Dynamic Lobbying: Empirical Evidence from Foreign Lobbying on Free Trade Agreements

Dynamic Lobbying: Empirical Evidence from Foreign Lobbying on Free Trade Agreements Hye Young You Abstract How do lobbying strategies change as legislation advances, and do lobbying activities influence the voting behaviors of legislators? Answering these questions has been severely constrained due to limitations in the comprehensive lobbying contacts data. Using a novel dataset constructed from foreign lobbying reports, I analyze lobbying activities by the governments of South Korea, Colombia, and Panama over ten years (2003-2012) regarding their respective free trade agreements. I find that lobbyists on behalf of foreign countries target different groups of legislators, both regarding their opinions on free trade and institutional positions, between the agenda-setting and voting stages. I also present evidence that lobbying influences voting patterns on free trade agreements among Democrats. I thank Larry Bartels, Brett Benson, Joshua Clinton, Jeffry Frieden, Karam Kang, Brenton Kenkel, Insong Kim, Melinda Ritchie, Michael Sances, and Alan Wiseman for their comments. Assistant Professor, Department of Political Science, Vanderbilt University. 230 Appleton Place, Nashville, TN 37203. Email: hyeyoung.you@vanderbilt.edu. 1

1 Introduction Lobbying is an important part of the legislative process but the extent of the intensity, depth, and impact of lobbying activities is difficult to study due to limited information about lobbying contacts. The Lobbying Disclosure Act of 1995 (LDA), which regulates the domestic lobbying process, only requires lobbyists to disclose the identity of the chamber of Congress or federal agency contacted, and does not require disclosure of the identity of contacted persons. As a result, scholars have relied on interviews with legislators and lobbyists to acquire information on the identities of targeted policy-makers. Furthermore, existing empirical studies are static in nature and focus on a relatively short window of time (Wright 1990; Hojnacki and Kimball 1998, 1999), in comparison to the whole period during which a bill is introduced to the policy community, discussed, and finally enacted (or not) into law. Given that targeting strategies can vary depending on legislative stages, it is important to understand the dynamic nature of lobbying. Lack of comprehensive lobbying contact information also significantly limits scholars endeavor to identify the effect of lobbying on the behavior of legislators. Although recent papers identify the effect of lobbying spending on various policy outcomes (de Figueiredo and Silverman 1006; Baumgartner et al. 2009; Richter, Samphantharak, and Timmons 2009; Facchini, Mayda, and Mishra 2011; Kang 2016; Goldstein and You 2016), little empirical research has been done to identify the influence of lobbying on individual legislators voting because lobbying reports under the LDA do not provide any information on how much money was spent to target a specific legislator. This is in stark contrast to the vast literature on the effect of campaign contributions on roll-call voting, where scholars can identity the amount of contributions made from special interest groups to individual legislators. 1 In this paper, I advance the literature by studying new data constructed from the lobbying reports mandated by the Foreign Agent Registration Act (FARA). In particular, I focus 1 See Ansolabehere, de Figueiredo, and Snyder (2003) for a summary of studies on the influence of campaign contributions on roll-call voting. 2

on the activities that the governments of South Korea, Colombia, and Panama undertook to lobby the United States government regarding their respective free trade agreements (FTA); all of which were ratified by the Congress at the same time in 2011. Unlike the majority of foreign government lobbying that does not target a specific legislation, lobbying activities by the governments of South Korea, Colombia, and Panama regarding free trade agreements with the US can be directly linked to roll-call voting on the ratification of each treaty in the Congress. The novelty of this dataset is twofold. First, I follow lobbying activities on the free trade agreements from the 108th Congress (2003-2004) to the 112th Congress (2011-2012). Hence I am able to follow a policy issue and the relevant lobbying activities for over a decade - from the beginning to the end of a policy enactment process. Second, I collected detailed information for over 5,800 individual contacts that lobbyists made with policymakers, including legislators, their staffers, and government officials at executive agencies such as the Office of the United States Trade Representative. The information for each contact includes (i) the contacted person s identity and position in the government, (ii) the type of contact (i.e., phone call, in-person meeting, or other), (iii) the date of the contact, and (iv) the purpose of the contact. This rich set of information allows me to measure the intensity of lobbying to each policymaker over time. Based on this dataset, I first conduct a dynamic analysis focusing on whom the lobbyists contacted, and, if they lobbied a policymaker, when and how they contacted her/him. To do this, I divide the period of study into two stages for each of the three trade agreements: one stage for determining the content of the trade agreement, which we call the agenda-setting stage ; and the other stage for determining the fate of the agreement via votes, which we call the voting stage. The point at which the two stages are divided is when the final trade agreement was signed by the executive branches in each country s government. This allows me to examine how lobbying strategies change between the two stages. My analysis yields four findings. First, a large fraction of total lobbying activities on each trade agreement occurred during the agenda-setting period, while lobbying intensity, as 3

measured by per-day expenditures and contact frequency, is much higher during the voting stage. Second, more powerful legislators, defined as those who are either committee chairs or party leaders, are more likely to be contacted during both stages. Third, staunch opponents of the trade agreements are lobbied in both stages, and the intensity of lobbying towards these opponents tends to decrease in the voting stage relative to the agenda-setting stage. Lastly, electorally vulnerable legislators are lobbied more heavily during the voting stage than in the agenda-setting stage. These patterns indicate that lobbying strategies about whom to target vary as legislation advances in Congress. Next, I examine whether lobbying contacts influence legislators votes. Assessing the impact of lobbying on voting behavior is challenging because lobbying decisions about whom to target are not random, and therefore it is possible that lobbyists simply target members who are more likely to support free trade. To address this selection bias, I use lobbying contacts by other countries on non-trade issues in years before voting on free trade agreements as an instrumental variable. There is strong evidence that lobbying contacts affect Democratic legislators voting choices. Although I employ foreign lobbying contact data to understand lobbying resource allocation and the relationship between lobbying and voting, the results presented in this paper can be generalized to understand the broader lobbying process. Half of the lobbying firms hired by foreign governments in the sample also represented domestic clients; and these 14 firms, combined, had over 10% of the total shares in the domestic lobbying industry and represented 1,515 domestic clients during the period between 2008 through 2010. 2 Also, 72% of total contacts and 78% of total congressional contacts made on behalf of the governments of Colombia, Panama, and South Korea were conducted through those 14 firms. These findings suggest that the conclusions from this study can have general implications for the lobbying process. 2 Calculations were made by the author based on the data compiled by The Center for Responsible Politics (opensecrets.org). Top lobbying firms, measured by total revenues in the domestic lobbying market (https://www.opensecrets.org/lobby/top.php?indextype=l), such as Patton Boggs LLP or Akin, Gump et al. are active intermediaries in the foreign lobbying market. 4

2 Lobbying Strategy and Its Influence One of the key questions in the lobbying literature is how to identify the types of legislators who are targeted by lobbyists. Examining the identities of these policy-makers who are lobbied allows us to study how interest groups allocate their scarce resources to influence the political process. The extant literature on targeting reveals some patterns in targeting strategy in the lobbying process (de Figueiredo and Richter (2014)). First, powerful legislators, such as committee chairs, ranking members, or members in leadership positions, are more likely to be targeted because they have agenda-setting power that can influence the prospects of legislation (Evans 1996; Hojnacki and Kimball 2001; Drope and Hansen 2004). Second, lobbyists tend to target sponsors or co-sponsors of bills who have greater knowledge of, and information about, legislation (Hojnacki and Kimball 1998). Third, legislators who are allies of interest groups, meaning those who have policy preferences similar to lobbying groups, tend to be heavily targeted (Kollman 1997; Caldeira and Wright 1998; Hojnacki and Kimball 1999; Heberlig 2005; Hall and Deadorff 2006; Hall and Miller 2008); and swing legislators are sometimes contacted (Holyoke 2003; Kelleher and Yackee 2009). Fourth, some argue that lobbyists target opposing legislators and this provides an incentive for other groups to lobby their allies to counteract the lobbying efforts by competing groups (Austen-Smith and Wright 1994, 1996). Scholars arrive at different conclusions on targeting strategies in part because they examine different points in time in the legislative process. Studies that examine lobbying activities during the committee stages find that lobbyists target committee chairs, legislative allies, and sponsors of bills (Evans 1996; Kollman 1997; Hojnacki and Kimball 1999). Since the committee stage matters for agenda setting and the contents of bills, it is natural for scholars to observe that lobbyists contact legislators who have the most influence on those activities. Scholars who examine the voting stage, however, find that interest groups lobby allies as well as swing voters to secure support for specific legislation (Hojnacki and Kimball 1998; Holyoke 2003; Kelleher and Yackee 2009). However, the legislative process can generate different lob- 5

bying intensities and dynamics and it is possible that the same interest group could change its tactics depending on the stage of the legislative process (Grossmann and Pyle 2013). Focusing on only a specific point in time within the entire legislative process potentially yields an incomplete understanding of resource allocation by interest groups. Following lobbying activities for the entire duration of the legislative process - from bill introduction to the enactment of law - is important to understand fully the targeting strategies employed by lobbyists. Most of the existing literature on targeting also relies on survey data from lobbyists to measure lobbying contacts due to a lack of systematic lobbying contact information. As noted above, lobbying reports submitted under the LDA are not required to disclose the identities of targeted legislators. While the survey method has its own advantages, such as flexibility in designing questions, there are several shortcomings with such an approach for studying lobbying. First, non-random response rates from lobbyists and a small number of observations complicate statistical inference (de Figueiredo and Richter 2014). Second, relying on survey data makes it challenging to measure the intensity of targeting, which is another important dimension of lobbying strategy. More specifically, while surveys of lobbyists reveal the identities of targeted legislators, it is difficult to identify the intensity of groups targeting efforts at various stages of the legislative process. Although the availability of lobbying disclosure reports allows measurement of the intensity of lobbying through the amount of lobbying money spent, these reports do not offer the chance to measure the intensity of targeting efforts. Hence, it is important to draw on a new data source, such as foreign lobbying data employed in this paper, that includes systematic lobbying contact information, including the identities of targeted legislators. Another important question in the literature on interest groups is how to identify the extent to which interest groups influence voting choices of legislators and policy outcomes through campaign contributions and lobbying activities (Grossman and Helpman 1994; Austen-Smith 1995; Wright 1996; Grossman and Helpman 2001; Baumgartner et al. 2009). 6

While a vast amount of studies have been done on the effect of campaign contributions, both on roll-call voting and policy outcomes (Stratmann 1991; Snyder 1992; McCarthy and Rothenberg 1996; Goldberg and Maggi 1999; Gawande and Bandyopadhyay 2000; Jayachandran 2006; Mian, Sufi, and Trebbi 2010), research on the effect of lobbying on policy outcomes has recently burgeoned. One feature that distinguishes empirical lobbying literature from the literature on campaign contributions is that the majority of the research focuses on lobbying s effect on policy outcomes (de Figueiredo and Silverman 1006; Baumgartner et al. 2009; Richter, Samphantharak, and Timmons 2009; Facchini, Mayda, and Mishra 2011; Kang 2016; Goldstein and You 2016), leaving the effect of lobbying on legislators voting choices unpacked. 3 Unlike campaign contributions, lobbying reports do not distinguish how much time is spent to lobby specific legislators. This limitation explains why there is a dearth of empirical literature on lobbying s influence on individual legislator s roll-call voting and why the extent of groups lobbying activities has been exclusively measured by their lobbying expenditures. While lobbying expenditures can show interest groups willingness to participate in the political process, these raw expenditure totals do not show how groups allocated their expenditures between legislative and executive bodies, both of which are critical institutions. Nor does it show how effectively expenditures were distributed across different members of the legislature. For example, targeting members of the House Ways and Means Committee or the Senate Finance Committee who have direct jurisdiction over trade policy in Congress, could be more effective in influencing trade policy than targeting members whose committee assignments are less relevant to international trade issues even though exactly the same amount was spent on lobbying each group. Hence, even if groups spend equal amounts on lobbying, the composition of lobbying contacts can have starkly different implications for the effect of lobbying on policy outcomes. Utilizing foreign lobbying reports submitted by lobbyists representing the governments of 3 Wright (1989, 1990) made an effort to compare the effect of campaign contributions and lobbying on legislators roll-call voting by using lobbying contact data from interviews with lobbyists. 7

South Korea, Colombia, and Panama regarding their respective free trade agreements allows me to overcome limitations in the extant literature outlined here. First, unlike lobbying reports submitted under the LDA, foreign lobbying reports include detailed information about who is contacted. Therefore, at each legislative level, I can measure the intensity of lobbying targeting activity. Second, detailed lobbying contact information allows me to control the contents of lobbying activities, along with total lobbying spending. Third, the majority of foreign government lobbying does not target specific legislation and sometimes lobbyists focus on multiple issues (Kang and You 2016). Therefore, it is hard to connect their lobbying activities to specific roll-call voting outcomes. However, lobbying activities by the governments of South Korea, Colombia, and Panama regarding free trade agreements with the US can be directly linked to roll-call voting on the ratification of each treaty in the Congress. With the specific voting choice outcome, I can estimate the influence of detailed lobbying activities on individual legislators voting choices. 3 Background of Free Trade Agreements and Data Description A free trade agreement (FTA) is one way to reduce trade barriers between two or more countries and establish a more stable and open trading and investment environment. The US entered its first FTA with Israel in 1985; since then, the United States has implemented 14 FTAs with 20 countries. 4 After enactment of the Trade Promotion Authority (known as fast track ) under the Bipartisan Trade Promotion Act of 2002 that granted the president and executive branches authority to negotiate international agreements with other countries, the Bush administration initiated discussions of potential free trade agreements with Colombia, Panama, and South Korea in 2003. Free trade agreements between the United States and each country were signed in Novem- 4 https://ustr.gov/trade-agreements (accessed on February 28th, 2016). 8

ber 2006 (Panama) and June 2007 (South Korea and Colombia) by the Bush administration; but due to various domestic political challenges and concerns, their ratification was delayed. First, by the time President George W. Bush signed the treaties, Democrats held majorities in both chambers of Congress and ratification was not considered likely before the 2008 presidential election: The South Korean and Panamanian FTAs were not even introduced for ratification votes. Second, during the 2008 presidential campaign, then-senator Obama expressed reservations about all three treaties; and many of his concerns were shared by other legislators. On the Colombian treaty, many Democrats were concerned that the treaty terms were unfavorable to local labor unions. On the Panamanian treaty, Democrats suspected that some of the treaty terms might make the Latin American nation a tax haven for American businesses. The South Korean treaty was the largest and most complicated of the three, and it was the source of significant special interest opposition in the US and in South Korea. These challenges were eventually engaged during the Obama administration. The administration started renegotiations with the South Korean government, and the parties reached a deal on December 3, 2010. This new agreement received support from US automobile companies and the United Auto Workers, both of which had opposed the previous agreement. On April 6, 2011, the government of Panama agreed to sign a tax treaty to clear potential tax haven issues. Finally, the administration received a letter from the Colombian president on April 19, 2011, which assured the US government of a plan to protect labor rights. These actions paved the way for ratification. Bills were introduced into Congress on October 3, 2011, and all three pacts were ratified ten days later without amendment. 5 I collected information on all lobbying contacts made by commercial lobbyists whose clients are one of the three countries studied, and whose contacts were related to the previously mentioned free trade agreements. 6 The data is drawn from the lobbying reports 5 Under the fast-track authority, the president and executive branches have complete authority to negotiate international treaties and Congress can only accept or reject the agreement. No amendments or filibusters are allowed. 6 There are reports made by non-lobbying firms on behalf of the three countries. Their lobbying activi- 9

mandated by the Foreign Agent Registration Act of 1938 (FARA). Under FARA, lobbyists hired by foreign governments and businesses must register with the Department of Justice and submit a semi-annual lobbying disclosure form. The FARA imposes stricter disclosure requirements than the Lobbying Disclosure Act of 1995 (LDA). Unlike the LDA, the FARA requires detailed disclosure of each contact to government officials including (i) the name of the contacted person and his/her position in the government, (ii) the date of the contact, (iii) the method by which the person was contacted (phone call, in-person meeting, email, etc.), and (iv) the issues discussed with the contact. 7 Among the FARA reports submitted by the lobbying firms that represented one of the three countries, I focus on the reports that specifically mentioned free trade agreement in their issue description. Table 1 presents the summary statistics of these data. There are 145 such reports covering 2003-2012, and three foreign governments paid $24 million to US lobbying firms for their services. 8 Table 1: Lobbying Activities by Foreign Governments, 2003-2012 Country # Firms a # Reports b Amount ($) Colombia 10 27 3,064,223 Panama 14 54 9,956,262 South Korea 17 73 11,742,543 Total 35 c 145 c 24,763,028 Note: a. Total number of lobbying firms that were hired by each country. b. Total number of FARA reports on free trade agreements. c. Because some lobbying firms represented more than one country, this number is not equal to the sum of the column. ties mainly comprised monitoring the process rather than contacting members of Congress or the executive branches. Ninety-nine percent of the contacts made to members of Congress and bureaucrats in federal agencies were made by lobbying firms. Since my focus is identifying types of legislators contacted and estimating the effect of lobbying on members voting, I concentrate on lobbying reports submitted by commercial lobbying firms that made the most contacts. 7 See Appendix F for a sample foreign lobbying report. 8 Among 145 reports, 38 do not include any contact information because those lobbying firms on behalf of the three countries did not engage in any political activity. Instead, those firms reported that they provided advice on general trade matters and analyses regarding trade developments in the US, including monitoring and reporting on developments of the FTA. 10

I extracted contact information from these lobbying reports, which were made via meetings, phone calls, emails, and through social events. To capture meaningful communication between lobbyists and policy-makers, I only used contacts made via meetings or phone calls. Over the time period, 5,831 such contacts were made. Table 2 presents the summary statistics of these data. Eighty percent of the contacts targeted members of Congress, but there were a substantial number of contacts made to those in federal agencies, such as US Trade Representatives (USTR) or members of the State Department. Moreover, 31% of the contacts were made in meetings, and there is significant variation across the three countries regarding targeted institutions and contact type. 9 Table 2: Lobbying Contacts by Foreign Governments, 2003-2012 Contact Agency Contact Type Country Total Congress(%) Executive Media Other Meeting(%) Phone Colombia 791 85.9 10.8 2.5 0.8 42.1 57.9 Panama 876 70.6 24.2 3.8 1.5 70.2 29.8 South Korea 4,164 82.1 11.4 2.7 3.8 21.0 79.0 Total 5,831 80.8 13.2 2.9 3.1 31.2 68.8 I also collected information from various sources on the 838 unique members of Congress who served during 2003-2012. For each member, I gather information on (i) ideology (party affiliation, DW-NOMINATE score), (ii) leadership position committee assignments, (iii) the winner s vote share in the most recent election, (iv) whether or not the member is affiliated with a Congressional caucus related to one of the three countries or the region to which any of the countries belong, (v) campaign contributions from union organizations, and (vi) district characteristics (trade volume with each country, population size, per capita income, manufacturing job growth, and China s import penetration). In addition, I have records of 9 Among congressional contacts, 67.4% targeted members of the House of Representatives and 21.1% targeted senators. Congressional staff members in the Senate Finance Committee or House Ways and Means Committee whose jurisdictions cover international trade were followed in terms of targeting (6.4% and 3.8%, respectively). Among the executive branch contacts, 67.7% targeted the USTR and 8.6% targeted officials at the White House. Contacts to the Department of State and Department of Commerce followed at 5.7% for each. 11

how members who served in the 112th Congress voted on ratification of all three free trade agreement ratifications. 4 Dynamic Lobbying Analysis In this section, I analyze how interest groups allocate their resources over time and over different targets. I divide the analysis into two stages: the agenda-setting stage, and the voting stage. The agenda-setting stage is defined as beginning on the date of the earliest lobbying report for one of the three nations studied that mentions free trade agreement as its lobbying issue. Given this definition, the agenda-setting stage started on November 29, 2002, for South Korea, on August 30, 2004; for Colombia, and on May 1, 2003; for Panama. Note that the treaties were signed in June 2007 for South Korea and Panama, and in November 2006 for Colombia. 10 Hence, lobbying on these treaties started at least two years before they were signed. The agenda-setting stage is defined as ending when the treaties and relevant agreements were finalized: December 3, 2010, for South Korea; April 6, 2011, for Colombia; and April 19, 2011, for Panama. Hence, the duration of the agendasetting period was 2,926 days for South Korea, 2,410 days for Colombia, and 2,910 days for Panama. The voting stage is defined as beginning immediately after the agenda-setting stage ends, and ending on the date of the votes in both the House and Senate: October 12, 2011, for all three countries. Given this definition, the duration of the voting stage was 313 days for South Korea, 189 days for Colombia, and 176 days for Panama. During this stage, no amendments to the final treaties and agreements were made. Table 3 shows the lobbying expenditures and lobbying contacts by each country during each period. Note that I have total lobbying expenditures for a given six-month period, per report. To calculate the lobbying expenditures for each stage from the data, I pro-rate the report-level data by the number of reported days. One interesting trend in the data is 10 The Bush administration started conversation on free trade agreements in early 2003. 12

that the majority (over 70%) of total lobbying expenditures was spent during the agendasetting stage. That said, the intensity of the lobbying activities, as measured by per-day expenditures, is much higher during the voting stage than during the agenda-setting stage. A similar pattern is obtained when measuring the number of contacts with Congress, executive agencies, and the media. Table 3: Lobbying Expenditures and Lobbying Contacts Agenda Setting Voting Country Total Per-Day Total Per-Day Panel A. Lobbying Expenditures ($) Colombia 2,701,223 1,168 362,999 1,614 Panama 7,704,098 2,474 2,252,163 7,513 South Korea 7,857,140 3,268 3,795,402 7,607 Panel B. Lobbying Contact Colombia 330 0.14 461 2.43 Panama 736 0.25 140 0.79 South Korea 2,709 0.92 1,455 4.64 In Figure 1, each point represents a contact record by lobbyists representing Colombia, Panama, and South Korea; the party affiliation of the contacted member is also shown. The vertical line indicates the date of the congressional votes. It is noteworthy that while free trade negotiation was handled by federal agencies such as the USTR, and the Department of State, lobbyists frequently targeted members of Congress during the process. Given that fast-track authority under the Trade Promotion Act grants all authority to negotiate international trade agreements to the president, and the Congress can only approve or reject the agreement (no amendment or filibuster is allowed) (Koh 1992), this intensive lobbying during the agenda-setting stage is puzzling. However, these lobbying patterns are consistent with theories arguing that interest groups press legislators to influence policy decisions made in executive branches (Fiorina 1977; Epstein and O Halloran 1995; Hall and Miller 2008; Jordan and Meirowitz 2012; Ban and You 2016; Ritchie and You 13

2016). Likewise, because trade agreements are very specific about the terms of treaties such as tariff schedules, legislators do not have much influence on the ex post implementation of trade agreements after they have been ratified. 11 Hence, interest groups have strong incentives to contact legislators before the voting stage occurs to design agreements that are favorable to their concerns (You 2016). Figure 1: Lobbying Contacts Over Time Note: Graph is jittered. Next, I present how lobbyists on behalf of Colombia, Panama, and South Korea changed their targets as legislation advanced, and focus on the identities of the targeted legislators. Among the 839 members of Congress who served during the 108th - 112th Congresses, 538 of them were contacted at least once via phone or in face-to-face meetings by lobbyists who represented one of the three foreign countries. In total, 4,681 contacts targeting members of 11 For example, the final text of the Korea-US FTA agreement is several thousand pages long and provides a very detailed tariff schedule at the HTS (Harmonized Tariff Schedule) 8-digit level (https://ustr.gov/ sites/default/files/uploads/agreements/fta/korus/asset_upload_file199_12753.pdf). 14

Congress were made. I seek to identify the types of legislators who were targeted during each stage of the legislative process. Specifically, I am interested in whether a member s position regarding the free trade agreement, meaning that he/she was either supportive, undecided, or opposed, is associated with lobbyists targeting in the agenda-setting or voting stage. To analyze lobbyists strategies in targeting legislators, I developed the following measure of how favorable or unfavorable legislators predispositions were regarding their opinions on trade agreements prior to being lobbied. First, I collected voting records for 74 trade-related bills from the 108th through 112th Congresses. 12 Drawn from voting records on these 74 measures, I constructed ideal points for each legislator that captures their ideological preferences on free trade. There may be concern that vote choices for those 74 bills may be affected by lobbying activities on behalf of South Korea, Colombia and Panama therefore it is inappropriate to use vote choices for those bills to construct the members ideal points on free trade. But, none of the 74 bills are specific to one of the three countries, and none of the lobbying reports submitted by these three foreign governments mentioned any of these bills. Figure A1 in Appendix B presents the distribution of these free trade ideal points (left), and the relationship between DW-NOMINATE scores and the free trade ideal points (right). 13 Having measured legislators preferences over free trade, I then placed the legislators into quartiles based on their free trade ideal point estimates. Legislators were classified (1) strong protectionist if the ideal point is below the first quartile, (2) weak protectionist if the ideal point is between the first quartile and the median, (3) weak free trader if the ideal point is between the median and the third quartile, and (4) strong free trader if the ideal point is above the third quartile. I also define a legislator as being part of the leadership if 12 Appendix A provides the list of legislation. I excluded FTA legislation on Colombia, Panama, and Korea because voting on those bills are the outcomes of interest. 13 Unsurprisingly, Democrats tend to be more protectionist and Republicans are more likely to support free trade. It is worth noting that DW-NOMINATE scores are highly correlated with voting behaviors on trade-related issues. 15

he/she had taken a leadership position, and code a committee variable for those who served on committees that have jurisdiction on international trade affairs: the House Ways and Means and Senate Finance committees. Figure 2 compares the proportion of the members of Congress who belong to each group - SP through SF depending on their predisposition on free trade, leadership, and their membership in the House Ways and Means Committee or the Senate Finance Committee - and the lobbying resource allocation on those members as measured by the number of lobbying contacts made by lobbyists on behalf of South Korea both in the agenda-setting and voting stages. 14 Figure 2: Composition of Congress and Lobbying Contacts I find three key trends in the table. First, legislators who held leadership positions, or who served on the House Ways and Means or the Senate Finance Committees are contacted frequently during both stages of the legislative process. Leaders consist of around 7% of the legislators during the period of study, but they were the targets of more than 15% of 14 In Appendix B, I also present the results for lobbying by Colombia and Panama that shows patterns similar to the lobbying by South Korea. 16

the lobbying contacts, and 18% of the total contacts targeted committee members who have direct jurisdiction on the issue. Second, a significant amount of contacts (23%) targeted strong opponents of free trade during both stages, and this pattern is more salient during the agenda-setting stage. Third, an increase in lobbying weak opponents and strong allies is shown as we move from the agenda-setting stage to the voting stage. Some of these findings comport with existing informational theories of lobbying, which hold that lobbyists provide costly private information to legislators on the policy at hand (Hansen, 1991; Austen-Smith, 1993; Lohmann, 1995; Austen-Smith, 1998). Using this comparative informational advantage, lobbyists gain access to legislators. However, most extant theories would suggest that targeting strong opponents is relatively rare, unless there is counteractive lobbying by competing interest groups taking place (i.e., Austen-Smith and Wright 1994, 1996). I argue that the extant literature does not fully capture the dynamics of lobbying because these informational theories ignore potential spillovers. More specifically, these theories fail to account for how the interactions between a lobbyist and a legislator may affect a lobbyist s future interactions with other legislators. This omission might have emerged because the focus in the literature has centered largely on the information that lobbyists provide to legislators. However, legislators may provide their own information to lobbyists, who may selectively spread that information to other legislators; and this kind of informational dynamic is more relevant when we take into account that policy making is often a collective process. With this view on the role of lobbyists, my findings for each stage of the legislative process can be easily explained. During the agenda-setting stage, we would expect to see a diverse set of legislators being contacted so as to collect information about the bill s prospects. Therefore, contacting strong opponents can help lobbyists and supporting legislators who push the policy through the legislative process, as they can provide information on likely sources of opposition to the bill. Furthermore, legislators in leadership positions, or who 17

serve on relevant committees are an important source of information about legislation. During the voting stage, it may be important for each legislator to know how other legislators intend to vote, and thus the potential electoral implications of her/his own vote. Legislators in leadership may want to keep updated about other legislators intentions to vote; lobbyists would have less incentive to lobby strong opponents in this stage because it is less likely that lobbying could change their voting decisions. That said, legislators who are on the fence (i.e., weak protectionists or weak free traders) may find lobbyists useful during the voting stage to help them make informed calculations about their votes impacts on their elections and party positions. To systematically investigate which legislator characteristics are associated with more lobbying contacts in addition to their ex ante positions for free trade, I conduct the following OLS analysis for each agenda-setting and voting stage. 15 Let y ijt denote the number of contacts between legislator i and a lobbyist during stage t (agenda-setting or voting). y it = β t X it + ε it. (1) I use two measures for lobbying contacts: (1) the total number of contacts made to a member s office, including contacts with member s staffers (Contact) and (2) the total number of contacts made directly with a member (Member). X it includes control variables such as party affiliation (Democrat), legislator i s ex ante position for free trade based on previous trade-related votes (Trade Ideal Point), whether the legislator has a leadership position 15 The three countries have slightly different thresholds that divide their agenda-setting and voting stages. For this analysis, I set a common threshold for the three countries and present the results for the Senate and the House of Representatives separately. The common threshold date for dividing the two stages is April 19, 2011 when Panama and the US finalized the treaties and relevant agreements. This is also the last date among the three agreements when executives of the US and foreign governments were involved in designing the contents of the agreement. South Korea and Colombia finalized their treaties on December 3, 2010 and April 6, 2011, respectively. After finalization of the treaties, the critical role moved to the Congress to ratify the treaties The results are similar when presented for individual countries. However, if we present the result for each country, separating the Senate and the House of Representatives, there are six different regression results that do not add much value. For clarity in presenting the result, I use the common threshold. 18

(Leadership), whether he/she belongs to the House Ways and Means or Senate Finance Committee (Committee), whether his/her vote share in the most recent election was less than 55% (Tight Race) (implying that he/she may be electorally vulnerable compared to other legislators), the Legislative Effectiveness Score for members in the House of Representatives (LES) (Volden and Wiseman 2014), and whether he/she is a member of a Congressional caucus regarding the respective country or region (Caucus). 16 The results are presented in Table 4, where Panel A presents the results for the House and Panel B presents the results for the Senate. By comparing the results between the two stages, several patterns emerge. First, legislators in the House who held leadership positions are contacted more frequently by lobbyists both in the agenda-setting and voting stages. Second, members who served on the relevant committees on international trade are contacted heavily, and this pattern is more salient during the agenda-setting stage. Although the president and executive agencies, such as the USTR, had the authority to negotiate the treaties, legislators who served on committees that have oversight authority over relevant federal agencies on international trade can influence the contents of agreements via communication with agency bureaucrats. Such oversight activities may explain why relevant committee members are heavily targeted during the agenda-setting stage. Third, lobbyists targeted members of the House with high Legislative Effectiveness Scores more often during the agenda-setting stage, and members who are associated with congressional caucuses on Colombia, Panama, or South Korea were heavily targeted over the course of the legislative process. Lastly, electorally vulnerable legislators tended to have more contacts with lobbyists during the voting stage than during the agendasetting stage. This finding would be consistent with the argument that legislators who face strong electoral competition may find lobbyists particularly useful during the voting stage, to help them make informed calculations about their votes impact on their elections. 16 LES and caucus membership data are only available for the House of Representatives. 19

Table 4: Number of Lobbying Contacts and Member Characteristics Panel A. House Agenda Setting Voting (1) Contact (2) Member (3) Contact (4) Member Democrat 2.717 0.148 1.040 0.122 (2.77) (0.89) (1.31) (1.76) Trade Ideal Point 1.743 0.219 0.894 0.00718 (3.42) (2.53) (2.52) (0.23) Leadership 3.371 0.471 3.555 0.229 (3.00) (2.48) (4.58) (3.40) Committee 5.543 1.247 0.673 0.159 (5.95) (7.89) (0.92) (2.50) Tight Race -0.537-0.00781 0.404 0.111 (-0.68) (-0.06) (0.67) (2.11) LES 0.816 0.120 0.0781 0.000868 (5.54) (4.78) (0.80) (0.10) Caucus 5.930 0.788 0.809 0.0902 (7.24) (5.67) (1.54) (1.97) N 695 695 436 436 Panel B. Senate Democrat 5.099 0.426 12.46-0.0563 (1.32) (1.86) (1.37) (-0.25) Trade Ideal Point 2.379 0.376 4.778-0.0380 (1.04) (2.77) (0.86) (-0.27) Leadership -1.154 0.298 1.732 0.219 (-0.45) (1.96) (0.29) (1.45) Committee 15.92 0.538 25.97 0.176 (3.72) (2.13) (2.90) (0.78) Tight Race 0.0767-0.0848-0.331-0.0596 (0.03) (-0.56) (-0.05) (-0.38) N 134 134 96 96 t statistics in parentheses. p < 0.05, p < 0.01 5 Lobbying and Voting Outcomes In this section, I investigate whether lobbying contacts are associated with voting decisions for free trade legislation in the 112th Congress. House majority leader Eric Cantor (R-VA) introduced the legislation into the House on October 3, 2011; and voting for free trade agreements with Colombia, Panama, and South Korea took place on October 12, 2011. The voting results for each agreement is presented in Table 5, and the voting results closely follow partisan lines. While the majority of Democrats both in the House and the Senate 20

opposed the free trade agreements, Republican members almost unanimously supported the treaties. Therefore, I only focus on Democrats where there is substantial variation in voting patterns. 17 Table 5: Voting Patterns on Free Trade Agreement Bills House Senate Democrats Republicans Total Democrats Republicans Total Panel A. Colombia FTA Number voting Yes 31 231 262 21 44 63 Number voting No 158 9 167 30 2 32 Panel B. Panama FTA Number voting Yes 66 234 300 30 46 76 Number voting No 123 6 129 21 1 22 Panel C. South Korea FTA Number voting Yes 59 219 278 37 45 82 Number voting No 130 21 151 14 1 15 To investigate the influence of lobbying contacts on voting behavior, I assume that member i s probability of supporting a bill v with a foreign country j is characterized by: P r(v ij = 1) = β 0 + β 1 ID i + β 2 CI ij + β 3 DL i + β 4 F L ij + ε ij (2) where ID i includes a member i s ideology and other characteristics, CI i is a proxy for constituent interest in congressional district i, DL i represents domestic lobbying pressure on free trade, and F L i denotes foreign lobbying activities. To measure a member s ideology and characteristics, I include her trade ideal points based on previous trade voting records, leadership and committee assignments, and vote share in the most recent election. To measure member i s district characteristics, I include member i s district level democratic presidential vote share for the 2008 election, and member i s district trade volumes with the three countries. The trade volume data were not disaggregated to the district level, so I constructed the district-level data based on the International 17 The main results do not change even if I include both Democrats and Republicans in the analysis and use an interaction term including Democrat. 21

Trade Commission state-industry level trade data and the zip-code level industry composition data. 18 I also employ the data from Autor, Dorn, and Hanson (2013) to construct the district-level changes in manufacturing employment, and the change in Chinese import exposure-per-worker in the district between 1990 and 2007. 19 To measure domestic groups pressure on free trade agreements, I include campaign contributions made by unions between 2003 and 2010. Given that opposition to the treaties by US unions was the biggest domestic obstacle in the negotiation process, I suspect that contributions from unions played a role in legislators voting decisions. Of course, union contributions is not a perfect measure to estimate domestic pressure on free trade agreements because campaign contributions are different from lobbying activities, and pro-free trade groups also lobbied during the period. An ideal measure for domestic groups lobbying would be detailed contact information from each domestic group to each legislator, as we have in the foreign lobbying data. Unfortunately domestic lobbying reports submitted under the LDA only provide the names of federal agencies, such as the House of Representatives or the USTR, which each domestic group contacted. Individual legislator-level lobbying contact data during the lobbying process by domestic groups do not exist. Hence, union contributions are used as a proxy to capture domestic pressure to oppose the trade deals. I use two measures for foreign lobbying contacts made to a legislator: total lobbying contacts made on behalf of a respective country, and the total lobbying contacts made by the three countries over the period. I consider total lobbying contacts, as well, because even if a member is contacted by a lobbyist on behalf of Colombia on free trade, this contact could influence not only the voting on the Colombian FTA, but also voting on the Panamanian or Korean FTA, since it pertains to the same issue. I introduce the interaction term, Position Lobbying Contact, to see whether lobbying has a heterogeneous effect on the voting 18 https://www.usitc.gov. 19 The change in Chinese import exposure per worker is measured by the difference in change in US imports from China per worker in the region. Autor, Dorn, and Hanson (2013) construct those measures at commuting zones for the period between 1990 and 2007. We use the relationship file between commuting zones and counties, and the relationship file between county and congressional district to assign values at the congressional district level. Population weight is applied. For senators, state-level measures are constructed. 22

behavior depending on legislators ex ante trade positions. Foreign lobbying may have more influence on members who can be persuaded than on stanch opponents of free trade. I estimate Equation (2) for each legislator i for voting for a free trade agreement with a foreign county j using OLS. 20 Table 6 presents the results. Columns (1) through (3) present the results when I measure lobbying intensity by the total number of lobbying contacts made by a specific foreign country; columns (4) through (6) present the results when I measure lobbying intensity by the total number of lobbying contacts made to each member by all three countries. 21 The coefficients on the interaction terms indicate that lobbying made Democrats who were more positively predisposed toward free trade based on their previous voting records lean more in favor of free trade legislation. Given that 51.6% of Democrats are categorized as strong protectionists, and only 0.8% are categorized as strong free traders, the results indicate that the effect of lobbying is mainly from the Democrats who are categorized as weak protectionists (39.1%) and weak free traders (8.5%). The interaction terms are statistically significant for all three voting patterns, regardless of how we measure lobbying intensity. Although I control for variables that could affect the vote choices of members, a decision to target a specific legislator is not random and therefore it is difficult to establish causality between lobbying contacts and vote outcomes. There may be some omitted variables that could affect both lobbying contacts by lobbyists and a propensity to vote for FTA bills. To address this issue, I use the contacts made by lobbying firms on behalf of other foreign countries to lobby members of Congress on non-trade issues before the 112th Congress (when voting on FTA legislation occurred) as an instrumental variable. The rationale behind this empirical strategy is that contacts made by lobbying firms on behalf of foreign countries other than Colombia, Panama, or South Korea on non-trade issues before the 112th Congress may be correlated with contacts made by lobbying firms representing Colombia, Panama, or South 20 Summary statistics for the variables are available in Table A2 in Appendix C. 21 As a robustness check, I also measure lobbying intensity by dollars spent to contact each member and the result is presented in Table A3 in Appendix D. The results are similar. I also run the analysis with the state fixed effect and separately for the House and the Senate. See Appendix D. 23

Table 6: Foreign Lobbying and Voting Patterns of Democrats on FTA Legislation Country Specific Lobbying Total Lobbying Dependent Variable = (1) (2) (3) (4) (5) (6) Voting for FTA Colombia Panama Korea Colombia Panama Korea (ln) Lobbying Contact -0.192-0.113-0.0637-0.140-0.0547-0.0276 (-1.92) (-1.11) (-1.23) (-3.29) (-1.14) (-0.57) (ln) Lobbying Contact Position a 0.167 0.121 0.0689 0.111 0.0829 0.0649 (2.93) (2.00) (2.31) (4.69) (3.10) (2.39) (ln) Trade Volume with j b ($K) 0.00757 0.0576 0.101 0.0119 0.0588 0.100 (0.33) (2.07) (3.70) (0.53) (2.16) (3.72) Trade Ideal Point 0.228 0.387 0.289 0.136 0.301 0.264 (4.87) (7.12) (4.82) (2.58) (5.03) (4.35) Leadership 0.0589 0.121 0.154 0.0836 0.121 0.142 (0.83) (1.51) (1.93) (1.22) (1.55) (1.80) Committee c -0.116 0.183 0.0682-0.119 0.200 0.0342 (-1.39) (1.79) (0.71) (-1.44) (2.13) (0.36) Tight Race d -0.0583-0.136-0.0874-0.0592-0.141-0.0783 (-0.78) (-1.61) (-1.03) (-0.81) (-1.71) (-0.94) Union Contributions ($M) -0.102-0.0881-0.0705-0.100-0.0859-0.0736 (-2.52) (-1.92) (-1.55) (-2.56) (-1.93) (-1.63) Dem. Presidential Vote Share -0.397 0.162 0.0467-0.446 0.0752 0.00876 (-1.65) (0.60) (0.17) (-1.87) (0.28) (0.03) % Change in Manufacturing Emp. e 0.0343 0.0316 0.0533 0.0313 0.0313 0.0542 (2.44) (2.02) (3.40) (2.30) (2.05) (3.50) Competition with China f -0.0105 0.00326-0.00386-0.00951 0.00445-0.00272 (-0.72) (0.20) (-0.23) (-0.66) (0.28) (-0.16) Constant 0.798 0.131-0.524 0.669 0.0170-0.561 (2.14) (0.33) (-1.17) (1.83) (0.04) (-1.27) N 252 252 252 252 252 252 adj. R 2 0.277 0.345 0.336 0.304 0.374 0.351 Note: t statistics in parentheses. p < 0.05, p < 0.01. a. Position indicates four groups depending on member i s trade ideal points. b. j {Colombia, Panama, South Korea} depending on the FTA vote. c. Committee = 1 if a member i serves on the House Ways and Means Committee or the Senate Finance Committee. d. 1 if the vote share in the most recent election is less than 0.55. e. % Changes in manufacturing employment, 1990-2007 (Autor, Dorn, and Hanson 2013). f. Change in Chinese import per worker, 1990-2007 (Autor, Dorn, and Hanson 2013). 24

Korea on free trade agreements. For example, a legislator may learn that meeting with a lobbyist on behalf of other foreign countries is actually useful in gathering policy-relevant information. Therefore, she may be more willing to give access to a lobbyist who represented either Colombia, or Panama, or South Korea. 22 A legislator contacted by a lobbyist on behalf of the United Arab Emirates (UAE) on arms sales in 2008, for example, may be more likely to be contacted by a lobbyist on behalf of South Korea on trade issues because of her willingness to meet lobbyists in general. However, the UAE s contact in 2008 on arms sales may not have any direct influence on voting on FTA legislation in 2012. If this exclusion restriction holds, lobbying contacts made before the 112th Congress meet the requirement to be a valid instrument (Angrist and Pischke 2009). 23 To implement this strategy, I collect the total lobbying contacts made to members by lobbyists on behalf of other foreign governments during the period between 2008 and 2010 on non-trade issues. 24 Seventy-five foreign countries - other than Colombia, Panama, and South Korea - made 7,005 contacts targeting members of Congress during the period. For each Democrat, I calculate the total contacts made by foreign countries between 2008 and 2010. I then use this measure as an instrument for the total contacts made on free trade agreements 22 In the IV framework (LATE), a treatment is the contacts by lobbyists who represented foreign governments other than Colombia, Panama, and South Korea on non-trade issues. The compliers in the LATE (local average treatment effect) framework are members who increased their contacts with lobbyists who represented Colombia, Panama, and Korea because they had lobbying contacts with other countries before the voting took place. 23 Some may be concerned that if a lobbying firm worked both for the UAE and for South Korea, a lobbyist may discuss the South Korean FTA with a member i while lobbying on behalf of the UAE. Except for a few large lobbying firms, lobbying firms that represent foreign governments have less than two foreign governments as their clients during the regular six-month contracting period. Even for the few large lobbying firms with multiple foreign clients at the same time, there is a clear division of labor among lobbyists representing a foreign client and the issue area, as a specific lobbyist s name is presented in the sample lobbying report in Figure F in Appendix F. According to my interviews with lobbyists who represented foreign governments, it is considered a violation of fiduciary duties to their clients if lobbyists discuss other client countries matters when meeting members of congress to discuss the assigned client country s issue. I also checked whether other foreign governments lobbied on Colombia, Panama, and South Korean free trade agreements during the period and none of the lobbying reports and contacts mentioned the issue. Foreign lobbying issues are very country specific. 24 The data source is image files posted in fara.gov under the Department of Justice. Non-trade issues include security (26.4%), administrative (16.4%), budget (11.8%), and general bilateral relationships (45.4%). 25

by Colombia, Panama, and South Korea for the Democratic legislators who served in the 112th Congress when the voting occurred. I have two endogenous regressors: Lobbying Contacts and an interaction term, Lobbying Contact Position. I instrument for the total contacts made to member i on free trade agreements by country j (Contact ij ) using total contacts made by other foreign governments on non-trade issues before the 112th Congress (Pre Contact i ). I fit models of the following form: 25 (FS) : Contact ij = γ 0 + γ 1 Pre Contact i + γ 2 X ij + ν ij (3) (RF) : Voting ij = β 0 + β 1 Contact ˆ ij + β 2 Contact ˆ ij Position i + β 3 X ij + ε ij The first-stage results are presented in Table A7 in Appendix E. The total number of contacts made by the three countries on free trade agreements are strongly associated with the total number of contacts made by lobbyists on behalf of other countries during the years between 2008 and 2010, and F-statistics are well above 10, which is a conventional threshold for a weak instrument (Stock, Wright, and Yogo 2002). Table 7 presents the 2SLS results from the reduced form model. The results are similar to those presented in Table 6. Lobbying contact has a significant effect on Democrats who possess ex ante modest positions on free trade. As lobbying contact increases, Democrats who are not staunch opponents of free trade tend to increase their support for free trade deals. Member s district democratic presidential vote share in 2008 and contributions from unions are associated with opposition to free trade agreements, whereas district s trade volumes with each foreign country and manufacturing employment growth are associated with support for free trade deals. 25 For the regressor with an interaction term (Contact ij Position i ), I instrumented it using Pre Contact i Position i. 26

Table 7: Foreign Lobbying and Voting Patterns on FTA Legislation: 2SLS Estimates Dependent Variable = (1) (2) (3) Voting for FTA Columbia Panama Korea (ln) Lobbying Contact 0.0372-0.0363 0.0130 (0.19) (-0.23) (0.08) (ln) Lobbying Contact Position a 0.240 0.166 0.124 (3.49) (3.03) (2.30) (ln) Trade Volume with j b ($K) 0.0157 0.0682 0.106 (0.47) (2.30) (3.75) Trade Ideal Point -0.211 0.129 0.127 (-1.34) (1.05) (1.03) Leadership -0.0481 0.0628 0.0924 (-0.40) (0.64) (0.96) Committee c -0.458 0.0656-0.0851 (-2.22) (0.40) (-0.52) Tight Race d 0.00518-0.116-0.0560 (0.05) (-1.28) (-0.62) Union Contribution ($M) -0.138-0.0975-0.0858 (-2.17) (-1.90) (-1.69) Dem. Presidential Vote Share 08-0.993-0.0946-0.162 (-2.13) (-0.27) (-0.46) % Change in Manufacturing Emp. e 0.0213 0.0278 0.0508 (1.01) (1.65) (3.05) Change in China Competition f -0.00792 0.00589-0.00216 (-0.37) (0.34) (-0.13) Contant 0.120-0.336-0.841 (0.20) (-0.70) (-1.64) N 252 252 252 adj. R 2-0.625 0.258 0.265 Note: t statistics in parentheses. p < 0.05, p < 0.01. a. Position indicates four groups depending on member i s trade ideal points. b. j {Colombia, Panama, South Korea} depending on the FTA vote. c. Committee = 1 if a member i serves on the House Ways and Means Committee or the Senate Finance Committee. d. 1 if the vote share in the most recent election is less than 0.55. e. % Changes in manufacturing employment, 1990-2007 (Autor, Dorn, and Hanson 2013). f. Change in Chinese import per worker, 1990-2007 (Autor, Dorn, and Hanson 2013). 27

6 Conclusion Despite its significant implications for democratic representation, research on how interest groups change their strategies during the legislative process and whether lobbying influences members voting has faced considerable constraints, mainly due to a lack of detailed lobbying contact information. In this paper, I construct and analyze a novel dataset on the activities by lobbyists that the governments of South Korea, Colombia, and Panama hired to lobby the United States government regarding their respective free trade agreements with the US. This data allows me to follow the lobbying activities for ten years - from the beginning to the end of a policy enactment process. Using the rich set of information from the data, I measure the intensity of lobbying towards each legislator over time. I find that targeting strategies by lobbyists vary depending on the legislative process. During the agenda-setting stage, lobbyists tend to target legislators who have a leadership position, or those who serve on committees that have jurisdiction over international trade, and those who are highly effective in the legislative process. As the legislation advances to the floor voting stage, those who are in leadership and legislators who faced tight races in recent elections tend to have more contacts with lobbyists. Regarding types of legislators, defined by their ex ante positions on free trade, I find that all types of legislators are targeted both in the agenda-setting and voting stages. During the agenda-setting process, strong opponents to free trade are often targeted but that strategy shifted to target undecided members and strong allies during the voting stage. I also present evidence that lobbying influences voting patterns. Lobbying has an especially significant effect on shifting the voting behaviors of Democrats who were strong protectionists to support free trade agreements. This study has several implications for study of special interest groups and the role of Congress in shaping trade policy. First, empirical patterns of targeting illustrate that lobbyists target a diverse set of legislators, and often, strong opponents of the legislation are contacted. This implies that a new framework on lobbying that incorporates potential spillover - in particular, that the interactions between a lobbyist and a legislator may affect a 28

lobbyist s future interactions with other legislators - should be considered, given that passing a bill is a collective process. Second, given that the president and the executive branch have complete authority to negotiate international trade with foreign countries, the sheer amount of lobbying targeting members of Congress during the negotiation stage is interesting and is consistent with theories that argue interest groups lobby Congress to influence the decisions made by federal agencies. This raises an interesting question of how interest groups lobby members of Congress to influence various bureaucratic decisions, not just the implementation of statutes after the legislation has passed the Congress. 29

References Angrist, Joshua, and Jörn-Steffen Pischke. 2009. Mostly Harmless Econometrics. Princeton, N.J.: Princeton University Press. Ansolabehere, Stephen, John M. de Figueiredo, and James M. Snyder. 2003. Why Is There so Little Money in U.S. Politics. Journal of Economic Perspectives 17 (1): 105-103. Austen-Smith, David. 1993. Information and Influence: Lobbying for Agendas and Votes. American Journal of Political Science 37 (3): 799 833. Austen-Smith, David. 1995. Campaign Contributions and Access. American Political Science Review 89 (3): 566-581. Austen-Smith, David. 1998. Allocating Access for Information and Contributions. Journal of Law, Economics, and Organization 14 (2): 277-303. Austen-Smith, David, and John R. Wright. 1996. Theory and Evidence for Counteractive Lobbying. American Journal of Political Science 40 (2): 543-564. Austen-Smith, David, and John Wright. 1994. Counteractive Lobbying. American Journal of Political Science 38 (1): 25 44. Autor, David, David Dorn, and Gordon Hanson. 2013. The China Syndrom: Local Labor Market Effects of Import Competition in the United States. American Economic Review 103 (6): 2121-2168. Ban, Pamela, and Hye Young You. 2016. Lobbying in Federal Agency Rulemaking: Efect on Policy Change. Vanderbilt University Working Paper. Baumgartner, Frank R., Jeffrey M. Berry, Marie Hojnacki, David C. Kimball, and Beth L. Leech. 2009. Lobbying and Policy Change: Who Wins, Who Loses, and Why. Chicago: University of Chicago Press. Caldeira, Gregory, and John Wright. 1998. Lobbying for Justice: Organized Interests, Supreme Court Nominations, and the United States Senate. American Journal of Political Science 42 (2): 499-523. de Figueiredo, John, and Brian Silverman. 1006. Academic Earmarks and the Returns to Lobbying. Journal of Law and Economics 49 (2): 597-625. 30

de Figueiredo, John M., and Brian Kelleher Richter. 2014. Advancing the Empirical Research on Lobbying. Annual Review of Political Science 17: 163-185. Drope, Jeffrey, and Wendy Hansen. 2004. Purchasing Protection? The Effect of Political Spending on U.S. Trade Policy. Political Research Quarterly 57 (1): 27-37. Epstein, David, and Sharyn O Halloran. 1995. A Theory of Strategic Oversight: Congress, Lobbyists, and the Bureaucracy. Journal of Law, Economics, and Organization 11 (2): 227-255. Evans, Diana. 1996. Before the Roll Call: Interest Group Lobbying and Public Policy Outcomes in House Committees. Political Research Quarterly 49 (2): 287-304. Facchini, Giovanni, Anna Maria Mayda, and Prachi Mishra. 2011. Do Interest Groups Affect US Immigration Policy? Journal of International Economics 85 (1): 114-128. Fiorina, Morris. 1977. The Case of the Vanishing Marginals: The Bureaucracy Did It. American Political Science Review 71 (1): 177-181. Gawande, Kishore, and Usree Bandyopadhyay. 2000. Is Protection for Sale? Evidence on the Grossman-Helpman Theory of Endogenous Protection. Review of Economics and Statistics 82 (1): 139-152. Goldberg, Pinelopi, and Giovanni Maggi. 1999. Protection for Sale: An Empirical Invesitgation. American Economic Review 89 (5): 1135-1155. Goldstein, Rebecca, and Hye Young You. 2016. Cities as Lobbyists. Working Paper. Grossman, Gene, and Elhanan Helpman. 1994. Protection for Sale. American Economic Review 84 (4): 833 850. Grossman, Gene, and Elhanan Helpman. 2001. Special Interest Politics. Cambridge: MIT Press. Grossmann, Matt, and Kurt Pyle. 2013. Lobbying and Congressional Bill Advancement. Interest Groups & Advocacy 2 (1): 91-111. Hall, Richard, and Alan Deadorff. 2006. Lobbying as Legislative Subsidy. American Political Science Review 100 (1): 69-84. Hall, Richard L., and Kristina C. Miller. 2008. What Happened After the Alarm? Interest Group Subsidies to Legislative Overseers. Journal of Politics 70 (4): 990 1005. 31

Hansen, John Mark. 1991. Gaining Access: Congress and the Farm Lobby, 1919-1981. Chicago: University of Chicago Press. Heberlig, Eric S. 2005. Getting to Know You and Getting Your Vote: Lobbyists Uncertainty and the Contacting of Legislators. Political Research Quarterly 58 (3): 511-520. Hojnacki, Marie, and David Kimball. 1998. Organized Interests and the Decision of Whom to Lobby in Congress. American Political Science Review 92 (4): 775 790. Hojnacki, Marie, and David Kimball. 1999. The Who and How of Organizations Lobbying Strategies in Committee. Journal of Politics 61 (4): 999 1024. Hojnacki, Marie, and David Kimball. 2001. PAC Contributions and Lobbying Contacts in Congressional Committees. Political Research Quarterly 54 (1): 161-180. Holyoke, Thomas. 2003. Choosing Battlegrounds: Interest Group Lobbying Across Multiple Venues. Political Research Quaterly 56 (3): 325-336. Jayachandran, Seema. 2006. The Jeffords Effect. Journal of Law and Economics 49: 397-425. Jordan, Stuart V., and Adam Meirowitz. 2012. Lobbying and Discretion. Economic Theory 49 (3): 683-702. Kang, Karam. 2016. Policy Influence and Private Returns from Lobbying in the Energy Sector. Review of Economic Studies 83 (1): 269-305. Kang, Karam, and Hye Young You. 2016. Lobbyists as Matchmakers in the Market for Access. Working Papar. Kelleher, Christine, and Susan Webb Yackee. 2009. A Political Consequence of Contracting: Organized Interests and State Agency Decision Making. Journal of Public Administration Research and Theory 19 (3): 579-602. Koh, Harold Hongju. 1992. Fast Track and United States Trade Policy. Brooklyn Journal of International Law 18 (1): 143-180. Kollman, Ken. 1997. Inviting Friends to Lobby: Interest Groups, Ideologial Bias, and Congressional Committees. American Journal of Political Science 41 (2): 519-544. Lohmann, Susanne. 1995. Information, Access, and Contributions: A Signaling Model of Lobbying. Public Choice 85 (3/4): 267 284. 32

McCarthy, Nolan, and Lawrence Rothenberg. 1996. Commitment and the Campaign Contribution Contract. American Journal of Political Science 40 (3): 872-904. Mian, Atif, Amir Sufi, and Francesco Trebbi. 2010. The Political Economy of the US Mortgage Default Crisis. American Economic Review 100 (6): 1967-1998. Richter, Brian, Krislert Samphantharak, and Jeffrey Timmons. 2009. Lobbying and Taxes. American Journal of Political Science 53 (4): 893-909. Ritchie, Melinda, and Hye Young You. 2016. Legislators as Lobbyists. Vanderbilt University Working Paper. Snyder, James. 1992. Long-Term Investing in Politicians; or, Give Early, Give Often. Journal of Law and Economics 35: 15 43. Stock, James, Jonathan Wright, and Motohiro Yogo. 2002. Survey of Weak Instruments and Weak Identification in Generalized Method of Moments. Journal of Business & Economic Statistics 20 (4): 518-529. Stratmann, Thomas. 1991. What Do Campaign Contributions Buy? Deciphering Causal Effects of Money and Votes. Southern Economic Journal 57 (3): 606-620. Volden, Craig, and Alan Wiseman. 2014. Legislative Effectiveness in the United States Congress. Cambridge University Press. Wright, John. 1989. PAC Contributions, Lobbying, and Representation. Journal of Politics 51 (3): 713-729. Wright, John. 1990. Contributions, Lobbying, and Committee Voting in the U.S. House of Representatives. American Political Science Review 84 (2): 417-438. Wright, John. 1996. Interest Groups and Congress: Lobbying, Contributions, and Influence. New York: Longman. You, Hye Young. 2016. Ex Post Lobbying. Vanderbilt University Working Paper. 33

A Appendix: The List of Trade-related Legislation Table A1: List of Trade-related Bills, 2003-2012 Congress Year Description Congress Year Description 108 2003 Burma Import Sanctions 109 2006 Miscellaneous Tariff Cuts 108 2003 Singapore FTA 109 2006 Approve Dubai Ports World Deal 108 2003 Chile FTA 109 2006 Reject Raising Airline Investment Cap 108 2003 Cuba Travel Ban 109 2006 Internet Gambling Payments 108 2003 Country of Origin Labeling 109 2006 Vietnam PNTR 108 2003 Computer Export Controls 109 2006 AGO, ATPA Extension 108 2003 Oppose EU GMO Ban 110 2007 Eliminate Worker Visas 108 2004 Restrict Federal Outsourcing 110 2007 Ban Mexican Trucks 108 2004 Australia FTA 110 2007 Peru FTA 108 2004 Morocco FTA 110 2007 Farm Bill 108 2004 Miscellaneous Tariff Cuts 110 2007 Defund Visa Waiver Program 108 2004 Increase Foreign Doctors 110 2007 Andean Trade Preference Act 108 2004 Cut Market Access Program 110 2007 Expand Fam Exports to Cuba 109 2005 China Currency Sanctions 110 2007 Reduce Sugar Protection 109 2005 Cuba Travel Ban 110 2008 Suspend TPA 109 2005 DR-CAFTA 110 2008 Reduce Cotton Subsidies 109 2005 Protect US Trade Laws 111 2009 Ending Offshoring Act 109 2005 Withdrawn US from WTO 111 2010 Currency Reform for Fair Trade 109 2005 Restrict Contract w/ Offshoring Firms 111 2010 US Manufacturing Act 109 2005 Defund Approval of CNOOC 112 2011 Currency Exchange Rate Reform 109 2005 Bahrain FTA 112 2012 Export-Import Bank Reauthorization 109 2005 Maintain Byrd Law 112 2012 Eliminated Sugar Program 109 2006 Study of Foreign Debt 112 2012 Russia and Moldova PNTR 109 2006 100% Container Scanning 112 2012 Farm Bill 109 2006 Orman FTA 112 2012 Applying Countervailing Duly Law A1

B Figures B.1 Ideal Points over Free Trade Figure A1: Trade Ideal Point Estimates B.2 Composition between Congress and Lobbying A2

Figure A2: Colombia Figure A3: Panama A3

C Appendix: Summary Statistics of Variables Table A2: Summary Statistics of the Variables (Democrats) Variable N Mean S.D. Min. Max Panel A. House Trade Ideal Point 204-0.82 0.56-2.52 0.43 Leadership 204 0.10 0.30 0 1 Committee 204 0.08 0.27 0 1 Tight Race 204 0.08 0.27 0 1 Union Contributions ($K) 204 860.7 511.9 0 5,392.5 Dem. Presidential Vote Share 08 204 0.66 0.11 0.34 0.95 (ln) Trade with Korea ($K) 204 12.65 0.59 9.79 14.09 (ln) Trade with Colombia ($K) 204 11.24 0.69 8.51 13.37 (ln) Trade with Panama ($K) 204 10.09 0.59 7.42 11.14 % Change in Manufacturing Employment 90-07 202-4.17 1.77-10.47 1.59 Change in Chinese Import per Worker, 90-07 202 3.45 1.87 0.65 12.32 Lobbying Contact by Korea 204 6.06 8.29 0 51 Lobbying Contact by Colombia 204 0.66 1.77 0 13 Lobbying Contact by Panama 204 1.12 3.50 0 35 Total Lobbying Contacts 204 7.85 11.83 0 98 Foreign Lobbying Contact 08-10 204 11.02 19.26 0 166 Panel B. Senate Trade Ideal Point 53-0.54 0.42-1.59 0.32 Leadership 53 0.26 0.44 0 1 Committee 53 0.07 0.26 0 1 Tight Race 53 0.26 0.44 0 1 Contributions from Union ($K) 53 811.2 798.5 36.5 5,392.5 Dem. Presidential Vote Share 08 53 0.55 0.079 0.38 0.72 (ln) Trade with Korea 53 14.54 1.16 11.83 17.01 (ln) Trade with Colombia 53 13.31 1.05 11.23 15.58 (ln) Trade with Panama 53 11.99 1.03 10.04 14.22 % Change in Manufacturing Employment 90-07 50-3.29 2.51-9.76 2.41 Change in Chinese Import per Worker, 90-07 50 3.16 1.77 0.67 7.36 Lobbying Contact by Korea 53 7.05 20.30 0 145 Lobbying Contact by Colombia 53 4.9 25.4 0 186 Lobbying Contact by Panama 53 1.8 4.81 0 26 Total Lobbying Contacts 53 13.8 49.1 0 357 Foreign Lobbying Contact 08-10 53 20.0 21.6 0 101 A4

D Appendix: Robustness Checks D.1 Measuring Lobbying Intensity by Lobbying Fees In this section, I present the results of the influence of lobbying on voting patterns when I measure lobbying intensity by money spent on contacting each member, instead of measuring lobbying intensity by number of contacts. To calculate the dollar amount spent on lobbying each member, I aggregate the total number of contacts made by each lobbying firm on behalf of each foreign country in every six-month period. Then I divide the total lobbying fees paid to lobbying firms by foreign governments for the six-month period by the total number of contacts made during the period. This creates a per-contact cost for each client country. Then I aggregate the per-contact cost at the member level, once for each foreign government and again for all foreign governments. As a final product, I now have the amount each government of Colombia, Panama, and South Korea spent to contact each member and the amount all three foreign governments spent to contact each member. I run the same analysis that produces Table 6 with a different measure of lobbying intensity - lobbying costs instead of number of contacts - and Table A3 presents the result. These results are consistent with the results presented in Table 6. D.2 Analysis with State Fixed Effect I estimate the equation (2) with a state fixed effect to control time-invariant state-level omitted variables. Standard errors are clustered at the state level. Table A4 presents the results. The main results are similar to the results presented in Table 6. D.3 Separate Analyses for the Senate and the House I run the analysis separately for the House and the Senate. Table A5 presents the results for the House and Table A6 presents the result for the Senate. A5

Table A3: Foreign Lobbying and Voting Patterns on FTA Legislation: Lobbying Intensity Measured by Lobbying Fees Country Specific Lobbying Total Lobbying Dependent Variable = (1) (2) (3) (4) (5) (6) Voting for FTA Colombia Panama Korea Colombia Panama Korea (ln) Lobbying Cost -0.0303-0.0269-0.0191-0.0321-0.0203-0.00702 (-2.11) (-1.80) (-1.62) (-3.12) (-1.75) (-0.60) (ln) Lobbying Cost Position a 0.0285 0.0233 0.0183 0.0247 0.0238 0.0176 (3.20) (2.52) (2.62) (4.25) (3.63) (2.67) (ln) Trade Volume with j b -0.00544 0.0409 0.100 0.00433 0.0423 0.0998 (-0.23) (1.50) (3.68) (0.18) (1.58) (3.71) Trade Ideal Point 0.215 0.367 0.271 0.131 0.274 0.252 (4.50) (6.56) (4.30) (2.33) (4.33) (3.95) Leadership 0.0553 0.140 0.157 0.0934 0.148 0.161 (0.78) (1.76) (1.98) (1.35) (1.90) (2.05) Committee c -0.0946 0.225 0.0897-0.0870 0.251 0.0764 (-1.14) (2.31) (0.95) (-1.06) (2.71) (0.82) Tight Race d -0.0534-0.134-0.102-0.0773-0.160-0.0935 (-0.72) (-1.58) (-1.21) (-1.05) (-1.92) (-1.12) Union Contributions ($M) -0.0996-0.0810-0.0678-0.0959-0.0803-0.0726 (-2.44) (-1.75) (-1.49) (-2.40) (-1.79) (-1.60) Dem. Presidential Vote Share 08-0.478 0.0997 0.0411-0.508-0.0120-0.0143 (-2.05) (0.37) (0.15) (-2.19) (-0.05) (-0.05) % Change in Manufacturing Emp. e 0.0361 0.0336 0.0556 0.0341 0.0347 0.0579 (2.60) (2.14) (3.56) (2.50) (2.26) (3.75) Change in China Competition f -0.0122 0.00298-0.00367-0.00930 0.00391-0.00164 (-0.83) (0.18) (-0.22) (-0.64) (0.24) (-0.10) Constant 1.000 0.227-0.521 0.806 0.123-0.578 (2.56) (0.51) (-1.16) (2.06) (0.28) (-1.30) N 252 252 252 252 252 252 adj. R 2 0.278 0.342 0.336 0.289 0.367 0.351 Note: t statistics in parentheses. p < 0.05, p < 0.01. a: Position indicates four groups depending on member i s trade ideal points. b. j {Colombia, Panama, South Korea} depending on the FTA vote. c. Committee = 1 if a member i serves on the House Ways and Means Committee or the Senate Finance Committee. d. 1 if the vote share in the most recent election is less than 0.55. e. % Changes in manufacturing employment, 1990-2007 (Autor, Dorn, and Hanson 2013). f. Change in Chinese import per worker, 1990-2007 (Autor, Dorn, and Hanson 2013). A6

Table A4: Foreign Lobbying and Voting Patterns on FTA Legislation: State-Fixed Effect Model Country Specific Lobbying Total Lobbying Dependent Variable = (1) (2) (3) (4) (5) (6) Voting for FTA Colombia Panama Korea Colombia Panama Korea (ln) Lobbying Contact -0.214-0.115-0.0552-0.138-0.0421-0.00241 (-2.19) (-1.01) (-0.91) (-5.51) (-0.81) (-0.04) (ln) Lobbying Contact Position 0.197 0.138 0.0775 0.121 0.0851 0.0645 (2.90) (2.13) (2.35) (6.47) (2.72) (1.93) (ln) Trade Volume with Each Country -0.00931 0.0360 0.0917 0.000564 0.0344 0.0884 (-0.31) (1.60) (4.47) (0.02) (1.49) (4.38) Trade Ideal Point 0.145 0.322 0.176 0.0394 0.226 0.153 (2.97) (7.01) (2.97) (0.98) (4.73) (2.46) Leadership 0.0687 0.0969 0.135 0.0952 0.105 0.132 (0.58) (0.98) (1.42) (0.81) (1.16) (1.53) Committee -0.0685 0.248 0.172-0.0778 0.293 0.137 (-0.69) (2.48) (1.54) (-0.69) (3.08) (1.25) Tight Race -0.0192-0.107-0.0956-0.0330-0.110-0.0843 (-0.23) (-0.92) (-1.06) (-0.45) (-1.02) (-0.98) Union Contribution ($M) -0.0792-0.0530-0.0128-0.0781-0.0483-0.0239 (-1.42) (-1.70) (-0.37) (-1.49) (-1.80) (-0.81) Dem. Presidential Vote Share 08-0.342 0.211 0.112-0.370 0.0874 0.0335 (-0.98) (0.63) (0.44) (-1.08) (0.27) (0.14) % Change in Manufacturing Emp. 0.0464 0.0706 0.101 0.0437 0.0687 0.102 (1.77) (3.14) (3.74) (1.61) (2.67) (4.16) Change in China Competition 0.00471 0.0299 0.0309 0.00664 0.0300 0.0297 (0.24) (2.31) (1.55) (0.38) (2.09) (1.62) Constant 0.854 0.295-0.542 0.617 0.203-0.519 (2.04) (0.82) (-1.45) (1.60) (0.54) (-1.44) State FE Y Y Y Y Y Y N 252 252 252 252 252 252 adj. R 2 0.393 0.401 0.456 0.429 0.433 0.474 t statistics in parentheses. p < 0.05, p < 0.01 A7

Table A5: Foreign Lobbying and Voting Patterns on FTA Legislation: House Democrats Only Country Specific Lobbying Total Lobbying Dependent Variable = (1) (2) (3) (4) (5) (6) Voting for FTA Colombia Panama Korea Colombia Panama Korea (ln) Lobbying Contact -0.316-0.0675-0.0766-0.148-0.0486-0.0443 (-2.62) (-0.09) (-1.36) (-3.44) (-0.92) (-0.82) (ln) Lobbying Contact Position 0.255 0.112 0.0878 0.123 0.0935 0.0845 (3.64) (1.57) (2.65) (5.06) (3.13) (2.78) (ln) Trade Volume with Each Country -0.0384-0.00486-0.0255-0.0535-0.00403-0.0261 (-1.00) (-0.09) (-0.45) (-1.44) (-0.07) (-0.47) Trade Ideal Point 0.184 0.347 0.242 0.0838 0.242 0.213 (3.91) (5.65) (3.66) (1.57) (3.63) (3.19) Leadership 0.0237 0.137 0.150 0.0568 0.119 0.132 (0.29) (1.40) (1.56) (0.74) (1.27) (1.38) Committee -0.137 0.198 0.134-0.147 0.241 0.110 (-1.62) (1.78) (1.28) (-1.77) (2.36) (1.06) Tight Race -0.166-0.236-0.148-0.158-0.211-0.136 (-1.85) (-2.15) (-1.35) (-1.81) (-1.98) (-1.25) Union Contribution ($M) -0.0910-0.0864-0.0285-0.0808-0.0750-0.0350 (-1.89) (-1.47) (-0.49) (-1.76) (-1.33) (-0.61) Dem. Presidential Vote Share 08-0.415-0.0203-0.180-0.489-0.111-0.246 (-1.68) (-0.07) (-0.60) (-2.01) (-0.37) (-0.83) % Change in Manufacturing Emp. 0.0346 0.0170 0.0395 0.0298 0.0129 0.0391 (2.23) (0.90) (2.08) (1.98) (0.70) (2.09) Change in China Competition -0.000599 0.000176 0.00390 0.00185-0.000526 0.00358 (-0.04) (0.01) (0.21) (0.13) (-0.03) (0.19) Constant 1.247 0.784 1.039 1.311 0.618 1.019 (2.34) (1.17) (1.32) (2.54) (0.95) (1.31) N 202 202 202 202 202 202 adj. R 2 0.255 0.312 0.275 0.292 0.354 0.293 t statistics in parentheses. p < 0.05, p < 0.01 A8

Table A6: Foreign Lobbying and Voting Patterns on FTA Legislation: Senate Democrats Only Country Specific Lobbying Total Lobbying Dependent Variable = (1) (2) (3) (4) (5) (6) Voting for FTA Colombia Panama Korea Colombia Panama Korea (ln) Lobbying Contact 0.0418-0.220-0.0593-0.0715-0.0859-0.00836 (0.18) (-1.06) (-0.39) (-0.47) (-0.61) (-0.06) (ln) Lobbying Contact Position -0.0160 0.161 0.0451 0.0390 0.0705 0.0431 (-0.13) (1.16) (0.55) (0.49) (0.95) (0.60) (ln) Trade Volume with Each Country -0.0237 0.0646 0.173-0.0209 0.0719 0.176 (-0.33) (0.94) (2.67) (-0.29) (1.04) (2.76) Trade Ideal Point 0.609 0.675 0.365 0.526 0.597 0.368 (2.96) (3.74) (1.79) (2.23) (2.66) (1.69) Leadership 0.0480 0.0280 0.0472 0.0691 0.0499 0.0349 (0.28) (0.18) (0.29) (0.39) (0.31) (0.22) Committee 0.0395-0.0399-0.177 0.0321 0.0169-0.269 (0.15) (-0.13) (-0.68) (0.11) (0.06) (-1.04) Tight Race 0.0948-0.0114-0.0476 0.0781-0.0344-0.0324 (0.58) (-0.07) (-0.32) (0.47) (-0.23) (-0.22) Union Contribution ($M) -0.0640-0.0480-0.145-0.0634-0.0688-0.145 (-0.69) (-0.53) (-1.69) (-0.69) (-0.78) (-1.72) Dem. Presidential Vote Share 08-0.737 0.585 0.0634-0.930 0.269 0.268 (-0.59) (0.51) (0.05) (-0.76) (0.23) (0.24) % Change in Manufacturing Emp. 0.0493 0.0836 0.0664 0.0478 0.0818 0.0748 (1.25) (2.30) (1.83) (1.21) (2.21) (2.08) Change in China Competition -0.0343 0.00868-0.0240-0.0358 0.00918-0.0166 (-0.74) (0.20) (-0.55) (-0.77) (0.21) (-0.39) Constant 1.740 0.138-1.271 1.788 0.157-1.510 (1.44) (0.15) (-1.30) (1.52) (0.17) (-1.54) N 50 50 50 50 50 50 adj. R 2 0.202 0.292 0.238 0.206 0.290 0.263 t statistics in parentheses. p < 0.05, p < 0.01 A9

E Appendix: First-Stage Results for Instrumental Variable Analysis Table A7: First-Stage Results (1) (2) DV= Endogenous Regressors (ln) Total Contact (ln) Total Contact Position Instruments (ln) Pre Contact (ln) Pre Contact Position 0.24 (5.50) 0.39 (9.84) Full Controls Y Y F-Statistics 19.17 52.94 N 252 252 t statistics in parentheses. p < 0.05, p < 0.01. Robust standard errors are used. Figure A4: Relationship between Pre Contact (IV) and Total Contact on FTA A10