahi Ben-David, Allen Berger, Souphala Chomsisengphet, Yongqiang Chu, Lou Dong, John Driscoll, Omrane Guedhami, Tatiana Kostova, Chuck Kwok, Sali Li, Gustavo Manso, Gerald McDermott, Jose-Luis Peydro, Tomek Piskorski, Eric Powers, Wenlan Qian, Paola Sapienza, Amit Seru, Sergey Tsyplakov, Bernard Yeung, Yu Yuan and seminar participants at Australian National University, Goethe University, National University of Singapore, University of South Carolina and World Bank. Contact author: Jian Zhang. † National University of Singapore, 15 Kent Ridge Drive, Singapore 119245, E-mail: bizagarw@nus.edu.sg ‡ Federal Reserve Board, E-mail: bernardo.c.morais@frb.gov § The World Bank, Email: cruizortega@worldbank.org ** National University of Singapore, 15 Kent Ridge Drive, Singapore 119245, Email: jian_zhang@nus.edu.sg I. Introduction Economists have long perceived political rent seeking as a major hurdle to efficient resource allocation in financial markets, as it can distort banks’ lending decisions.1 Theoretically, it is easy to model political rent seeking in the banking sector, with government-owned banks providing favors to politically connected firms (e.g. Sapienza, 2004; Khwaja and Mian, 2005). Advocates of bank privatization argue that it should eliminate the politically motivated resource allocation, engender competition, boost efficiency and ultimately diminish rent-seeking behavior (e.g. Shleifer, 1998; Roland, 2008). However, critics claim that privatization actually does not change the requirement of government management and will lead to the loss of benefits provided via public services to the whole society (e.g. Moe, 1996). In light of the this debate, it is surprising that there are no studies examining the effect of rent seeking on credit supply in an economy with a fully privatized banking sector. In this paper, we try to fill this gap by studying the existence of political rents in bank lending in Mexico, a country with no government-owned commercial banks, using the universe of commercial loans from 2003 to 2012.2 To the best of our knowledge, this paper is among the very few studies employing a panel loan-level data set representing overall lending in one particular country (also see Khwaja and Mian, 2005, 2008). For identification, we rely on changes in senate committee chairmanship as exogenous variation of firms’ political relationship and use a geographical-based measure of political connection: whether the firm is located in a state that elected a senator who at a particular time chaired an important senate committee.3 Such political connection is likely to be valued by commercial banks since chairing a committee is a signal of political importance. Chairmanship is entrusted with discretional powers over economic resources, and chairmen are perceived to be able to influence policies and legislation affecting specific industries, among which banking.4 We start the analysis by establishing that firms headquartered in the state that elected a senator chairing a commission receive more generous loan terms. Since the chairmanship is determined through the bargaining among different political parties once all senators are elected, the ascension to chairmanship is basically unrelated to either tactical voting or events and conditions in the chairman’s state. We isolate the bank-lending channel by focusing on a firm’s borrowing pattern                                                              1 See Olken and Pande (2011) for a review of the literature on political rent seeking and efficient resource allocation and also La Porta and Lopez-de-Silanes (2003) for a detailed description of banking in Mexico. 2 The banking sector in Mexico provides a unique setting for us to examine the issue because the privatization process of the late 1980s and 1990s has eliminated public ownership of commercial banks. 3 Similar measures were adopted in previous studies (e.g. Cohen, Coval and Malloy, 2011; Kostovetsky, 2015). 4  As an example of chairman’s power in influencing fund allocation, when senator Jorge Nordhausen González took over as the Chairman of the Energy Committee in 2003, his private energy company SICSA increased by 1,300 percent the average annual contracts received from the state-owned oil company.  http://contralinea.com.mx/archivo/2005/julio/htm/senador/trafico.htm  2    over time using firm*bank fixed-effects. These fixed-effects allow us to fully absorb invariant firm-specific fundamentals that proxy for credit demand, by comparing how a firm's loan terms vary from when a chairman resides in the same state to when the chairman does not. Additionally, the fixed-effects control for banking relationships, since previous studies have found that repeated borrowing from the same lender can translate into better loan terms (i.e. Baharth et al., 2011). Finally, we saturate our specification with bank*quarter and industry*quarter fixed-effects to control for unobserved time-varying shocks to the bank and industry fundamentals such as risk or investment opportunities. In the baseline analysis, we employ a broad definition of political connection identifying 45 percent of loans as politically connected. According to this classification, being politically connected leads to a 0.7 percent increase in the loan volume lent by commercial banks, and to a 5 and 6 percent reduction in the interest rate spread and probability of collateral requirement respectively. We also find significant effect, albeit economically marginal, on loan duration. Finally, we show that political loans exhibit significantly worse performance with the default probability rising 12 percent. This fact challenges the hypothesis that improved loan terms reflect better credit quality of politically connected firms. To validate our identification, we narrow down the political connection measure by focusing on firms that, in addition to being headquartered in the same state, have stronger links to the politicians (see Khwaja and Mian 2008). These stronger relations arise either because a firm operates in an industry related to the purview of the chairman’s commission, or because it is located in the same municipality in which the chairman lives.5 The economic magnitude of the political lending result becomes substantially larger. For example, a narrower political relation – either same sector or same municipality – increases loan volume by around 2.7 percent. As an illustrative example of the importance of political connections in the banking sector, we consider the case when a senator from the state of Colima was appointed as the chairman of the agriculture committee in September of 2006. Figure 1, contrasts the dynamics of credit supply to agriculture and all industries of Colima with that of the average state in Mexico (excluding Colima) around the appointment time. In the year leading up to the appointment, lending to Colima exhibited a similar pattern to the rest of the states of Mexico. However, in the quarters immediately after the election, lending to the agriculture industry in Colima more than doubled, while lending to other industries in Colima actually decreased a touch. Elsewhere, lending in the remaining Mexican states, both in agriculture and in other industries, increased at a more moderate pace. We next explore the relative importance of bank characteristics in the context of political lending. More concretely, we contrast the lending patterns across two dimensions: bank size and nationality. We find that large and domestic banks engage more in political lending. This is consistent with                                                              5 An example of a narrower political connection is that of a firm operating in the agricultural sector that is headquartered in the state that elected the chairman of the agriculture commission. 3    the presence of economies of scale in political lending, and with the fact that the leadership of domestic banks tends to have stronger relations with national politicians. We also study the dynamics of political lending along the political cycle by focusing on the chairman's three-year term. We find that the impact of the chairman's presence contributes to a persistent increase in lending during the legislative session, reaching a peak as the upcoming congressional elections near. Specifically, on average there is little effect of political connections in the first and second years of a chairman’s term, but in the third year the impact of political loans increases with average loan volume increasing on average 1.5 percent. We further investigate the effect of political lending on the firm’s real economic activity and find a strong and significant effect on employment outcomes. More concretely, the chairmanship of a given committee is associated with a 2.1 percent increase in employment in firms headquartered in the chairman’s state. We also find an impact on the firm’s finances, including total assets, liabilities and revenues. Consistent with frictions due to asymmetric information in the “large versus small” literature, the impact of credit supply channel varies with firm size, and the effect is significantly stronger for small firms. As reward for offering political rents, we document that banks that engage more in political lending receive significantly more government borrowings, and with better quality. This finding suggests that politicians use their political power over government borrowing to offer benefits to banks in return for political loans. To explore the robustness of our findings, we conduct a series of tests to rule out alternative explanations and show the pattern holds under a variety of specifications. First, we address the concern that the chairmanship may be selected based on the economic performance of the chairman’s state. Second, we find that the politically connected firm receives greater preferential treatment when its politician has stronger political power. Third, we show that the political rent does not vary with the firm size, contradicting the social purpose lending hypothesis predicting relatively more political lending for small firms. Finally, we run falsification tests by randomly assigning the state of each chairman and show that the political lending pattern disappears. Finally, we rerun the same specifications using the less relevant commissions when constructing the measure of political connection and find that the effect wanes. Given the comprehensive coverage of our loan data, we estimate the economic cost of the documented political rents by focusing on the two dimensions that we can infer from the data. First, the deadweight loss from the increase in default loans is estimated to be 0.4 basis point of Mexico’s annual GDP. Given an OECD report that the cost of political rents averages about 5 percent of the total GDP, political lending comprises 0.08 percent of total amount of political rents in Mexico.6 Second, we document evidence of additional distortion of commercial credit allocation: firms in                                                              6 The report can be found at http://www.oecd.org/cleangovbiz/49693613.pdf. 4    the chairman’s state but not in an industry under its purview suffer from credit under-provision as a result of the political lending. Finally, we show that banks reap a net benefit from engaging in the rent provision activity of 1.6 percent of their annual net income. Two potential mechanisms can explain our results: i) Politicians exert influence on banks through bank management appointments (Khwaja and Mian, 2005; Carvalho, 2014). This mechanism however is most likely prevalent in economies with public banks where management appointments are controlled by politicians; ii) Banks channel lending to politically connected firms to exchange favors with politicians. In these settings, politicians may channel additional lending to their state and sector to benefit in upcoming elections. Our results provide support for this view. That is, though the lending decision of private banks is not subject to direct pressure from politicians, banks have incentives to adapt to political developments and adjust lending to cater to the needs of politicians as an alternative way of campaign contributions and lobbying. The benefits may include additional revenue to the bank’s business, that we provide evidence for in the paper, favorable policies and regulations (Dinc, 2005; Kostovetsky, 2015) and personal “perks” to bank top management (Bertrand et al, 2007).7 Our paper builds on the work by Sapienza (2004) and Khwaja and Mian (2005) and makes four important contributions to the literature. First, unlike the prior literature studying political rent seeking in public sector banks, we focus on an economy with a fully privatized banking sector where the existence of political rent seeking is not obvious. While privatization should help eliminate the politically motivated resource allocation, increase efficiency and ultimately diminish rent-seeking behavior, we document the existence of significant political rents in the Mexican banking sector. Second, using a time-varying measure, we exploit within-firm variation by comparing a firm’s loan terms and performance when it is politically connected and when it is not. The political connection measure used in our paper, firms operating in the politician's state, allows us to target a broader sample of borrowers and thus provide a more general evidence of political rents in banking. Third, using a matched bank-firm data, we show how political considerations in bank lending have important effects on firm’s real economic outcomes. Specifically, we show that firms use the excess credit supply to hire new employees and to increase their assets. Finally, we provide evidence of quid pro quo as banks that extend more political loans receive more and better public loans. Related Research - Our paper adds to the literature of political distortions in the supply of credit. It most closely relates to Sapienza (2004) and Khwaja and Mian (2005) as they also use loan-level data to study how political connections affect bank lending. Sapienza (2004) presents evidence of Italian state-owned banks used as instruments in the supply of political patronage: firms located in an area with a relatively stronger political party affiliation obtained lower loan spread. Similarly,                                                              7 One concern is that private banks do not have the advantage of soft budget constraints like government organizations. However, given that our estimates are smaller relative to the case of government lending, the concern of whether the rent provision can sustain is less of a problem. 5    Khwaja and Mian (2005) show that government banks offer preferential loan terms to politically connected firms (those with directors participating in an election). Their findings suggest that on average politically connected firms borrow 45 percent more and have 50 percent higher default rates. Our study differs from these two studies since we investigate the political influence on lending in a fully privatized economy. Other papers have examined preferential access to capital due to political connections in the context of campaign contribution in Brazil (Claessens, Feijen and Laeven, 2008), agricultural lending in India (Cole, 2009) and by comparing cross-country data (Dinc, 2005). As our sample covers lending to both large and small firms, our study also enriches the literature on small business lending (i.e. Petersen and Rajan, 2002; Agarwal and Hauswald, 2010). This work is also related to papers focusing on distributive politics and regional favoritism. Hodler and Raschky (2014) document that the regions where current political leaders are born have more intense nighttime light and GDP growth. Our study complements theirs by unveiling a specific but important channel through which the politician's support can benefit local economic development: greater credit supply to local firms. Other policies favoring politically connected regions include legislation (Mian, Sufi and Trebbi, 2013), mortgage and lending prior to the financial crisis (Mian, Sufi and Trebbi, 2010), biased taxation (Kasara, 2007) as well as the localized public employment schemes of state run enterprises (e.g., Shleifer and Vishny, 1994). Finally, our paper relates to the studies on rent seeking and corruption. Olken and Pande (2011) provide a complete review on the evidence of corruption in developing countries, including the definition, determinants and consequences. Due to its secretive and illicit nature, measuring corruption often entails deductive reasoning based on indirect evidence (e.g. Reinikka and Svennson, 2004; Olken, 2007; Fang et al., 2014).8 Fisman (2001) is among the first studies to use the market inference approach to calculate the value of political connections. Houston et al. (2014) find consistent evidence that banks charge lower rates because they recognize the enhancement of firm’s credit worthiness due to political connection. Faccio (2006) has similar findings in a cross- country study with 20,202 publicly traded firms in 47 countries. Rent seeking is socially harmful since it can lead to resource misallocation and bring significant cost. Evidence on the distortive consequences includes poor public good provisions (Campante and Do, 2014), high workplace fatalities (Fisman and Wang, 2015) and consumer credit under-provision (Agarwal et al., 2015). We add to the literature by offering a lower-bound estimate of the economy-wide deadweight loss induced by this form of political rents. By documenting the existence of political rents in a fully privatized economy, we also contribute to the broad literature on privatizations.9                                                              8 The few notable exceptions that can have a more direct measurement of corruption include McMillan and Zoido (2004), Olken and Barron (2009) and Sequeira and Djankov (2014). 9 Privatization takes different forms in different countries and thus an enormous literature devotes to evaluate the efficiency of privatization in the single-country case including Russia (Barberis et al,1998; Boycko, Shleifer and Vishny, 1993), Mexico (La Porta and Lopez-de-Silanes, 1994), France (Degeorge, et al, 2004), China (Sun and Tong, 6    The remainder of the paper is organized as follows: Section II introduces the Mexican banking sector and election systems. Section III describes the data and methodology. Sections IV and V report the main empirical results. Section VI presents a series of robustness checks, and Section VII concludes. II. Institutional Details on the Mexican Political System The Mexican Congress comprises the legislative branch of the government and as such, its main function is to write new laws and modify existing ones. As in the United States, the Congress of Mexico is a bicameral body composed of two separate assemblies: the Chamber of Deputies (or the lower house) and the Senate (the upper chamber). Whereas the lower house has 500 deputies elected every legislative session, or every three years, the upper chamber consists of 128 senators serving for two sessions coinciding with the presidential six-year term.10 To investigate the political influence on bank lending to firms, we focus on connections to senators – rather than deputies – for several reasons. First, as is the case in the United States, senators are perceived to be more influential than deputies since they have the power to review and approve/reject appointees and treaties made by the executive branch.11 Second, whereas deputies represent proportionally their constituents within a state, senators represent states.12 Third, when firms relocate their headquarters, they often move within the same state (possibly in a different electoral district) and rarely across state boundaries (Kostovetsky, 2015). The most relevant work of senators is done through permanent committees. Committees are formed during the first month of a legislature and last for its three-year duration. 13 These groups are regulated by the Ley Organica del Congreso and consist of a team of senators, headed by a chairman, supervising and drafting legislative bills on topics within their purview. We focus on the president of the commission, and not on all committee members, as committee members in Mexico have considerably less power compared to the chair (Rivera Sanchez, 2004). Presiding a committee is politically important since chairmen propose and direct the bills, have formal contact with the government, and access to important information and economic resources to be distributed in a                                                              2003) and Nigeria (Beck, Cull and Jerome, 2005). For the complete review of research work on privatization, please refer to Meggison (2011). 10 The 128 senators in Mexico are selected at the state level. Out of all senators, 96 are elected by popular vote with every political party nominating two candidates for each of the 32 states. The party that obtained the highest vote elects two senators, while the second most voted party elects the candidate heading the party list. The 32 remaining seats are distributed among the parties in proportion to their share of the national vote. 11 Additionally, senators serve for longer terms and have higher compensations. 12 While the most and least populous states in Mexico (Estado de Mexico and Colima) have 66 and 6 deputies respectively, they both elect three senators. To guarantee that deputies represent the constituents, deputies are selected at the electoral district level. Based on the distribution of population across the country, the territory is divided into 300 electoral districts, each with a similar number of constituents. 13 Our data set spans through three legislative sessions: LIX-LXI. Legislative session LIX starts in September of 2003, while legislatures LX and LXI start in September of 2006 and September of 2009 respectively. 7    relatively discretionary way. 14 Therefore, as chairmen of important committees, senators gain greater media and political exposure and as a result, are better able to support firms from their state.15 Moreover, being chair of a relevant committee signals substantial political influence. Presiding a committee gives political power not only to the chairman but also to his party. Therefore, several weeks of intense negotiations in the senate take place behind doors to select the chairmen of the relevant committees. This process can be summarized in two stages: First, each party postulates senators to chair the committees that best match their expertise and interest. In the second stage, the political groups in the senate negotiate the committee chairs, guaranteeing that its distribution is proportional to the seats that each party holds in the senate.16 The relative importance of the committees varies substantially depending on the topics they legislate, the operating funds they receive, and more importantly, the number and influence of bills in their purview (Langston and Aparicio, 2009). The most important and contested committees are those that deal with economic, fiscal and budgetary topics (Garcia and Sanchez, 2002). While there is no unique ranking of the more influential committees, Ballinas and Becerril (2012) identify the following committees as the most influential: Administration, Agriculture, Commerce, Constitution, Energy, Governance, Legislation, Legislation I, Legislation II, Justice, Telecommunications, Transport and Treasury.17, 18 In terms of legislative work, these committees concentrate around 78 percent of all bills received in the Senate in the period we examine. 19 According to this definition, and after hand collecting information on all 256 senators elected in 2000 and 2006, we divide senators into three types, increasing in political importance. The type I includes all senators that did not chair any commission; the second type includes all senators who have only chaired non-influential commissions; while the third type includes all the remaining senators that have presided an influential committee. As shown in Table A3, 44 percent of senators are type I, 42 percent of senators are type II, and the remaining 14 percent are type III. In this                                                              14 Recent studies show evidence that leakage of important macro-news and policies affects the distribution of private information in the financial market. (e.g. Bernile, Hu and Tang, 2015; Agarwal, et al.,2015b). 15 Senators can only preside one commission per legislature. 16 This negotiation takes place through the Junta de Coordinacion Politica, a group within the Senate formed by the senators leading each parliamentary group, as well as three additional senators: two from the largest parliamentary group and one from the smallest one. By law, the votes of the members of the Junta are weighted by the number of senators that each parliamentary group has (Articulo 81, Ley Organica del Congreso General De los Estados Unidos Mexicanos). 17 With 60 permanent committees, the Mexican senate is the upper chamber with more committees worldwide (Power, 2012; Morales, 2015). 18 For each committee, Table A1 displays a brief description of its main areas of work as well as the state of its chairman per legislature. These committees can be further divided into ones that are related to specific economic sectors (e.g. Energy) and those that purview a more global agenda (e.g. Treasury). In the last column of Table A1 we display the economic sectors that are directly related with the purview of a committee using the NAICS 2007 classification. 19  Detailed data on all the bills received in the Senate for the years of our study was downloaded from the website of the Legislative Information System of the Mexican Ministry of Governance. The information for each bill includes its date, type of bill, topic, person/institution/political party that presented it, and the committee the bill was channeled to.  8    paper, we focus our analysis on the political connection that firms have with senators chairing the most relevant committees. These chairmen in principle have access to more political resources and are perceived by banks as important political figures. To formally verify the importance of chairing a commission and validate our classification of influential commissions, we evaluate three indicators of the political importance of the three types of senators. The first measure indicates whether the senator headed their party’s state list for the senate elections. The second measure is an indicator of whether the senator held an important political position in the past, like state governor, federal senator, federal deputy, ambassador or secretary at the state or federal level. Finally, the third measure is an indicator of whether the senator will hold an important political position after chairing the commission, including state governor, federal senator or ambassador. The results are displayed in Table A3. All measures support our classification and corroborate the relative importance of chairing a commission. In the second line, we find that only 44 percent of type I senators headed their party’s list for the senate elections, contrasted with 67 percent and 81 percent of type II and III senators respectively. Regarding their previous occupations we find in the third line that only 4 percent of type I senators had relevant past positions, compared with 30 percent and 56 percent of type II and type III senators respectively. Finally, we find that once their term as senators ends, chairmen of important committees are also more likely to hold important federal or state appointments in the future. In line four, we find that 5 percent of both type I and type II senators end up with important political positions, compared with 39 percent of type III senators. III. Data and Methodology III.A. Data The data in our paper come from four primary sources. The first is the loan-level data for the universe of commercial lending in Mexico from 2003 to 2012. The second is the monthly bank balance sheet data set provided by the Bank of Mexico. The third is hand-collected data on personal and professional information on all 256 senators elected in 2000 and 2006. This information includes all political occupations as well as the commission memberships. Finally, and to document variation in real economic outcomes, we use yearly information on firms’ balance sheets using the Orbis data set compiled by Bureau van Dijk. The loan-level data are a supervisory data set on the entire universe of commercial loans in Mexico from September 2003 to March 2012. The data set was compiled by the Mexican Banking and Securities Commission (CNBV) in its role as bank supervisor. The data come from regulatory reports (known as R04) sent monthly by every commercial bank to the CNBV. Reports are mandatory, updated electronically, and include detailed characteristics of all new and continuing 9    loans made to firms by every bank in Mexico.20 All business loans, regardless of their size, have to be reported.21 For each loan in the CNBV data set, we have information on the issuing bank, the borrower (firm), the outstanding amount, the annualized-loan spread, and the start and end date of the loan (maturity). There is information regarding each borrower, such as its identifier, location, industry and number of employees. Since loans are tracked every month, we are able to see their evolution until their maturity. Specifically, we can observe whether the debtor obligation is being fulfilled, and in case it is not, by how much and for how long the loan has been under-performing.22 The final sample covers three legislatures LIX-LXI from October 2003 to March 2012.23 For each loan, we keep the information at the initial period when the loan was extended. However, we also record whether the loan is more than 90 days in arrears within the 36 months after origination. Some borrowers had more than one loan outstanding with the same bank in one quarter. Therefore we adopt a similar approach as in La Porta et al. (2003) and aggregate individual loans at the firm- bank-quarter level and report loan characteristics such as loan spread and maturity at origination using a weighted average by loan volume. The only exception is loan volume, which is the sum of the value of all new loans that a firm initiated from a certain bank in a given quarter. Our final data set contains 594,534 loans from 88,117 firms from 21 banks. During the first month of each legislative session, committees are formed with one senator appointed as chairman. For each borrower and quarter, we define the dummy variable chairman that equals one if the borrower is headquartered in a state with at least one senator serving as the chairman in that quarter and zero otherwise. III.B. Empirical Strategy We now discuss the methodology used to analyze whether banks offered better loans to firms in a state with a relevant chairman, and whether these loans are more likely to eventually default in the future. We focus on four credit margins: loan size (in logs), price of the loan (interest rate spread), original maturity (in months) and a dummy on the existence of collateral.24 We also examine the loan's ex-post performance by creating an indicator variable of whether loans were more than 90 days in arrears. Appendix A presents the detailed definitions of each variable.                                                              20 To ensure the homogeneity of the data, we exclude sole proprietorships from the sample because consumption credit can be confused for commercial credit. 21 Many loan-level credit registry data sets are truncated from below. For example, the threshold is 1.5 million euro in Germany and 41,300 euros in Italy (Sapienza, 2004), which suggests that small and medium firms may be excluded in those credit registries. 22 Therefore, we are able to calculate the recovery rate of defaulting loans. 23 Legislative session LIX started in September of 2003, while LX and LXI started in September of 2006 and September of 2009 respectively. 24 We calculate the interest rate spread as the difference between the loan interest rate charged and the annual cost of funds reported monthly by commercial banks to the Bank of Mexico. 10    As discussed earlier, the major empirical challenge to identify political rents is that it is difficult to find an exogenous measure of political connection. A widely used method is to define a firm as politically connected when one of its large shareholders or top officers is a politician such as a parliament member, minister or election participant (e.g. Khwaja and Mian, 2005; Faccio, 2006). Nevertheless, such a static measure brings several identification concerns. First, including firm fixed-effects becomes infeasible with such a relatively time-invariant variable for each firm. Therefore, selection bias is present in the estimates of political effect, which may be overstated. For example, large enterprises are more likely to attract politicians as shareholders or board members. Therefore, the observed preferential treatment is an inaccurate reflection of political rent. Second, without considering firm-specific credit-demand shocks, both the demand and supply channels may confound the estimation (Berger, Molyneux and Wilson, 2014). Third, the analysis with the time-invariant measure of political connection is equivalent to a cross-sectional procedure since there is no time variation in the main variable of interest. Given the above concerns, we use a strategy exploiting within-firm variation, and examine whether each firm receives preferential treatment when its headquarters are located in a chairman’s state, a proxy for political connection, versus when they are not. Therefore, we use a time-varying measure of political connection including firm*bank fixed-effects to exploit changes in the credit availability to a firm-bank pair over time and across states. 25 Specifically, the baseline identification strategy relates the credit outcome of each firm-bank pair in a given quarter, to an indicator of whether the state where the firm is headquartered has any senator presiding a committee in that same quarter. Our benchmark specification is given by the following equation: , , , , , , , , , 1 where , , corresponds to each of the credit margins (logarithm of loan volume, loan spread, original maturity and collateral) or the ex-post loan performance (an indicator variable that equals one if the loan is in arrears for more than 90 days) for a loan obtained by firm f from bank b in quarter t. The variable of interest , is an indicator of whether the state where the firm f is located has at least one senator as the chairman during quarter t. , includes state-level annual GDP growth and unemployment rate to control for time-varying economic conditions in the firm’s state. To isolate the bank lending channel, we include firm*bank fixed-effects , , to exploit the variation within the same firm-bank pair over time. This fixed-effect controls for all time-invariant characteristics of firms and banks that proxy for credit demand (such as geographical location, bank nationality or firms' economic activity), as well as for bank-firm relationships. Shocks to credit supply can affect the banks' decision to extend loans (Khwaja and Mian, 2008), so we further isolate the unobserved time-varying factor in the supply of credit by including                                                              25 In our data set, 99.2 percent of the firms remain in the same state throughout our sample period. We use these firms in our analysis. Therefore, including firm-bank fixed-effects, would absorb any state fixed-effects. 11    bank*period fixed-effects , . Finally, we add fixed-effects at the industry*quarter level , to control for unobserved time-varying shocks to the industry’s fundamentals such as risk or investment opportunities.26 Therefore, we explore the variation on the outcomes of loans extended to firms within the same industry in states with and without any senator presiding a committee. In other words, this specification compares, at a given time, the credit conditions of firms from the chairman's states with those in the same sector but headquartered in other states. Therefore, captures the preferential lending the bank extended to the same firm when it is politically connected versus when it is not. Since some committees are industry-specific, we also estimate whether the preferential treatment extended to politically connected firms differs across industries within the same state. The effect is represented by an interaction term of , and , in the regression outlined in equation (2) , , , , , , , , , , 2 where equals one if for a given firm f, a senator from its state chairs the committee that directly supervises its industry. The coefficient of interest offers a difference-in-difference estimate of this preferential treatment. It contrasts the credit terms of politically connected firms within the same industry versus firms in other industries. As with the benchmark specification, we saturate our specification with fixed-effects at the firm*bank, bank*quarter and industry*quarter level. III.C. Summary Statistics Table 1 presents the summary statistics of our sample. Panel A shows the composition of the loan- level data according to loan attributes. We create six dummies for each loan according to whether the loan is: (1) made to politically connected firms (firms from the chairman's state); (2) made to politically connected firms within the same industry as the chairman; (3) made to the politically connected firms located in the municipality in which the chairman lives; (4) provided by large banks; (5) provided by domestic banks; (6) made to large firms. The table displays the mean and number of observations for each dummy variable. The variable Chairman indicates whether a loan is in a state-sector-period with a chairman. On average, 45 percent of loans are extended to politically connected firms. The variables Chairman*SameInd and Chairman*SameMuni represent whether loans are made to politically connected firms in the same industry or municipality as their chairman. 5 percent of the loans in our sample are from firms in the same industry as the chairman, while 6 percent of the loans are obtained by firms from the chairman’s municipality.                                                              26 Industries are classified at the three digit 2007 NAICS. 12    As discussed earlier, the impact of a chairman may vary with bank size since there may be scale advantages. A bank is defined as large if it has on average more than 150 billion pesos, roughly 13 billion dollars, in assets at 2011 prices. Under this classification, 7 out of the 21 banks are large. The list of banks is presented in Table A2. The segmentation of banks is relatively stable throughout the sample period, given that the biggest bank in the small group has substantially lower assets, 93 million pesos, than the threshold. Under this division, large banks have 81 percent of the total number of commercial loans. Furthermore, 35 percent of the number of loans are provided by domestic banks. The extent of the political rent may also vary with firm-level characteristics such as the size of borrowing firm. We divide the sample into large and small firms, using 50 employees as the threshold. In our sample period, 19 percent of the number of loans are extended to large firms. Panel B shows summary statistics of loan-level variables including loan volume, loan spread, maturity, collateral and default rate for the full sample, as well as for the loans to politically connected and non-connected firms. The average loan is worth 2,776,000 Mexican pesos (roughly 237,000 dollars). On average, politically connected firms receive loans worth 2,857,000 Mexican pesos which are 5.5 percent larger than the average for non-connected firms. Furthermore, political loans are also associated with preferential treatment in other dimensions. Compared to the rest of loans, political loans have lower loan spread spreads (11.6 versus 12.3 percent) with the difference being statistically significant at 1 percent. In other words, politically connected firms pay average loan spreads that are 5.6 percent lower than those paid by firms without political connections. The magnitude of this estimate is economically comparable to that of loan spread in the literature of relationship lending (i.e. Baharth et al. 2011). The difference in collateral requirement is also large: the probability of having collateral requirement is 5 percentage points lower for political loans.27 The political rent hypothesis predicts that political lending takes the form of advantageous terms and politically connected firms default on the loans that are accumulated with the intention of not being returned (Khwaja and Mian, 2005). The last row shows the incidence of bad loans in our sample. Consistent with the rent-offering hypothesis, the default probability for political loans is 0.7 percentage points higher than that of non-political loans. In other words, political loans are 12 percent more likely to fall into arrears. IV. Results: Preferable Lending Terms This section investigates the link between credit supply and political connections. As noted earlier, in order to exclude concerns regarding unobserved heterogeneity, all specifications include firm*bank and bank*quarter fixed-effects. Standard errors are clustered at the state level. IV.A. Lending Terms                                                              27 Politically connected firms also receive statistically significant longer maturities on their loans relative to non- connected firms. However the magnitude on this dimension is fairly small (0.1 months). 13    Table 2 presents evidence of preferential lending to politically connected firms. Column 1 shows the baseline results in terms of credit availability. The dependent variable is the logarithm of loan volume multiplied by 100. The point estimate is 0.76 and statistically significant at the 1 percent level, suggesting that credit conditions loosen significantly for the average firm in states with a senator heading the committee. To control for unobserved time-variant shocks to the industry, we further include industry*quarter fixed-effects in column 2. The estimate decreases to 0.72 remaining significant at 1 percent. Therefore, for the same firm, relative to the status of not being connected, having the political connection (that is during periods-states with a chairman) leads to a 0.72 percent increase on the average loan volume. Loan spreads and maturity are the least affected margins of politically connected firms, since the economic magnitudes are relatively small. Column 3 suggests that the loan spread charged for loans made to politically connected firms is, on average, 2 basis points lower. When we include industry*quarter fixed-effects in column 4, the estimate remains stable and statistically significant at 10 percent level. Columns 5 and 6 report the results for loan maturity. The positive coefficient corroborates the evidence of preferential treatment to politically connected firms: political loans on average obtain loans with maturity of 0.09 more months than non-political ones. Relative to the mean, political loans have a loan spread 0.2 percent lower and a maturity 0.6 percent longer. Finally, the last two columns indicate that loans offered to politically connected firms are 1.8 percentage points less likely to be collateralized. In relative terms, this translates to a lower probability of 5 percent. In summary, we present evidence of favorable lending to politically connected firms by showing that they have more access to credit, can borrower at lower rates, with longer maturities and lower collateral requirement than their counterparts. This preferential treatment persists after controlling for the demand effect of the firm, bank-firm relationship and unobserved time-varying shock to the bank and industry. Compared to the results in previous studies, our estimates of the advantageous loan terms are smaller for at least two reasons. First, we are exploiting the within-firm variation and documenting the existence of political rents offered by private banks rather than by government banks, where politicians are more likely to exert influences on the lending process. Second, we are using a more generalized and much broader measure of political connection to draw inferences on the value of connection (firms from the politician’s state), classifying almost half (45 percent) of the loans as politically connected. As we narrow down the measure of political connection in the following sections, the effects become much larger. IV.B. Ex Post Performance We interpret the documented preferential lending as evidence of political rents provided to politically connected firms. However, a plausible concern is that this may simply reflect better credit quality of politically connected firms since banks may have access to “soft information” to support the favorable lending decisions (Keys et al, 2010). In this section, we conduct a direct 14    examination of this hypothesis by looking at subsequent performance after the loan origination. If the preferential lending is motivated by the higher credit quality rather than political connections, one should expect a better performance for these political loans. We follow the literature by focusing on the share of loans that is more than 90 days past due (Gross and Souleles, 2002; Agarwal et al., 2011), and use default dummy as the dependent variable in Table 3. The default dummy is an indicator variable that equals to one if loans of a firm-bank pair extended in a given quarter were ever more than 90 days in arrears. The result is consistent with the political lending hypothesis. The default probability for political loan is 0.7 percentage points significantly higher or, equivalently, 12 percent relative to the average default rate. IV.C. Narrowing Down the Political Connection Measure In this section, we impose a stricter restriction on the measure of political connection in the baseline analysis and characterize a firm to be politically connected if it has stronger links to the politicians in addition to being located in the chairman’s state. Specifically, we use two different methods to measure the connection strength. First, we use an industry-specific measure. That is, operating in the industry related to the purview of the chairman’s committee leads to a stronger political connection, since a significant portion of the chairman's power is linked to the specific industry for which the committee writes legislation and provides direct regulation. Our second measure tests whether the beneficiaries of political rent seeking are more likely to be firms headquartered close to the politician’s hometown. Therefore, our second proxy is whether firms are located in even narrower geographical areas around where the political leaders live. We formally test the impact of connection strength on access to credit based on these two measures and report the results in Panels A and B of Table 4 respectively. In columns 1 to 8, we find that politically connected firms with stronger measure of political connectedness, obtain on average much better loan terms than firms that merely have their headquarters in the chairman’s state. For example, when chairman’s areas of governance are mapped to firms in their states, these connected firms are able to obtain loans 0.72 percent larger. When we use the second measure, being in the same municipality, loans are 2.2 percent larger and have a 6.4 percent lower probability of collateral requirement. In columns 9 and 10, we analyze the ex post performance following similar specifications as to the previous section. We find that the political loans offered to connected firms with same industry (municipality) are more likely to fall into default by 1.8 (2.6) percent in Panel A (B). Overall, the results provide an internal validation for the baseline measure of political connection and suggest that politically connected firms with stronger connection obtain even larger rents from banks. V. Further Results on Political Rents V.A. Bank Size 15    The incentive to extend political lending varies across banks with different size. Large banks should have stronger motivation to engage in political lending as they have the advantage of economies of scale and should benefit more from regulatory favor policies from the government ex-post. We identify a bank as large if it has on average during our sample period more than 150 billion Mexican pesos in assets, in 2011 prices (roughly 13 billion dollars). Such threshold is based on the observation that the largest of the small banks has substantially lower assets, 93 billion, than the threshold. This contributes to a more precise identification since this segmentation of banks is stable throughout the sample period. Table 5 presents the results of tests on loan terms in columns 1 through 8 and ex post performance in the last two columns. The coefficient of interest is the interaction term that reveals whether politically connected firms are able to earn even higher rents from large banks. The result is consistent with the above hypothesis and illustrates that political rents to firms located in the chairmen's state are mainly provided by large banks. Specifically, compared to small banks, we observe that the loans made by large banks to politically connected firms are larger by 1.2 percent, charged a loan spread 0.2 basis points lower, have a longer maturity by 0.1 months, and are less likely to require collateral by 1.6 percent. Columns 9 and 10 confirm the heterogeneity of political rents across bank size in terms of ex-post performance. Political loans are more likely to be offered by larger banks: the coefficient before the interaction term is positive and statistically significant. The estimate in column 10 indicates that compared with loans made by small banks to politically connected firms, loans made to the same firm by large banks have significantly higher default probability by 0.6 percentage points in absolute terms. V.B. Domestic versus Foreign Bank We examine the impact of bank ownership structure on rent-offering by comparing the lending of domestic and foreign banks.28 There exist two conflicting views on the impact of bank ownership. On the one hand, the leadership of domestic banks tends to have stronger political relations relative to that of foreign banks. Therefore, we would expect domestic banks to provide more political loans than foreign banks. On the other hand, previous studies suggest that domestic agents have a comparative advantage over foreign agents in overcoming some of the obstacles associated with corruption and weak institutions (e.g. Aizenman and Spiegel, 2002). To the extent that domestic banks are more likely to have alternative options, the domestic ownership may hinder bank’s incentive to cater to the politicians and offer political rents. Table 6 presents evidence suggesting that domestic ownership of the bank is more conducive to political rent provision. As can be observed in columns 1 through 8, politically connected firms receive 1.1 percent larger loans from domestic banks than from foreign banks, and the difference is significant at 1 percent. Political loans by domestic banks are charged with a similar loan spread, maturity and probability of collateral requirement. Most importantly, we find that loans made by                                                              28 Table A2 presents the characteristics of domestic and foreign banks. 16    domestic banks to politically connected firms have significantly higher default probability, roughly four times, as demonstrated in columns 9 and 10, which lends further support to the notion that domestic banks provide greater rents to politically connected firms. V.C. Lending during the Election Cycles Rent-offering activities like political lending documented in this paper depend on political considerations and thus are supposed to vary according to regular election cycles. We next exploit the time-series variation of the political lending within the three-year term of the chairman. Mexico has a decades-old election system forbidding a second term of a senator. As a result, if banks offer political rents through preferential lending, one should expect such patterns to be relatively invariant throughout the term. We plot the time series of political lending in Figure 2. The horizontal axis presents the length in quarters since an election for the chairman while the vertical axis shows the difference in the average loan volume in states with and without a chairman relative to their overall mean. There is a persistent increase in difference of credit supply between states with and without a chairman throughout the election cycle. The graph suggests that, two quarters after an election, the loan volume in states with a chairman is 0.7 percent higher relative to its mean than in states without a chairman, and the gap increases to over 2.5 percent ten quarters after the election. To formally test this conjecture in the regression setting, we break the dummy Chairman into three dummy variables based on the timing and estimate the following specification: , , ∑ _ , , , , , , , , 3 where the dependent variables are the same as in equations (1) and (2) and include the loan terms plus ex post performance. The variable of interest, Chairman_Yeari,f,t, is an indicator for the year i during the chairman’s term. As before, in this regression we control for the unobservable firm- specific credit demand and firm-bank relation by using industry-quarter fixed-effects. The coefficients of interest are β1, β2 and β3 and the results are presented in Table 7. The results indicate that political lending becomes gradually expansive throughout the legislature. For example, in columns 1 to 4, being politically connected only weakly affects the loan volume and loan spread in the first year of a legislature. However, the impact becomes substantially larger in the second and third years with an increase from 0.2 to 1.6 percent for loan volume and a decrease of 0.2 to 3.1 basis points in loan spread. The expansive pattern of political lending is best reflected in the results for collateral and ex post loan performance displayed in the last four columns. The lower probability of collateral and higher default rate of political loans exhibit a persistent pattern among the loans made throughout the whole legislature. The only exception is the result for maturity in columns 5 to 6 where political loans made in the final year of the legislature do not have significant longer duration unlike during the first two years. Our results further validate our 17    identification and exclude the concern that the documented political lending is driven by the power of being the senator instead of serving as the committee chairman. V.D. Quid Pro Quo: Rewards to the bank The above results suggest that politically connected private firms receive favorable lending from commercial banks. Consequently, one expects banks to reap some benefits from this increased loan supply. One possible reward from government officials, is for them to influence loan demand by public firms to these commercial banks. In other words, we investigate whether politicians reward the rent providers through public loans (i.e. by extending more "business" to them, in the form of public loans). Public loans are an important part in the banks’ overall businesses and on average represent 15 percent of all bank commercial lending.29 What is more, the choice of lender is largely at the politician’s discretion. We first classify banks into two types according to their share of public loans as a fraction of total commercial loans, and create a dummy variable LargePubLoan depending on whether the bank receives high level of public loans relative to the mean.30 To the extent that banks receive benefits in terms of large government borrowing, we would expect that the political lending is more likely to be provided by banks with high level of public loans. We formally test this in Panel A of Table 8. As observed in column 2 of Panel A, the lending from banks with high level of public loans is associated with a further 0.6 percentage point increase in the volume lent, roughly doubling the impact of a chairman. Columns 7-8 similarly show that these loans also have a lower probability of collateral requirement. Finally, columns 9 and 10 confirm a higher probability of default for loans made by banks with high level of public loans. These results indicate that banks receive rewards in terms of more government borrowing for extending the political lending to the local firms in the chairman’s state. We now offer evidence that banks that have a larger share of public loans on their balance sheet tend to offer more public loans in the states that have a chairman. More specifically, we aggregate all the public loans at the state-bank type (LargePubLoan/SmallPubLoan)-quarter level. The panel structure of the data enables us to examine whether the political rent provider receives any benefit in the public loan market. Panel B contrasts the government borrowing from the two types of banks and is saturated with state-level controls and a rich set of fixed-effects. First, the state*banktype level fixed-effects is included to exclude the possibility of differences across states and bank size and explores the variation within each banktype-state pair.31 We also include quarter fixed-effects to control for any time-varying shocks to the macroeconomic environment.                                                              29  Table A2 presents the average bank-level shares of public loans. While there is a wide variation across banks, the largest banks tend to have the highest shares of public loans in their loan portfolios.  30 Under this classification, 8 banks are considered to receive a high level of public loans. SmallPubLoan = 1 – LargePubLoan. 31 For example, there may be significantly more public loan originations from larger banks in well-developed states. 18    The results are largely consistent with the “Quid Pro Quo” hypothesis. As expected, banks with high level of public loans get more "business" from the government and lend out a significantly larger volume of public loans (by 13 percent as shown in columns 1 and 2 of Panel A). More interestingly, we find that these public loans are generally of better credit quality, as indicated by the significantly lower default rate (0.8 percent lower) in column 10. The results on other loan terms are not robust across different specifications, though column 6 of Panel A suggests significantly longer maturity for public loans made by banks with high level of public loans. To sum up, our results suggest a specific channel of rewards in the form of government borrowing, with larger volume and better credit quality for banks that offer political rents and extend favorable lending to politically connected firms. Nevertheless, we should caution that the results here do not aim to establish the sequence of whether “the politicians offer rewards to the bank” or the “bank provides favorable lending to politically connected firms” comes first. V.E. Real Effects of the Political Lending In this section, we investigate whether the effects of political connections in the loan market can translate into effects on real outcomes for the borrowers. Starting from Bernanke (1983), much of the macroeconomic and finance literature examines how the credit channel can propagate shocks to lending institutions into outcomes in the real economy. However, the data limitations have made it difficult to identify the effect on the borrower’s economic outcomes with a few exceptions.32 We study the impact of political lending on firms’ outcomes using a vast sample of matched bank-firm data that merges the loan database with balance sheet information of the borrower. The specification is presented in Equation 4: , , , 4 where , corresponds to the firm-level information including assets, liabilities, revenue and employment. The variables are available at the annual frequency and collected from Orbis, which includes a rich set of financial and ownership information for both public and private firms around the world. The variable of interest , indicates if the state s, where the firm f is located, has at least one senator as the chairman at the year t. We cannot include firm*year fixed-effects as they would absorb the main independent variable. So we saturate the specification with fixed- effects at the firm level to control for time-invariant unobservable firm characteristics and also use year fixed-effects to exclude the overall effect of time. β1 is our main coefficient of interest and an estimate statistically significant from zero would imply that the preferential treatment to politically                                                              32 Gan (2007) and Almeida et al (2012) document contraction in investment in affected borrowers but focus only on firms that have regulatory filings with borrower-level information. Chodorow-Reich (2014) relies on the 2008 financial crisis as shock to credit supply and show that it has a significant effect on employment at small and medium firms. Carvalho (2014) shows in the context of Brazil that firms expand employment near election in exchange for government bank loans. 19    connected firms has a real impact on real economic activity. Table 9 presents the results of the impact of political lending on firm-level economic activity. Though the merge with Orbis data set leaves fewer observations, we still find that the preferential treatment does translate into real effects on firms. The excess credit supply due to political connection has a large effect on employment outcomes, which is statistically significant at the 1 percent level. The economic magnitude of the estimate is also meaningful. Having at least one senator serving as the chairman of any committee results in an additional increase in employment of 2.1 percentage points in the ordinary least squares specification. This estimate is comparable to that of Chodorow-Reich (2014), examining the impact of a reduction in credit supply due to the financial crisis on firm-level employment. With higher credit availability, firms become more capable of taking full advantage of profitable opportunities; thus, they can increase investment that leads to better performance. In columns 3 to 8 we also examine the impact of excess credit supply on firm’s financial information. More specifically, political lending increases total assets, liabilities and revenue of firms by 1.2, 1.0 and 0.8 percent respectively. However, the last two estimates are no longer statistically significant. The “small versus large” literature has documented that due to greater asymmetric information or smaller buffer savings, smaller borrowers exhibit greater sensitivity to changes in credit supply (Gertler and Gilchrist, 1994; Chodorow-Reich, 2014). Therefore, we conduct a subsample analysis by only looking at the sample of small firms (firms with less than 50 employees), and document a much stronger effect. For example, political connection leads to 2.5 percent increase in employment and 2.7(2.4) percent rise in total assets (liabilities). This greater sensitivity for small borrowers provides supporting evidence of asymmetric information in the lending markets. VI. Extension and Robustness Tests VI.A. Endogenous Selection of the Chairman One threat to the identification strategy is an endogenous selection hypothesis. More concretely, the possibility that the committee chair is selected based on the economic performance of the corresponding sector in his or her state. For example, if a senator from Tamaulipas is selected to chair the energy committee because of the outperformance, or expected outperformance, of the energy sector in Tamaulipas, the empirical strategy would confound the effect of the chairman with the fact that the energy sector of Tamaulipas would have bloomed regardless of who the chair was. Such an explanation is unlikely given the process by which the committee members are selected, as noted in Section II. However, and while the institutional background makes it difficult to believe the endogenous selection hypothesis, we present several checks disproving this explanation. To test whether the selection of chairman actually depends on the relative performance of the sector in his or her state, 20    we first compare sectors across states with and without senator serving as the chairman in the year when committees are formed (2003, 2006 and 2009). Given the data availability, we focus on five sectors: Agriculture, Commerce, Transport, Energy and Telecommunications. We obtain sector indicators from the Banco de Informacion of the National Institute of Statistics and Geography (INEGI) and compile data at the state-year level on the performance of different sectors across states and over time. The data set measures the level of activities in each sector with different indicators and contains information on total agricultural production, gross production of the commerce sector, registered vehicles in circulation, users of electric energy, and length of highway network, proxying for the performance of the agriculture, commerce, transport, energy and telecommunications sectors respectively. The results presented in Table 10 contradict the endogenous selection hypothesis. For states with a senator chairing the committee, the level of economic activity in the corresponding sector is not significantly different from that of the same sector in the other states. For example, the share of agricultural production relative to the state’s GDP in states with a senator heading the agriculture commission is slightly higher than that in the other states (6.4 percent vs 4.6 percent), but this difference is not statistically significant. To exclude the possibility that the appointment is based on expectation of sector performance instead of current level, we also compare the growth rate of the sector-level activities in the bottom of Panel A. Again, we find no evidence suggesting that the economic activity in the corresponding sector of the chairman’s sector is significantly different from that of the same sector in other states of Mexico. In addition to the sector activity, we also examine in Panel B general economic conditions of the state and compare two key economic indicators, GDP and employment growth, for the chairman’s state versus other states when the committee is formed. As an extension, we plot the time series of GDP and employment growth rate throughout our sample periods from 2004 to 2012 in Figure 3. As can be observed, the economic development for the chairman’s state exhibits no significant difference when compared to the rest of the Mexican states. Taken together, the results support our identification strategy of using the appointment of chairman as an exogenous variation of political connection as they are not likely to be endogenously determined. VI.B. The Gradient of Political Power We perform two other robustness checks of our results. Specifically, we focus on two types of states that should be at the top of the power pyramids. That is, those states that have more than one senator chairing a committee, and those states with a chairman serving in a core committee with relatively higher importance, according to a more systematic measure. The political lending hypothesis predicts that, all else being equal, we should observe a stronger pattern for those states given their greater influence on bank lending. Table 11 reports the regression results showing the effect of having more than one chairman at a time. In all specifications, we include a full set of controls and fixed-effects similar to our benchmark exercise. In Panel A, we find that firms 21    headquartered in states with more than one committee chairman receive even more favorable lending, roughly three times larger, and the political loans also have significantly worse ex-post performance, with default rates four times higher. In Panel B, we limit our attention to the core committees that are more relevant to the banking industry, such as Legislation, Governance, Treasury, Justice, Commerce, Constitution and Labor, and find similar results: firms located in states with a chairman from those committees are more likely to receive political rents.33 VI.C. Social Lending Hypothesis Another potential explanation for the political lending result is the "social purpose lending" hypothesis: to fulfill the goal of promoting development in a certain sector, the new chairman is likely to engage in projects with social gains but low returns, pushing banks to extend lending to such projects. This hypothesis may be particularly stronger for government banks since, according to the social view, they bear the responsibility to maximize social welfare. However, given that there are no government-owned banks in Mexico, it is infeasible for politicians to enforce such type of purpose. And even if such channel existed, this should be extremely costly. Moreover, though the social purpose-lending hypothesis can explain the higher default rate of political loans, it cannot reconcile with other empirical results established in the data. For example, with the motive to promote sector development and social welfare, one would expect the lending to spread nationwide instead of targeting only the province from which the chairman comes. We further provide evidence to dispel the social purpose-lending hypothesis. If politicians use political lending to promote development in the sector they directly supervise, one should expect the effect to be stronger for small firms, which are more credit-constrained and have limited nonbank financing channel. As observed in Table 12, favorable lending in columns 1-8 and the ex- post performance of political loans does not vary with firm size (where the Small-Firm indicator classifies firms with less than 50 employees as small). These results are in stark contrast to the social lending hypothesis and lend further support to the political rent-seeking hypothesis. VI.D. Randomization of the Chairman Selection Finally, we perform a falsification test to further examine the robustness of the results. We examine the effects of political connections on lending by randomly assigning states with chairmen. This specification checks the validity of our political connection measure to exclude the possibility that we are establishing a spurious relation between chairmen and bank lending. Table 13 reports the results for the geographical falsification test. Specifically, instead of using the real value of whether there is a chairman for each state, we create an indicator variable and randomly assign 13                                                              33  We sorted commissions by their relative importance using a composite of three rankings: i) share of chairmen that headed their party list in the senate elections; ii) share of chairmen that had a relevant past political position; and iii) share of chairmen that will have an important political position in the future. We then averaged the three rankings and chose the top five commissions. These are Legislation, Governance, Treasury, Justice, Commerce, Constitution and Labor. As robustness check, we have also used the top 10 commissions. Results, not shown, are similar.  22    commissions per legislature to states.34 We re-estimate equation (1) based on this randomization. A positive relationship would raise the concern that the documented impact of a chairman is driven by simple spurious variation. However, the coefficients on Chairman in Table 13 are indistinguishable from zero and indicate that identification by the change in congressional committee chairman is not random variation. To validate the power of our falsification test, we do the randomization process 100 times and repeat the analysis for each sample to obtain a set of coefficients. We plot the coefficient estimates along with the 95 percent confidence intervals in Figure A3. As can be observed, the 95 percent confidence interval contains zero for most of the randomized samples, indicating that the coefficient estimate from the randomization sample is not significantly different from zero. We also conduct t-test for all coefficients to check whether the mean of the coefficient estimate differs from zero and display the result in Table A4 of Online Appendix. None of the t-statistics indicates that the coefficient estimates are statistically significant, which further supports the results in Table 13. VII. Cost of the Political Rents and Banks’ Incentive VII.A. Estimating the Cost The political rents identified can introduce a variety of costs, including wasteful activities used to build political connections and the distortion in entry and competition of firms (Khwaja and Mian, 2005). In this section, we focus on two types of costs that can be directly inferred from our data: (1) loss due to writing off the bad loans; (2) distortion in credit allocation. While precise data are not readily available, we perform a back-of-the-envelope analysis and estimate the upper and lower bound derived from our universe data set of commercial lending. Our calculation here can provide an alternative solution for estimating corruption more precisely as recent studies (i.e. Kraay and Murrell, 2015) suggested that survey-based estimation can lead to downward bias. In Section V.D., we established that banks with higher level of public loans relative to the mean are the major rent providers, so the default on the loans of these banks to politically connected firms suggests a lower bound of this estimate. Table A2 indicates that the outstanding commercial lending from these banks is 866 billion Mexican pesos (74 billion dollars) while 45 percent is extended to politically connected firms as shown in Table 1. From Table 3, we also note that the additional default rate of political loans is 0.7 percent and the recovery rate on all loans is 69 percent. 35 Therefore, the total losses from political loans are 0.85 billion pesos (866 *0.45*0.007*(1-0.69)), or roughly 73 million dollars per year. To assess the economic significance                                                              34 Even through there are 13 committees overall, in some cases a state has more than one senator heading a committee in a legislature. 35  The recovery rate of 69 cents on the dollar is calculated as the weighted average, by loan size, of the share of outstanding debt relative to the debt amount at origination. As a comparison, the recovery rate estimates from Doing Business for Mexico are of around 62 cents. 23    of the estimate, we compare this number with annual average Mexico’s GDP from 2003 to 2012 (11,148 billion MXP) and the total cost of political lending is around 0.4 basis point of GDP annually. According to an OECD report, the cost of political rents is about 5 percent of total GDP worldwide on average. Therefore, our estimate can explain 0.08 percent of Mexico’s total cost of political rents per year. In addition to the cost of writing off bad debts, excess credit supply driven by the political lending can also cause distortions of credit allocation in the economy by crowding out lending to other firms. Under the plausible assumption of fixed supply of commercial credit to the same state, firms in the chairman’s state but not the same industry that under the chairman’s purview are likely to be the “victims” and suffer from the credit under provision. Figure 1 offers a consistent lending pattern with this conjecture: in contrast to the sharp increase in the lending to the agricultural firms, the credit supply for firms in non-agricultural sectors has a slight decrease after the appointment of the chairman. In an unreported test, we examine the within-firm variation in the credit outcomes for those firms and find that the lending distortion leads to a significant decrease in credit supply of firms located in the chairman’s state but operating in an industry out of the supervision of the chairman’s committee. VII.B. Banks’ Incentive – Net Revenue Estimation A soft budget constraint enables government-owned banks to remain solvent even in the face of high levels of non-performing loans. Moreover, government-owned banks often bear the role of social welfare maximization. In contrast, private banks face harder budget constraints and adopt profit-maximizing strategies to maintain operations. In this section, we present evidence on the incentives of banks by conducting a cost-of-benefit analysis for banks that provide political rents. To estimate the benefits to the bank, we follow the quid pro quo results and focus on the additional interest income from the extra government borrowing. We are being conservative in this estimate as there may exist many other benefits that are either unobservable or hard to measure. For example, political connection can bring favorable policies and discretionary government decisions like lower regulation, weaker capital and liquidity requirements or a bailout of failing banks. Again, we focus on the subset of banks with government borrowing above the median and treat them as the major rent providers. Table 8 suggests that banks with high level of public loans receive 13 percent more government borrowing, so we are able to infer a ballpark number for additional government borrowing relative to other banks that are not engaged in such activities. Based on our estimate of cost due to writing off bad debts (in the above section), and the fact that the average interest rate on public loans is 15 percent, the annual net income is 1.1 billion pesos in total or 0.13 billion for each bank. This represents roughly 1.6 percent of the banks’ annual net income during our sample period. VIII. Conclusions 24    While there is a long tradition in the literature documenting that political rent seeking can distort the capital allocation process by government-owned banks, economists know little about how it affects the credit supply in an economy without public banks. This paper aims to fill this gap using loan-level data containing all commercial loans in Mexico, which has a fully privatized banking sector. We include a rich set of fixed-effects to absorb any factor driven by firm-specific credit demand, banking relationship and time-varying industry and bank fundamentals. Relying on changes in the committee chairmanship to identify variation in political connection, this study finds compelling evidence that loan terms offered to politically connected firms are substantially better than those available to the counterparts. At the same time, political loans exhibit much worse performance ex-post. Such political lending is mainly offered by large and domestic banks, concentrated in politically connected firms with stronger links to the politician (same industry or municipality) and in the final year of the chairman's term. Banks receive significantly more government borrowings with better credit quality as the reward for offering political rents. We also find that the excess credit supply leads to a large and significant increase in firm’s employment, assets and total liabilities. The results provide support for the view that banks use lending to politically connected firms to exchange favors with politicians, and confirm the importance of rent-seeking in influencing the lending decisions of financial intermediaries, in this case private banks. The study also provides an insight on the long-standing debate on the long-run benefit of privatization. Privatization advocates often assume that it will necessarily eliminate rent-seeking behavior through competition and crowding out political interventions. However, we still find evidence of political rent in bank lending of a privatized economy. A better way to overcome the root problem of rent seeking may be stricter regulation in terms of increased reporting requirements, scrutiny of illegal transactions, and closer supervision. 25    References Aizenman, Joshua, and Mark, Spiegel, 2006, Instutitional Efficiency, Monitoring Costs and The Investment Share of FDI, Review of International Economics, 14(4), 683-697. Agarwal, Sumit and Robert Hauswald, 2010, Distance and Private Information in Lending, Review of Financial Studies 23(7), 2757-2788. Agarwal, Sumit, Gene Amromin, Itzhak Ben-David, Souphala Chomsisengphet and Douglas Evanoff, 2011, The Role of Securitization in Mortages Renegotiation, Journal of Financial Economics 102(3), 559-578. Agarwal, Sumit, Wenlan, Qian, Amit Seru and Jian Zhang, 2015a, Disguised Corruption: Evidence from Consumer Credit in China, ABFER Working Paper Agarwal, Sumit, Tomek Piskorski, Amit Seru, Tienfoo Sing and Jian Zhang, 2015b, Information Leakage and Insider Trading in the Housing Market, SSRN Working Paper. Almeida, Heitor, Murillo Campello, Bruno Laranjeira, and Scott Weisbenner, 2012, Corporate Debt Maturity and the Real Effects of the 2007 Credit Crisis, Critical Finance Review 1, 3-58. Aparicio, Francisco Javier, and Joy Langston. Committee leadership selection without seniority: The Mexican case. Centro de Investigación y Docencia Económicas (CIDE), División de Estudios Políticos, 2009. Ballinas, Victor and Andrea, Becerril, 2012, Integration of 59 Ordinary Commission in the Senate, Politica. Barth, James R., Chen Lin, Ping Lin, and Frank M. Song, 2009, Corruption in Bank Lending to Firms: Cross-country Micro Evidence on the Beneficial Role of Competition and Information Sharing, Journal of Financial Economics 91, 361-388. Baharth, Sreedhar, Sandeep Dahiya, Anthony, Saunders and Anand Srinivasan, 2011, Lending Relationships and Loan Contract Terms, Review of Financial Studies 24, 1141-1203. Barberis, Nicholas, Maxim Boycko, Andrei Shleifer, and Natalia Tsukanova, 1996, How does Privatization Work? Evidence from the Russian Shops. Journal of Political Economy 104(4), 764-790. Battaglini, Marco, and Stephen, Coate, 2008, A Dynamic Thoery of Public Spending, Taxation and Debt. American Economic Review 98(1), 201-236. Beck, Thorsten, Aslı Demirgüç-Kunt, and Ross Levine, 2006, Bank Supervision and Corruption in Lending, Journal of Monetary Economics 53, 2131-2163. Beck, Thorsten, Robert Cull, and Afeikhena Jerome, 2005, Bank Privatization and Performance: Empirical Evidence from Nigeria, Journal of Banking and Finance 29, 2355-2379. Berger, N. Allen, Philip Molyneux, and John O.S. Wilson, 2014, The Oxford Handbook of Banking Oxford: Oxford University Press. 26    Bernanke, Ben, 1983, Nonmonetary Effects of the Financial Crisis in the Propagation of the Great Depression. American Economic Review 73 (3), 257-276. Bernile, Gennardo, Jianfeng Hu, and Yuehua Tang, 2015, Can Information be Locked Up? Informed Trading Ahead of Macro-news Announcement. Journal of Financial Economics forthcoming. Bertrand, Marianne, Francis Kramarz, Antoinette Schoar, and David Thesmar, 2008, Politicians, Firms and the Political Business Cycle: Evidence from France, University of Chicago Working Paper. Boycko, Maxim, Andrei Shleifer, and Robert W. Vishny, 1993, Privatizing Russia, Brookings Papers on Ecomomic Activities No 3. Caballero, Ricardo, J. and Pierrer Yard, 2010, Future Rent-seeking and Current Public Saving, Journal of International Economics 82(2), 124-136. Campante, Filipe R. and Quoc-Anh Do, 2014, Isolated Capital Cities, Accountability and Corruption: Evidence from US States, American Economic Review 104(8), 2456-81. Carvalho, Daniel R., 2014, The Real Effects of Government Owned Banks: Evience from An Emerging Market, Journal of Finance 69, 577-609. Claessens, Stijn, Erik Feijen, and Luc Laeven, 2008, Political Connections and Preferential Access to Finance:the Role of Campaign Contributions. Journal of Financial Economics 88, 554-580. Chodorow-Reich, Gabriel, 2011, The Employment Effects of Credit Market Disruptions: Firm-Level Evidence from the 2008-9 Financial Crisis, Quarterly Journal of Economics, 129(1), 1-59. Cohen, Lauren, Joshua Coval, and Christopher Malloy, 2011, Do Powerful Politicians Cause Corporate Downsizing?, Journal of Political Economy 119, 1015-1060. Cole, Shawn, 2009, Fixing Market Failures or Fixing Elections? Election, Banks and Agricultural Lending in India. American Economic Journal:Applied Economics 1(1), 219-250. Degeorge, Francois, Dirk Jenter, Alberto Moel, and Peter Tufano, 2004, Selling Company Shares to Reluctant Employees: France Telecom's Experience, Journal of Financial Economics 77, 169-202. Dinc, Serdar, 2005, Politicians and Banks: Political Influences on Government-Owned Banks, Journal of Financial Economics 77, 453-479. Faccio, Mara, 2006, Politically Connected Firms, American Economic Review 96(1), 369-386. Fang, Hanming, Quanlin Gu and Li-An Zhou, 2014, The Gradient of Power: Evidence from the Chinese Housing Market, NBER working paper No.20317. Fisman, Raymond, 2001, Estimating the Value of Political Connections, American Economic Review 1095- 1102. Fisman, Raymond, and Yongxiang Wang, 2015, The Mortality Cost of Political Connections, Review of Economic Studies, forthcoming 27    Gan, Jie, 2007, The Real Effects of Asset Market Bublles: Loan- and Firm-Level Evidence of a Lending Channel, Review of Financial Studies 20, 1941-73. Gertler, Mark, and Simon Gilchrist, 1994, Monetary Policy, Business Cycles and the Behavior of Small Manufacturing Firms, Quarter Journal of Economics 109, 309-340. Gross, David, B. and Nicholas,S., Souleles, 2002, An Empirical Analysis of Personal Bankruptcy and Delinquency, Review of Financial Studies 15, 319-347. Holder, Roland and Paul, Raschky, 2014, Regional Favortitism, Quaterly journal of economics 129:995- 1033. Houston, Joel, Liangliang Jiang, Chen Lin and Yue Ma, 2014, Political Connections and the Cost of Bank Loans, Journal of Accounting and Economics 52(1): 193-243. Kasara, Kimuli, 2007, Tax Me if You Can: Ethnic Geography, Democracy, and the Taxation of Agriculture in Africa, American Political Science Review 101, 159-172. Keys, Benjamin J., Tanmoy Mukherjee, Amit Seru, and Vikrant Vig, 2010, Did Securitization Lead to Lax Screening? Evidence from Subprime Loans, Quaterly Journal of Economics 125(1), 307-362. Khwaja, Asim Ijaz and Mian, Atif, 2005, Do Lenders Favor Politically Connected Firms? Rent Provision In an Emerging Financial Market, Quaterly Journal of Economics 1371-1411. ---, 2008, Tracing the Impact of Bank Liquidity Shock. American Economic Review 98,1413-1442. Kikeri, Sunita, John Nellis, and Mary Shirley, 1994, Privatization: the Lessons from the Market Economies. The World Bank Research Observer 9,241-271. Kostovetsky, Leonard, 2015, Political capital and moral hazard, Journal of Financial Economics 116, 144- 159. Kraay, Aart, and Peter Murrell, 2015, Misunderestimating Corruption, Review of Economics and Statistics forthcoming. La Porta, Rafael, and Florencio López-de-Silanes, 2003, The Benefits of Privatization: Evidence from Mexico, Quarterly Journal of Economics 114(4), 1193-1242. La Porta, Rafael, Florencio López-de-Silanes, and Guillermo Zamarripa, 2003, Related Lending, Quarterly Journal of Economics 118, 231-268. Martin, Kelly, John, Cullen, Jean Johnson, and Praveen Parboteeah, 2007, Deciding to Bribe: a Cross-Level Analysis of Firm and Home Country Influences on Bribery Activity, Academy of Management Journal, 50(6),1401-1422. McMillan, John, and Pablo Zoido, 2004, How to Subvert Democracy: Montesinos in Peru, Journal of Economic Perspectives, 18(4), 69-92. 28    Meggison, L. William, 2011, Privatization and Finance, Annals Review of Financial Economics 2, 145- 174. Mian, Atif, Amir Sufi, and Francesco Trebbia, 2010, The Political Economy of the US Mortgage Default Crisis, The American Economic Review 100, 1967-1998. Mian, Atif, Amir Sufi, and Francesco Trebbia, 2013, The Political Economy of the Subprime Mortgage Credit Expansion, Quarterly Journal of Political Science 8, 373-408. Moe, Roland, 1996, Managing Privatization: A New Challenge to Public Administration, In Agenda for Excellence 2: Administering the State, edited by B.G. Peters and B.A. Rockman, 135-148. Chatham, NJ: Chatham House. Montero, Mercedes García, and Francisco Sánchez López. Las comisiones legislativas en América Latina: una clasificación institucional y empírica. Vol. 212. Institut de Ciències Polítiques i Socials, 2002. Morales, Alberto. “Congreso lidera en Comisiones: Integralia”. El Universal. August 31, 2015. Olken, Benjamin A, 2007, Monitoring Corruption: Evidence from a Filed Experimen in Indonesia, Journal of Political Economy 115, 200-249. Olken, Benjamin A., and Patrick Barron, 2009, The Simple Economics of Extortion: Evidence from Trucking in Aceh, Journal of Political Economy 117(3), 417-452. Olken, Benjamin A, and Rohini Pande, 2011, Corruption in Developing Countries, Annal Review of Economics 4, 479-505. Petersen, Mitchell A, and Raghuram G. Rajan, 2002, Does Distance Still Matter? The Information Revolution in Business Lending, Journal of Finance 57(4), 2533-2570. Reinikka, Ritva and Jakob, Svensson, 2004, Local Capture: Evidence from a Central Government Transfer Program in Uganda Quaterly Journal of Economics 119(2), 678-705. Rendón Arias, Iliana. "Las prácticas parlamentarias en el congreso mexicano: los casos de las comisiones de hacienda y de gobernación durante el primer año de la LXII legislatura (2012-2013)." (2014). Roland, Gerard, 2008, Privatization:Success and Failtures New York: Columbia University Press. Sapienza, Paola, 2004, The Effects of Government Ownership on Bank Lending, Journal of Financial Economics 72, 357-384. Sequeira, Sandra, and Simeon Djankov, 2014, Corruption and Firm Behavior: Evidence from African Ports, Journal of International Economics 94(2), 277-329. Shleifer, Andrei and Robert, Vishny, 1994, Politicians and Firms, Quarterly Journal of Economics 109,995- 1025. Sun, Qian, and Wilson H.S Tong, 2003, China Share Issue Privatization, Journal of Financial Economics 70(2), 183-222. 29    Wu, Xun, 2005, Corporate Governance and Corruption: A Cross-Country Analysis, Governance: An International Journal of Policy, Administration and Institutions, 18 (2). 151-170. Yared, Pierrer, 2010, Politicians, Taxes and Debt, Review of Economic Studies 77(2), 806-840. 30    Appendix A: Variable Definitions Dependent Variables Loan Volume is the amount of outstanding bank credit of a firm-bank pair (thousands of Mexican pesos). Loan Spread is the difference between the average annualized loan rate of a firm-bank pair, weighted by loan volume, and the average annual cost of funds of the bank. Maturity is the average loan duration (in months) at origination of a firm-bank pair, weighted by loan volume. Collateral is the average fraction of loans of a firm-bank pair that is covered by the firm’s assets, weighted by loan volume. D(Collateral) is a dummy variable that equals one if collateral is used. Default Rate is the average share of bank loans, weighted by loan volume, of a firm-bank pair that is more than 90 days in arrears within the 36 months after origination. D(Default) is a dummy variable that equals one if the loan is more than 90 days in arrears. Employment is the yearly number of employees of a firm. Assets are the total assets of the firm in a year (in thousands Mexican pesos). Liabilities are the total liabilities of the firm in a year (in thousands Mexican pesos). Revenue is the total revenue of the firm in a year (in thousands Mexican pesos). Independent Variables Chairman is a dummy variable that equals one if the state where the firm is headquartered has a senator serving as chairman. SameInd is a dummy variable that equals one if the firm operates in an industry that is the focus of a commission (e.g. Energy). SameMuni is a dummy variable that equals one if the firm is headquartered in the same municipality as the chairman. LargeBank is a dummy variable that equals one if the lending bank has more than 150 billion pesos on assets, in 2011 prices, on average during our sample period. DomBank is a dummy variable that equals one if the bank is domestic. LargePubLoan is a dummy variable that equals one if the bank extends a higher level of loans to public firms (as a share of total loans) relative to the median bank during our sample period. Please refer to third column of Table A2 for the list of banks and share of public loans. LargeFirm is a dummy variable that equals one if the firm has more than 50 employees. Chairman_Yeari is a dummy variable that equals one if the loan was extended during the i-th year of a legislative session. 31      Figure 1 - Lending in Colima versus Rest of Mexico Loan Volume 2.4 2.2 2 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 -4 -3 -2 -1 0 1 2 3 4 Quarters since chairmanship appointment Colima Agri Colima NonAgri All Agri All NonAgri The figure compares the total lending to the agriculture sector in the state of Colima with the total lending to the other sectors in Colima and to the other states in Mexico excluding Colima, around the time when the senator from Colima was appointed chairman of the agricultural committee. The horizontal axis presents the length in quarters since the election and the vertical axis shows loan volume normalized to 1 in the quarter of the chairmanship appointment. 32    Figure 2 - Dynamics of Political Lending through the Election Cycle Loan Volume 3 2.5 Vol. in states with minus state without(%) 2 1.5 1 0.5 0 0 2 4 6 8 10 12 Quarters since chairmanship appointment The figure shows the volume of loans at the bank-state level in states with chairman (relative to those without) through the election cycle. The horizontal axis presents the length in quarters since the election for the chairman and the vertical axis shows the difference in the average loan volume in states with and without chairman relative to their overall mean.       33    Figure 3 - Economic Developments and Chairmanship Appointment  Annual GDP Growth 0.06 0.04 0.02 0 -0.02 -0.04 -0.06 -0.08 2004 2005 2006 2007 2008 2009 2010 2011 Year Annual Employment Growth 0.04 0.035 0.03 0.025 0.02 0.015 0.01 0.005 2004 2005 2006 2007 2008 2009 2010 2011 Year States with Chairman States without Chairman The figure compares the annual GDP and employment growth rate for states with and without a senator serving as chairman.   34    Table 1 - Summary Statistics This table presents the loan-level summary statistics for the data used in this paper. Panel A reports the composition of the data according to the loan type while Panel B compares the sample mean and difference of loan terms for political and non-political loans. SameInd is a dummy variable indicating whether the firm operates in an industry that is the focus of a committee (e.g. energy) and SameMuni is a dummy variable that equals one if the firm is headquartered in the same municipality as a senator. Please refer to Appendix A for the variable definitions. *, **, *** indicate significance at the 10%, 5% and 1% levels. Panel A Composition by Loan Type N Mean SD Chairman 594,534 0.45 0.5 Chairman*SameInd 594,534 0.05 0.24 Chairman*SameMuni 594,534 0.06 0.27 Large Bank 594,534 0.81 0.48 Domestic Bank 594,534 0.35 0.48 Large Firm 594,534 0.19 0.39 Panel B Loan Characteristics Difference All loans Loans with chairman Loans without chairman (t-stats) (1) (2) (3) (4) Loan Volume (in ‘000MXP) 2,776 2,857 2,708 149*** (-11.86) Loan Spread (%) 12.1 11.6 12.3 -0.67*** (2.98) Maturity (in months) 15.6 15.6 15.5 0.09** (2.33) D(Collateral) 0.36 0.33 0.38 -0.05*** (-39.94) D(Default) (%) 6.30 6.7 6.0 0.70*** (9.80) 35    Table 2 - Political Connection and Loan Terms The table reports the coefficients of ordinary least squares regressions of loan terms (volume, spread, maturity and collateral) on an indicator variable, Chairman, which equals to one if the firm is headquartered in the state that has a senator serving as committee chairman. The sample period spans from October 2003 to March 2012. A unit of observation is a loan at the firm-bank-quarter level. The dependent variables (ordered by column) are the firm's loan volume (in logarithm multiplied by 100), loan spread (in percent), original maturity (in months) and collateral requirement dummy. Please refer to Appendix A for variable definitions. The yearly state-level GDP growth and unemployment rate are included as controls in all specifications. All regressions include firm*bank and bank*quarter fixed-effects. Standard errors are reported in parentheses and are clustered at the state level. *, **, *** indicate significance at the 10%, 5% and 1% levels. Loan Volume Loan spread Maturity D (Collateral) (1) (2) (3) (4) (5) (6) (7) (8) Chairman 0.762** 0.720** -0.019* -0.020* 0.106*** 0.091*** -0.021*** -0.018*** (0.285) (0.287) (0.011) (0.011) (0.036) (0.033) (0.007) (0.006) GDP growth rate 13.723* 13.686* -0.250 -0.233 1.961** 1.782** -0.176* -0.164* (7.240) (7.099) (0.239) (0.234) (0.738) (0.735) (0.095) (0.087) Unemployment rate 0.083 0.067 -0.009*** -0.009*** 0.021 0.019 -0.002 -0.002 (0.116) (0.118) (0.003) (0.003) (0.017) (0.016) (0.002) (0.002) Bank*Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Bank*Quarter FE Yes Yes Yes Yes Yes Yes Yes Yes Industry*Quarter FE No Yes No Yes No Yes No Yes R-Square 0.021 0.021 0.414 0.415 0.201 0.202 0.368 0.372 N 594,534 594,534 594,534 594,534 594,534 594,534 594,534 594,534 36    Table 3 - Political Connection and Ex Post Loan Performance The table studies the ex post performance for political and non-political loans from October 2003 to March 2012. The dependent variable, D(Default), is a dummy variable that equals one if the loan extended by a bank to a firm in a given quarter ever becomes more than 90 days in arrears. Chairman is a dummy indicator that equals one if the firm is headquartered in the state that has a senator serving as committee chairman. The state-level GDP growth and unemployment rate are included as controls in all specifications. All regressions include firm*bank and bank*quarter fixed-effects. Standard errors are reported in parentheses and are clustered at the state level. *, **, *** indicate significance at the 10 %, 5% and 1 % levels. D (Default) (1) (2) Chairman 0.006*** 0.007*** (0.002) (0.002) GDP growth rate -0.058 -0.054 (0.042) (0.041) Unemployment rate 0.002** 0.002** (0.001) (0.001) Bank*Firm Fixed-effects Yes Yes Bank*Quarter Fixed-effects Yes Yes Industry*Quarter Fixed-effects No Yes R-Square 0.023 0.025 N 480,354 480,354 37    Table 4 - Preferential Treatment and Political Connection Strength The table investigates the extent to which the political connection strength affects the favorable treatment of politically connected firms by banks. Panels A and B show the results using two proxies for connection strength respectively: being in the same industry that is under the chairman’s supervision and being headquarted in the same municipality as the chairman. A unit of observation is a loan at the firm-bank-quarter level. The dependent variables are loan terms at origination and ex post performance respectively. Please refer to Appendix A for variable definitions. The state- level yearly GDP growth and unemployment rate are included as controls in all specifications. All regressions include firm*bank and bank*quarter fixed-effects. Standard errors are reported in parentheses and are clustered at the state level. *, **, *** indicate significance at the 10%, 5% and 1% levels, respectively. Loan Volume Loan spread Maturity D(Collateral) D(Default) Panel A Same Industry (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Chairman 0.386 0.345 -0.015 -0.016 0.118*** 0.099** -0.023*** -0.020*** 0.008*** 0.008*** (0.280) (0.271) (0.012) (0.012) (0.040) (0.038) (0.007) (0.006) (0.002) (0.002) Chairman*SameInd 2.341*** 2.414*** -0.023 -0.022 -0.069 -0.051 0.012 0.013* 0.013*** 0.010*** (0.232) (0.220) (0.018) (0.019) (0.045) (0.049) (0.009) (0.007) (0.001) (0.002) Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Bank*Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Bank*Quarter FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Industry*Quarter FE No Yes No Yes No Yes No Yes No Yes R-Square 0.021 0.021 0.414 0.415 0.201 0.202 0.368 0.372 0.021 0.023 N 594,534 594,534 594,534 594,534 594,534 594,534 594,534 594,534 480,354 480,354 Panel B Same Municipality (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Chairman 0.700*** 0.658** -0.019* -0.019* 0.108*** 0.092** -0.023*** -0.020*** 0.006*** 0.007*** (0.239) (0.245) (0.011) (0.012) (0.039) (0.035) (0.006) (0.005) (0.002) (0.002) Chairman* SameMuni 1.544* 1.568* -0.012 -0.010 -0.034 -0.014 -0.045** -0.044** 0.021*** 0.019*** (0.925) (0.923) (0.019) (0.018) (0.128) (0.120) (0.019) (0.016) (0.009) (0.008) Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Bank*Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Bank*Quarter FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Industry*Quarter FE No Yes No Yes No Yes No Yes No Yes R-Square 0.021 0.021 0.414 0.415 0.201 0.202 0.368 0.372 0.021 0.022 N 594,534 594,534 594,534 594,534 594,534 594,534 594,534 594,534 480,354 480,354 38    Table 5 - Preferential Treatment and Bank Size The table investigates whether the preferential treatment to politically connected firms depends on bank size. A unit of observation is a loan at the firm-bank-quarter level. The dependent variables in columns 1 to 8 are loan terms at origination. Ex post performance is presented in columns 9 and 10. LargeBank is a dummy variable that equals one if the lending bank has more than 150 billion pesos in assets, in 2011 dollars, on average during our sample period. Please refer to Appendix A for variable definitions. The state-level yearly GDP growth and unemployment rate are included as controls in all specifications. All regressions include firm*bank and bank*quarter fixed-effects. Standard errors are reported in parentheses and are clustered at the state level. *, **, *** indicate significance at the 10%, 5% and 1% levels, respectively. Loan Volume Loan Spread Maturity D(Collateral) D(Default) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Chairman 0.868*** 0.768*** -0.015* -0.015* 0.083*** 0.069*** -0.019*** -0.016*** 0.007** 0.008*** (0.253) (0.193) (0.008) (0.008) (0.025) (0.021) (0.007) (0.006) (0.003) (0.003) Chairman*LargeBank 1.237** 1.155* -0.006 -0.002 0.134** 0.127** -0.016* -0.016* 0.007** 0.006* (0.601) (0.603) (0.022) (0.021) (0.053) (0.056) (0.010) (0.009) (0.003) (0.004) Bank*Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Bank*Quarter FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Industry*Quarter FE No Yes No Yes No Yes No Yes No Yes R-Square 0.024 0.025 0.416 0.417 0.206 0.207 0.445 0.449 0.020 0.022 N 594,534 594,534 594,534 594,534 594,534 594,534 594,534 594,534 480,354 480,354 39    Table 6 - Preferential Treatment and Bank Ownership: Domestic versus Foreign Bank The table investigates whether the preferential treatment to politically connected firms differs between domestic and foreign banks. A unit of observation is a loan at the firm-bank-quarter level. The dependent variables in Panels A and B are loan terms at origination and ex-post performance respectively. DomBank indicates whether the loan is extended by a domestic bank. Please refer to Appendix A for variable definitions. The state- level yearly GDP growth and unemployment rate are included as controls in all specifications. All regressions include firm*bank and bank*quarter fixed-effects. Standard errors are reported in parentheses and are clustered at the state level. *, **, *** indicate significance at the 10%, 5% and 1% levels, respectively. Loan Volume Loan Spread Maturity D(Collateral) D(Default) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Chairman 0.761*** 0.721*** -0.019* -0.020* 0.106*** 0.091*** -0.021*** -0.018*** 0.005** 0.006*** (0.254) (0.259) (0.011) (0.011) (0.037) (0.033) (0.007) (0.006) (0.002) (0.002) Chairman*DomBank 1.083* 1.081* 0.007 0.006 0.033 0.029 0.000 0.001 0.014*** 0.013*** (0.622) (0.609) (0.022) (0.022) (0.062) (0.062) (0.008) (0.008) (0.003) (0.003) Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Bank*Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Bank*Quarter FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Industry*Quarter FE No Yes No Yes No Yes No Yes No Yes R-Square 0.021 0.021 0.414 0.415 0.201 0.202 0.368 0.372 0.021 0.022 N 594,534 594,534 594,534 594,534 594,534 594,534 594,534 594,534 480,354 480,354 40    Table 7 - Political Lending through the Election Cycle The table examines how political lending varies along the election cycle. A unit of observation is a loan at the firm-bank-quarter level. The dependent variables in Panels A and B are loan terms at origination and ex post performance respectively. Chairman_Year1, Chairman_Year2 and Chairman_Year3 are all dummy variables that equal one for the first, second and third year after the start of the legislative session, respectively. Please refer to Appendix A for variable definitions. The state-level GDP growth and unemployment rate are included as controls in all specifications. All regressions include firm*bank and bank*quarter fixed-effects. Standard errors are reported in parentheses and are clustered at the state level. *, **, *** indicate significance at the 10%, 5% and 1% levels, respectively. Loan Volume Loan Spread Maturity D(Collateral) D(Default) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Chairman_Year1 0.199 0.231 0.004 0.002 0.138*** 0.127*** -0.017** -0.014** 0.008** 0.008*** (0.355) (0.350) (0.020) (0.022) (0.044) (0.040) (0.007) (0.007) (0.003) (0.003) Chairman_Year2 0.231 0.198 -0.025 -0.026 0.173*** 0.156*** -0.024*** -0.021*** 0.006** 0.006*** (0.367) (0.372) (0.016) (0.017) (0.039) (0.037) (0.007) (0.007) (0.002) (0.002) Chairman_Year3 1.721*** 1.628*** -0.034* -0.033* 0.017* 0.000 -0.022*** -0.019*** 0.006** 0.006** (0.498) (0.488) (0.020) (0.020) (0.047) (0.045) (0.007) (0.006) (0.003) (0.002) Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Bank*Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Bank*Quarter FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Industry*Quarter FE No Yes No Yes No Yes No Yes No Yes R-Square 0.021 0.021 0.414 0.415 0.368 0.372 0.073 0.074 0.201 0.203 N 594,534 594,534 594,534 594,534 594,534 594,534 594,534 594,534 480,354 480,354 41    Table 8 - Quid Pro Quo: Rewards to the Bank The table investigates whether banks who extend more favorable loans to politically connected firms also extend more loans to public firms as a reward. The dependent variables in Panels A and B are loan terms at origination and ex post performance for public loans respectively. LargePubLoan (SmallPubLoan) is a dummy variable that equals one if the bank receives high (low) level of public loans relative to the median. In Panel A, we examine political lending made by banks that extend a relatively high level of public loans. In Panel B, we focus on the public loans lent by type of bank, where a unit of observation is a loan at the state-banktype(LargePubLoan/SmallPubLoan)-quarter level. Please refer to Appendix A for variable definitions. State-level GDP growth and unemployment rate are included as controls in all specifications. All regressions include state*banktype fixed-effects. Standard errors are reported in parentheses and are clustered at the state level. *, **, *** indicate significance at the 10%, 5% and 1% levels, respectively. Loan Volume Loan Spread Maturity D(Collateral) D(Default) Panel A Lending by Banks to Private Firms (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Chairman 0.762** 0.719** -0.019* -0.020* 0.106*** 0.091*** -0.021*** -0.018*** 0.007*** 0.007*** (0.290) (0.290) (0.011) (0.012) (0.036) (0.033) (0.007) (0.006) (0.002) (0.002) Chairman*LargePubLoan 0.658* 0.639* 0.052 0.053 -0.005 -0.010 -0.036*** -0.036*** 0.009*** 0.008** (0.398) (0.379) (0.032) (0.033) (0.067) (0.066) (0.011) (0.010) (0.003) (0.003) Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Bank*Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Bank*Quarter FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Industry*Quarter FE No Yes No Yes No Yes No Yes No Yes R-Square 0.021 0.021 0.414 0.415 0.201 0.202 0.368 0.372 0.021 0.022 N 594,534 594,534 594,534 594,534 594,534 594,534 594,534 594,534 480,354 480,354 Panel B Public Loans (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Chairman*SmallPubLoan -0.065 -0.030 -0.107 -0.062 0.284 0.481* 0.004 0.015* -0.008 -0.007 (0.109) (0.100) (0.139) (0.083) (0.426) (0.286) (0.017) (0.008) (0.005) (0.005) Chairman*LargePubLoan 0.149* 0.132* -0.101 -0.060 0.242 0.487* 0.004 0.008 -0.009* -0.008* (0.091) (0.079) (0.135) (0.081) (0.423) (0.290) (0.017) (0.009) (0.005) (0.005) Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes State*BankType FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Quarter FE No Yes No Yes No Yes No Yes No Yes R-Square 0.768 0.815 0.122 0.694 0.305 0.668 0.304 0.845 0.255 0.422 N 2,181 2,181 2,181 2,181 2,181 2,181 2,181 2,181 2,181 2,181 42    Table 9 - Real Effect of the Political Lending The table examines whether political lending has real impacts on the economic activity of firms. The unit of observation is at the firm-year level. Columns 1, 3, 5 and 7 report the estimates for the full sample while C2, 4, 6 and 8 present the results for small firms (firms with less than 50 employees). The dependent variables are in logs and include number of employees, total assets, total liabilities and revenue. Chairman equals to one if the firm is headquartered in the state that has a senator serving as committee chairman. Please refer to Appendix A for variable definitions. All regressions include firm and year fixed-effects. Standard errors are reported in parentheses and are clustered at the state level. *, **, *** indicate significance at the 10%, 5% and 1% levels, respectively. Employment Assets Liabilities Revenue (1) (2) (3) (4) (5) (6) (7) (8) All Small All Small All Small All Small Chairman 0.021*** 0.025* 0.013* 0.027** 0.010 0.024* 0.008 0.008 (0.007) (0.015) (0.008) (0.010) (0.012) (0.014) (0.011) (0.010) Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Yes R-Square 0.037 0.041 0.073 0.068 0.055 0.043 0.038 0.040 N 13,004 6,635 11,488 6,006 11,408 5,974 21,022 10,726 43    Table 10 - Chairman Appointment and Sector Activities This table investigates whether the selection of chairman is endogenously determined by the economic development of his state of origin. Panel A compares sector-level productivity (both in level and growth rate) in five sectors (Agriculture, Commerce, Transport, Energy and Telecommunications) for states with and without a senator serving as chairman in the corresponding sector in the year when committees are formed (2003, 2006 and 2009). We use indicators (by row order) to proxy for productivity of the sectors reported for a given state and year. The indicators are obtained from the Banco de Informacion of the National Institute of Statistics and Geography (INEGI). Panel B displays a similar comparison by focusing on two important macroeconomic indicators at the state level: growth rates of GDP and of employment. Columns 1 and 2 report the mean of each sector's productivity indicator for states with and without at least one senator presiding the corresponding committee while column 3 shows the mean difference between the two groups with the t-statistics reported in the underlying parentheses. (1) (2) (3) Difference States with Chairman States without (t-statistics) Panel A Productivity of Sectors Level Total agricultural production 0.064 0.046 0.018 (relative to state's GDP) (0.83) Gross production of commerce sector 0.124 0.125 -0.001 (relative to state's GDP) (0.06) Length of highway network 0.114 0.069 -0.045 (relative to state's GDP) (1.17) Registered vehicles in circulation 0.227 0.155 0.072 (relative to state's population size) (1.14) Users of electric energy 0.296 0.294 0.002 (relative to state's population size) (0.05) Growth Rate Total agricultural production -0.040 0.069 -0.109 (relative to state's GDP) (-1.10) Gross production of commerce sector 0.076 0.078 -0.002 (relative to state's GDP) (-0.05) Length of highway network -0.028 0.026 -0.054 (relative to state's GDP) (-0.69) Registered vehicles in circulation -0.104 0.079 -0.183 (relative to state's population size) (-2.41) Users of electric energy 0.017 -0.004 0.020 (relative to state's population size) (0.20) Panel B Economic Indicator GDP Growth -0.013 0.006 -0.018 (-1.16) Employment growth 0.019 0.020 -0.001 (-0.24) 44    Table 11 - The Gradient of Political Power The table examines whether potentially stronger measures of political power result in higher political lending to connected firms. A unit of observation is a loan at the firm-bank-quarter level. The dependent variables in Panels A and B are loan terms at origination and ex post performance respectively. MoreThanOne is a dummy variable that equals one if the state where the firm is located has more than one senator chairing committees. The state-level GDP growth and unemployment rate are included as controls in all specifications. Please refer to Appendix A for variable definitions. All regressions include firm*bank and bank*quarter fixed-effects. Standard errors are reported in parentheses and are clustered at the state level. *, **, *** indicate significance at the 10%, 5% and 1% levels, respectively. Loan Volume Loan Spread Maturity D(Collateral) D(Default) Panel A More than One Chairman (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Chairman 0.742** 0.701** -0.018 -0.019 0.105*** 0.089** -0.022*** -0.019*** 0.006*** 0.007*** (0.294) (0.296) (0.011) (0.012) (0.037) (0.034) (0.007) (0.006) (0.002) (0.002) Chairman*MoreThanOne 1.328* 1.237* -0.043 -0.041 0.090 0.154 0.007 0.009 0.022** 0.020** (0.697) (0.724) (0.055) (0.056) (0.234) (0.202) (0.028) (0.025) (0.008) (0.009) Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Bank*Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Bank*Quarter FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Industry*Quarter FE No Yes No Yes No Yes No Yes No Yes R-Square 0.021 0.021 0.414 0.415 0.201 0.202 0.368 0.372 0.021 0.022 N 594,534 594,534 594,534 594,534 594,534 594,534 594,534 594,534 480,354 480,354 Panel B Alternative Definition of Core Committees (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Chairman 0.845*** 0.803** -0.022* -0.022* 0.108*** 0.093*** -0.021*** -0.018*** 0.007*** 0.007*** (0.291) (0.293) (0.011) (0.011) (0.037) (0.033) (0.007) (0.006) (0.002) (0.002) Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Bank*Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Bank*Quarter FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Industry*Quarter FE No Yes No Yes No Yes No Yes No Yes R-Square 0.021 0.021 0.414 0.415 0.201 0.202 0.368 0.372 0.021 0.022 N 594,534 594,534 594,534 594,534 594,534 594,534 594,534 594,534 480,354 480,354 45    Table 12 - Social Lending Hypothesis The table presents a test for the social lending hypothesis and investigates whether favorable treatment is related to borrower size. A unit of observation is a loan at the firm-bank-quarter level. The dependent variables in Panels A and B are loan terms at origination and ex post performance respectively. SmallFirm is a dummy variable that equals one if the firm has less than 50 employees. Please refer to Appendix A for variable definitions. All regressions include firm*bank and bank*quarter fixed-effects. Standard errors are reported in parentheses and are clustered at the state level. *, **, *** indicate significance at the 10%, 5% and 1% levels, respectively. Loan Volume Loan Spread Maturity D(Collateral) D(Default) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Chairman 0.763** 0.721** -0.019* -0.019* 0.105*** 0.090** -0.021*** -0.018** 0.006*** 0.007*** (0.281) (0.284) (0.011) (0.011) (0.038) (0.034) (0.007) (0.007) (0.002) (0.002) Chairman* SmallFirm -0.037 -0.084 -0.021 -0.019 0.146 0.126 -0.010 -0.008 -0.005 -0.005 (0.967) (0.927) (0.027) (0.026) (0.195) (0.184) (0.086) (0.083) (0.011) (0.011) Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Bank*Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Bank*Quarter FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Industry*Quarter FE No Yes No Yes No Yes No Yes No Yes R-Square 0.021 0.021 0.414 0.415 0.201 0.202 0.368 0.372 0.021 0.022 N 594,534 594,534 594,534 594,534 594,534 594,534 594,534 594,534 480,354 480,354 46    Table 13 - Robustness Test: Geographic Randomization The table presents results for the falsification test in Section VI where we randomly assign a chairman for each state. A unit of observation is a loan at the firm-bank-quarter level. The dependent variables include loan terms at origination and ex post performance. Please refer to Appendix A for variable definitions. All regressions include firm*bank and bank*quarter fixed-effects. Standard errors are reported in parentheses and are clustered at the state level. *, **, *** indicate significance at the 10%, 5% and 1% levels, respectively. Loan Volume Loan Spread Maturity D(Collateral) D(Default) (1) (2) (3) (4) (5) Chairman 0.008 -0.004 0.061 0.003 0.001   (0.306) (0.008) (0.051) (0.005) (0.003) Bank*Firm FE Yes Yes Yes Yes Yes Bank*Quarter FE Yes Yes Yes Yes Yes Industry*Quarter FE Yes Yes Yes Yes Yes R-Square 0.024 0.416 0.206 0.445 0.020 N 594,534 594,534 594,534 594,534 480,354 47    48    Online Appendix Table A1 - List of Senate Committees The table displays the list of relevant committees studied in this paper. Column 2 presents a brief description of their main objectives, while the states and Legislature term for the chairman are indicated in columns 3 and 4. For each committee, the last column lists the sectors that are directly supervised according to 2007 North American Industry Classification System (NAICS) classification. Committee Objective Legislature State of Chairman NAICS LIX Tlaxcala Supervises the senate Administration LX Tlaxcala All budget. LXI Chiapas LIX Hidalgo Development of 111.112, Agriculture LX Colima agricultural sector 11,511,152 LXI Colima LIX Coahuila 21,222,123 Promote national and Commerce LX Nuevo Leon 31-33 international trade LXI Nuevo Leon 43,46,52 LIX Puebla Reforms and additions to Constitution LX Quintana Roo All the constitution LXI Puebla LIX Campeche 21,22,2371 Supervise and develop the Energy LX Sinaloa 324,325 energy sector LXI Sinaloa 486,148,624,869 LIX Jalisco Legislates on matters Governance LX Hidalgo All related with transparency LXI Hidalgo LIX Michoacán Coordinate initiatives of Legislation LX Coahuila All other committees LXI San Luis Potosi LIX Colima Coordinate initiatives of Legislation I LX Baja California All other committees LXI Colima LIX DF Coordinate initiatives of Legislation II LX Zacatecas All other committees LXI Michoacán LIX Coahuila Supervise the judicial Justice LX Baja California All branch LXI Baja California LIX - Supervise and disseminate Telecommunications LX Colima 512,515 information LXI Zacatecas LIX - Develop transportation 2373.2379 Transport LX Guerrero and telecommunications 48,49,51 LXI Baja California LIX DF Treasury Supervise public spending LX Zacatecas All LXI Zacatecas 49            Table A2 - List of Banks The table displays the list of banks in this paper sorted by average assets from 2003-2012 in constant 2011 prices. Column 2 presents total assets in millions of pesos. The first 7 banks have on average assets above 150 billion pesos (roughly 13 billion dollars) and thus, are considered large. Column 3 displays the average value of outstanding commercial loans in millions of pesos. Column 4 lists the share of public loans over total loans. The last column indicates whether the bank is a domestic bank or not. Bank Assets Commercial Loans Shr. Public Loans (%) Domestic B1 1,086,784 190,218 23.8 No B2 1,028,502 131,049 14.0 No B3 663,033 112,146 18.2 No B4 530,906 95,201 21.3 Yes B5 419,654 72,051 23.6 No B6 205,239 121,882 5.9 Yes B7 165,414 32,178 20.8 No B8 75,505 35,188 12.7 Yes B9 71,592 9,596 7.9 Yes B10 64,031 15,340 62.7 Yes B11 51,582 5,131 19.3 Yes B12 38,431 12,826 1.0 Yes B13 30,564 8,140 17.7 Yes B14 25,483 4,850 0.0 Yes B15 14,990 5,472 0.3 Yes B16 13,879 4,631 21.5 Yes B17 12,299 6,508 5.9 Yes B18 5,326 1,636 1.8 Yes B19 3,496 572 0.0 Yes B20 2,267 430 0.0 Yes B21 1,797 708 0.0 Yes Average 214,799 41,228 13.3 -         50      Table A3 - Political Importance of Chairing the Committees The table analyzes the relative importance of chairing a commission by comparing senators that have presided an influential committee with two other types of senators: Type I is composed of all senators that never chaired any commission and type II comprises senators that have only chaired less influential commissions. The statistics are based on hand-collected information on all 256 senators elected in 2000 and 2006. Fraction indicates the proportion of senators under each type. Head of Party List is an indicator of whether the senator was the top candidate from a party in the senate elections of a given state. Past and Future position are indicators whether the senator has had important political positions (including governor, senator, ambassador or deputy) in the past and future.    Type I Type II Type III Not chairing any Chairing less important Chairing important committee committee committee Fraction (%) 44 42 14 Head of Party List (%) 44 67 81 Past Position (%) 4 30 56 Future Position (%) 5 6 39   51       Table A4 - Geographic Randomization and T-Test for the Coefficient Estimates As an extension for the falsification test in VI.D, we do the randomization process 100 times and repeat the analysis for each sample to obtain a set of coefficients. The table presents the results for the t-test of whether the mean of the coefficient estimate differs from zero. The dependent variables include loan terms at origination and ex post performance. Please refer to Appendix A for variable definitions. State-level GDP growth and unemployment rate are included as controls in all specifications. Standard errors are reported in parentheses and are clustered at the state level. *, **, *** indicate significance at the 10%, 5% and 1% levels, respectively.    Loan Volume Loan Spread Maturity D(Collateral) D(Default) (1) (2) (3) (4) (5) Chairman 0.064 -0.003* 0.005 -0.001 0.000 (0.041) (0.002) (0.007) (0.001) (0.000) Bank*Firm FE Yes Yes Yes Yes Yes Bank*Quarter FE Yes Yes Yes Yes Yes Industry*Quarter FE Yes Yes Yes Yes Yes N 594,534 594,534 594,534 594,534 480,354       52    Figure A1 - Lending in Colima versus Rest of Mexico-Other Loan Terms  Interest Rate Maturity 2 0.4 1.5 0.3 1 0.2 0.5 0.1 0 0 -0.5 -0.1 -1 -0.2 -4 -2 0 2 4 -4 -2 0 2 4 Collateral Default 0.15 0.3 0.1 0.2 0.05 0.1 0 0 -0.05 -0.1 -0.1 -0.2 -0.15 -0.3 -4 -2 0 2 4 -4 -2 0 2 4 Quarters since chairmanship appointment Quarters since chairmanship appointment Colima Agri Colima NonAgri All Agri All NonAgri The figure compares other loan terms including loan spread, maturity, collateral requirement and ex post performance to agriculture and all industries for the state of Colima with that of other states in Mexico (excluding Colima) around the time when the senator from Colima is appointed as chairman of the agricultural committee. The horizontal axis presents the length in quarters since the election of the chairman, and the vertical axis shows loan volumes which are normalized to 100 in the quarter before the chairmanship appointment.     53    Figure A2 Dynamics of Political Lending through the Election Cycle-Other Loan Terms  The figure shows shares of other loan terms including loan spread, maturity, collateral requirement and ex post performance at the bank-state level in states with chairman (relative to those without) through the election cycle. The horizontal axis presents the length in quarters since the election for the chairman and the vertical axis shows the difference in the average loan terms in states with and without chairman relative to their overall mean.     54      Figure A3 - Mean and Confidence Interval for the Estimates from the Randomized Sample   Loan Volume Interest Rate 2 0.1 1 0.05 0 0 -1 -0.05 -2 -0.1 0 20 40 60 80 100 0 20 40 60 80 100 Maturity Collateral Requirement 0.4 0.06 0.2 0.04 0.02 0 0 -0.2 -0.02 -0.4 -0.04 0 20 40 60 80 100 0 20 40 60 80 100 Default 0.02 0.01 0 -0.01 -0.02 0 20 40 60 80 100 Confidence Bounds Mean   The figure plots the mean and confidence interval for the set of coefficient estimates from the generalized random samples. We achieve this by first doing the randomization process 100 times to generate 100 random samples then repeating the analysis for each sample to obtain a set of coefficient estimates and statistics. The horizontal red line indicates the zero level.   55