Lobbying and Delegation in Financial Regulation

Transcription

1 Lobbying and Delegation in Financial Regulation Thomas Groll Columbia University Sharyn O Halloran Columbia University August 2017 Geraldine McAllister Columbia University Preliminary Version Please Do Not Circulate or Cite Abstract We explore the determinants of financial market regulation with a formal model of the policy-making process in which the legislature delegates authority to a government agency and special interests can lobby both the legislature and executive agency. We show that the mere threat of administrative lobbying by the industry may be sufficient to induce the agency to set policies preferred by the industry. Our analysis also shows that policy conflict, the difference between Congress preferred policy and the agency s implemented policy, is increasing in the agency s vulnerability to lobbying but decreasing in the interest group s lobbying cost when Congress prefers more extreme policies. Administrative lobbying either amplifies or mitigates the conflict between Congress and the agency. Relatedly, our formal findings show that the ally principle does not hold and Congress prefers an agency that is slightly more biased against the industry. Congress delegates greater discretion to the agency when policy uncertainty is higher, when policy conflict between Congress and the agency is a lower, and when administrative lobbying mitigates the policy conflict between Congress and agency. We will test our hypotheses using a comprehensive data set of U.S. financial regulatory laws and lobbying activities. The authors are grateful to Samantha Rhoads and Jadrian Guy for excellent research assistance. All errors are our own. School of International and Public Affairs, Columbia University, New York, NY; tgroll@columbia.edu. Department of Political Science and School of International and Public Affairs, Columbia University, New York, NY; so33@columbia.edu. Columbia University, New York, NY; gam2116@columbia.edu.

2 1 Introduction Ten years after the financial crisis, the U.S. economy has come full cycle. We are now in the recently unthinkable position where over-regulation is the cry and the current administration is seeking to deregulate the banking and financial services sector once again. We ask if these regulatory reforms will reflect expert opinion and constituent interests or the special interests of the banking sector they are meant to oversee. The answer to this question is important as it will likely predict the probability and severity of the next financial crisis. If expertise drives regulatory decision-making, leading to smart, efficient reforms, than the anticipated effect would be an increase in economic growth and investment, thereby freeing capital and increasing market liquidity. On the other hand, if regulatory reform is driven by exactly the same special interests as before, then the forces that lead to the collapse of the financial system in the first place will be again at play. In our analysis we focus on the legislative design of statutory laws as well as bureaucratic rulemaking pursuant to regulatory authority. We use a delegation framework of joint policymaking and derive the constraints on administrative discretion. We also recognize that interest groups may engage in lobbying at both stages of policy-making and analyze industry groups choice of lobbying venue. Interest groups may choose to engage in policy making by lobbying at the legislative stage or at the administrative rule-making stage to shape both policy making and policy outcomes. It is this variation in lobbying patterns and its impact on regulatory design that the present paper seeks to explain. In recent empirical work Boehmke et al. (2013) show that interest groups engage in both legislative and administrative lobbying and that most of these groups lobby at both levels. We incorporate these insights into our formal analysis and examine how interest groups engage in hierarchical multi-arena lobbying and derive their optimal mix of legislative and administrative lobbying activity. We then apply our analysis to a particular example, financial regulation. For example, the link between administrative lobbying and financial regulation can be seen in the number of lobbying reports and in the number of clients that lobby regulators across various agencies. Recent trends in Figure 1 illustrate tremendous changes in lobbying activity in the aftermath of the financial crisis when regulators where overhauling the rule making process. The first panel of Figure 1 shows the number of lobbying reports overtime for the four regulatory agencies, SEC, CFTC, OTC, and the OCC. The second panel denotes the number of clients 1

3 Figure 1: Trends in Administrative Lobbying. (banks, industry associations, community groups, etc.) on whose behalf the reports in panel one were filed. The last panel shows the resources associated with such lobbying activity, as denoted by the amount of money expended in millions of USD. We can see from the time series across the four agencies that the SEC and the CFTC experience the largest variation in lobbying pressure, meaning that they are highly susceptible to interest group demands. If we look at the OCC and OTC, on the other hand, we do not see much variation in the number of client reports. We do, however, see a spike in the amount of resources spent in lobbying the OCC around the passage of the Dodd-Frank legislation, indicating a greater intensity in lobbying effort during times of regulatory reform. In this current analysis, we model the delegation of regulatory authority from Congress to a bureaucratic agency. As it is standard in models of delegating policymaking to government agencies, we assume that Congress has the legislative authority over policymaking but may prefer to delegated the regulation to a better informed agency. The classical trade-off is then between the agency s superior expertise and a potential policy conflict when Congress and the agency have different ideal points in policy and rule making. We extend the classical delegation setup and consider how a regulated industry may choose to undertake lobbying efforts at both the legislative stage and the administrative stage. At the administrative stage the regulated industry may exercise costly lobbying pressure towards the agency and influence the agency s rulemaking. The agency facing a costly burden of lobbying may then choose to propose and implement a different policy that is within its delegated discretion. In an extension of the model we consider how the regulated industry chooses to engage in legislative lobbying by offering campaign contributions to the legislature in exchange for policy favors. These policy favors are then in the form of a status quo policy closer to the industry s preferred as a lower amount of delegated discretion to the regulating agency. 2

4 Our model highlights that the regulatory agency chooses a lower regulatory policy level in response to greater external policy shocks and lobbying pressure by the industry. We show that a mere threat of administrative lobbying by the industry may be sufficient to induce the agency to choose policies that are preferred by the industry. Our analysis also shows that the policy conflict, the difference between Congress preferred policy and the agency s implemented policy, is increasing in the agency s lobbying burden but decreasing in the lobby s lobbying cost if Congress prefers a higher policy level. This implies that administrative lobbying can either amplify or mitigate the conflict between Congress and the agency. Our formal findings also show that the ally principle does not hold (Bertelli and Feldmann (2006)) and Congress prefers an agency that is slightly more biased against the industry and exercises resistance against lobbying pressure. Congress delegates greater discretion to the agency when there is more policy uncertainty, when there is a lower policy conflict between Congress and the agency, and when administrative lobbying mitigates the preference conflict between Congress and agency. In an extension of our formal model we consider legislative lobbying by the industry and show that campaign contributions induce Congress to set lower regulatory policies and to delegate less discretion to the agency when Congress prefers less regulation than the agency. Campaign contributions then amplify the preference conflict between the pressured agency and Congress and constrain the agency s abilities in rulemaking. 1.1 Literature Review A primary theme in the bureaucratic design literature has been the willingness of Congress to delegate discretionary authority to executive agencies. One consistent finding, for instance, has been that Congress delegates less authority under conditions of divided as opposed to unified government. 1 In other work on the delegation question, Volden (2002a) notes an asymmetry: it is easier to raise executive discretion than to lower it, so the regulatory state tends to grow over time. Bendor and Meirowitz (2004) examine a number of variations on the standard delegation model and note in which cases the ally principle, their term for delegating more to actors with similar preferences, holds. Boehmke et al. (2006) analyze a model in which interest groups choose whether to pursue policy reform via legislative action or agency rule-making, thus restricting 1 See Epstein and O Halloran (1994, 1999), Volden (2002b), Huber and Shipan (2002), and Wiseman (2009). 3

5 agency policy making to those issues which Congress has the hardest time addressing. And Gailmard and Patty (2007) show that with endogenous agency expertise, grants of discretionary authority encourage investment in information gathering, but at the cost of a neutrally competent bureaucracy. 2 After determining how much authority Congress delegates, the next step is to examine how this delegated authority is structured within the executive branch. This is a basic issue of the industrial organization of regulatory policy making: how are agency budgets, staff, and resources determined? Where within the executive branch hierarchy are the agencies located: closer to presidential control (Executive Office of the President or cabinet-level) or more independent (independent agencies and commissions)? How do we understand the division of tasks among agencies when is a single issue split among multiple, possibly competing agencies, and when is a single agency given multiple tasks to perform? And how do all these details of agency design affect Congress s willingness to delegate authority in the first place? Some of these questions have received scholarly attention in recent years. Gailmard and Patty (2010) for instance, examine the resource question: Why would Congress have incentives to increase bureaucratic capacity even though they know that it will be used to pursue the president s policy agenda rather than their own? The answer they provide is that when the president has the option of acting unilaterally, Congress might as well give them the expertise to act in an informed rather than uninformed manner. Regarding agency location, Lewis (2003) argues that presidents will want politically responsive agencies to implement their preferred policies, while Congress will prefer to insulate agencies from outside political pressure, all the better to serve the favored interest groups that lobbied for government action. Stephenson (2006) examines the similar question of when authority will be delegated to politically responsive agencies as opposed to courts, arguing that courts are more ideologically heterogeneous across issue areas but more stable over time. And Stephenson (2008) argues that partially insulated bureaucrats actually make policy more similar, on average, to public opinion, as politically responsive politicians may tend to swing policy from one extreme to the other. 2 Some interesting technical work has been done as well on the optimal type of discretion to offer agencies. Melumad and Shibano (1991) and Alonso and Matouschek (2008) provide instances where a principal would prefer to offer a menu of discontinuous choices to an agent receiving authority. Gailmard (2009), though, demonstrates that in situations where the principal cannot precommit to certain courses of action, interval-type delegation regimes, such as that used here, are optimal. 4

6 The distribution of tasks across agencies has been examined in two papers by Ting (2002, 2003). The former analyzes the circumstances under which a legislature will choose to assign multiple tasks to a single agency, as opposed to dividing them across several different agencies. This line of research finds that agencies should be tasked with projects that are complementary, in the sense that the agency itself should prefer that they all be completed rather than concentrate on a few at the expense of others. It is under these conditions that legislatures can design rewards and punishments for the agency to extract more work for less cost. Ting (2003) provides the flip side of this analysis, exploring the incentive to give a single job to multiple agencies, thus setting them up in competition with each other. The major finding is that legislators will only rarely wish to set up competing power-centers in different agencies for the same policy, since agencies might strategically shirk, letting the other factors produce policy in the given area and concentrate their own resources elsewhere. Our study takes this process one step further, from the delegation of power to agency design and on into the agency policy making process itself. As required by the Administrative Procedure Act of 1946, executive agencies engaged in rule making activities must announce a notice and comment period, notifying affected interests of the proposed regulations and giving them a chance to comment. Furthermore, agencies must respond to all comments received, defending their proposals or changing the regulations in response to concerns raised. Otherwise, the rule making process could be judged arbitrary and capricious by the courts and the rules themselves struck down. These requirements thus formally enfranchise outside interests into the rule making process. However, few studies have formalized the role of interests in agency rule making. As mentioned in the introduction, many observers have worried that regulated interests end up effectively controlling the agencies, rather than vice-versa. Several Chicago school economists model the capture of agencies by various interests: Stigler (1971), Peltzman (1975), and Becker (1983) all examine the market for regulatory protection, assuming that legislators will accommodate industry demands for protection up to the point where the marginal votes gained from doing so equal the marginal votes lost. More recently, Gailmard and Patty (2014) provide an informational rational for semi-captured, independent agencies. They argue that in regulatory environments where it is crucial to elicit information from the regulated industry, agencies with ideal points close to the industry may be more effective in coaxing out the necessary information. 5

7 We also examine the impact of industries on regulation. In our model, however, the industry s role is different than the benign information provision of former models. 3 Rather, we assume that industries can spend resources to directly mitigate the impact of proposed regulations, for instance, by entering multiple objections during the notice and comment period or by making implicit costly threats of potential legal actions later. Our approach, then, admits for the raw exercise of industry power to weaken or negate altogether regulatory policy making at the legislative and executive stage. 4 We extend the canonical model of delegation (Epstein and O Halloran (1999)) and consider in a first step how interest group lobbying effort is costly to the industry and also possibly costly to regulators as well. We assume, that is, that regulators would rather reach a given outcome by announcing a rule and having it accepted by all parties, rather than initially propose a tougher regulation and have it weakened to the same point by industry lobbying. 5 Bennedsen and Feldmann (2006b) follow a similar starting point but model lobbying by interest groups at the agency level as menu-auction interaction between bureaucrats and interest groups. 6 Our approach of costly lobbying efforts provides similar predictions but we provide an empirical test for our predictions of costly lobbying efforts. We use the lobbying reports filed by interest groups and firms under the Lobbying Disclosure Act (1995) that actually document their costly lobbying efforts and which are not direct transfers to bureaucrats. 7 In a second step we extend our baseline model and consider the multi-tier lobbying activities of interest group lobbying at the legislative and executive stage. Mazza and van Winden (2008) analyze multi-tier lobbying as menu-auction exchanges at both levels and focus on the binary delegation choice of Congress in budget allocations. In our analysis we consider interest group s campaign contributions to legis- 3 McCubbins and Schwartz (1984) introduced the role of interest groups as watch dogs for Congress which provide potential fire alarms that inform about the bureaucracy s performance. For studies in this spirit see Laffont and Tirole (1993), Lupia and McCubbins (1994), Epstein and O Halloran (1995), Hopenhayn and Lohmann (1996), as well as Milner and Rosendorff (1996). 4 A few theoretical studies have considered the direct impact of interest groups on the choices of bureaucrats (Laffont and Tirole (1993), Sloof (2000), Bennedsen and Feldmann (2006b)). 5 We assume that any policy that is initially proposed by the agency can be influenced by interest groups lobbying efforts. However, the finally implemented policy has to be within the discretionary authority delegated by Congress. In this assumption we differ from Gailmard (2002) who considers policies announced by the agency that are outside its discretionary authority and are costly and may not necessarily be overturned by courts. 6 Many studies have focused on the direct influence of interest groups on legislative decisions. Legislative decisions may be shaped by monetary transfers or other private benefits provided by interest groups to legislators. A common modeling choice is the menu-auction approach introduced by Bernheim and Whinston (1986). For popular studies see for example Grossman and Helpman (1994, 2001). Other lobbying models have focused on the direct interactions between information provision and financial transfers (Bennedsen and Feldmann (2006a), Dahm and Porteiro (2008), Groll and Ellis (2014, 2017). 7 The empirical challenge for the menu-auction approach in this setting is lack of transparency in the exchanges between bureaucrats and interest groups as well as the legal limitations on gifts that can be received by bureaucrats. 6

8 lators as means of influence and focus on Congress choice of delegating regulatory authority by choosing a status quo policy and the explicit extent of discretion limits as constraints on agencies policy choices. Our empirical analysis of our lobbying-delegation model is related to recent studies on the political economy of financial regulation around the recent financial crisis. 8 In a series of papers Mian et al. (2010, 2013) investigate how constituents and special interests as well as legislators ideologies shaped the mortgage credit expansion in the mid-2000s as well as the implementation of the Emergency Economic Stabilization Act (2008) and American Housing Rescue and Foreclosure Prevention Act (2008). They show that policymakers responded to both constituent interests as well as campaign contributions from industry representatives. Another recent series of papers by Agarwal et al. (2014) and Lucca et al. (2014) focuses on federal and state U.S. banking regulators incentives to implement rules on banking regulations. Besides finding how institutional differences and salience about local economic conditions shape differences and offsetting actions in rule implementations between federal and state regulators, they do not find evidence that is a revolving door for regulators to banking and quid-pro-quo exchanges. This provides empirical support for our modeling choice that lobbying at the regulatory agency level does not follow a traditional menu-auction approach but rather a lobbying pressure in form of information provision, opinion and comment statements, as well as threats of legal action against regulators rulemaking. 2 Model Our model examines strategic interactions among Congress (C), an executive agency (A), and an interest group (I) representing a regulated industry. As in the standard delegation model, all actors have ideal points and quadratic preferences over outcomes in a uni-dimensional outcome space: u i (x) = (x x i ) 2 for x X = R 1. Without loss of generality we assume that x I = 0 and x A > 0. We further assume that x C > 0, so if the value of x represents the strength of regulation, the industry prefers less regulation than either Congress or the agency. Congress may delegate authority to make policy decisions to the agency, and final policy outcomes are a function of both the policy p chosen and an external shock ω according to the 8 In earlier work Groll et al. (2016) have applied a delegation framework to understand and test the political factors that impacted Congress delegation of regulatory authority to address market and systemic risk in financial markets in the time period of 1950 to

9 C chooses (p 0, d) Nature draws ω A proposes p A I chooses e p enacted, x realized Figure 2: Order of Events. equation x = p + ω. The external shock ω is uniformly distributed: ω U [ R, R]. When delegating, Congress can also place discretionary constraints on the agency s policy choice. Thus Congress can set a status quo policy p 0 and discretion limit d so that p p 0 d. Where we differ from the standard model is our assumption that the interest group can affect outcomes directly by lobbying the agency after the agency announces its proposed rule of p A. In particular, we assume that p = p A e, where e is the amount of costly effort exerted by the interest group (Richardson, 2010). We envision this effort coming in the form of presenting analyses and testimony at the notice and comment stage of rule-making, as well as broader lobbying efforts aimed at legislators, executive officials, and the public at large to weaken industry regulations. The cost to the group of this effort is c(e), with c > 0 and c > 0. For the sake of concreteness, we take c(e) = αe 2, so that α > 0 measures the relative cost of lobbying to the interest group, and low values of α indicate the ability to exert greater pressure on regulators. This lobbying also reduces the agency s utility by an amount βe, where β 0 is the cost to the agency of having its original proposals moved back towards the interest group s ideal point. It is possible to set β = 0, so that β > 0 indicates that the agency would prefer to implement a given policy outcome directly, rather than propose a tougher regulation and have the industry lobby to weaken the agency s proposal. Overall, then, u C = (x x C ) 2, u I = x 2 αe 2, and u A = (x x A ) 2 βe, where x = p A e + ω. The order of events is illustrated in Figure 2 and is the following: First, Congress sets the status quo and the discretion limit, (p0, d). Then nature draws ω which is observed by both the agency and the interest group. Third, the agency proposes its policy rule p A. Fourth, the interest group observes p A and chooses its lobbying effort level e. Finally, policy p is enacted and policy outcome x with corresponding utility levels is realized. We solve the game for its subgame perfect Bayesian-Nash equilibria. 2.1 Interest Group s Administrative Lobbying Starting at the end of the game and working backwards, for a given policy proposal p A and shock ω, the industry will set its lobbying effort e to maximize (p A e + ω) 2 αe 2. This leads to 8

10 lobbying in the amount of e (p A ) = p A + ω 1 + α. (2.1) Thus positive amounts of lobbying are exerted whenever p A + ω > 0 and it goes to zero when the agency accommodates the industry by making final policy outcomes equal to the interest group s ideal point. Note that e / p A > 0, so that the interest group spends less resources lobbying an agency with preferences closer to their own. Further, the greater the shock, e / ω > 0, the more lobbying effort the interest group undertakes. 2.2 Agency s Policy Choice Knowing the interest group s best response to the announced policy rule, e (p A ), the agency will propose policy rule p A to maximize [p A e (p A ) + ω x A ] 2 βe (p A ), yielding p A = (1 + α)(2x Aα β) 2α 2 ω (2.2) iff p A p0 d and constrained by (p 0, d), otherwise. Combining equations (2.1) and (2.2), final policy outcomes will be: x = p A e (p A) + ω = x A β 2α. (2.3) Notice that this point lies in the interval between the agency s and the interest group s ideal points as long as x A > β/2α. We refer to the term β/2α as the effective administrative lobbying pressure which is a combination of the agency s lobbying burden and industry s lobbying cost. For values of β greater than or equal to 2αx A, the agency sets p A such that x = 0 and the industry does not lobby. Consequently, if the burden to the agency of industry lobbying is high relative to the industry s lobbying cost, the agency may propose a policy rule so that the industry gets its ideal point of no regulation without the interest group having to actively lobby to obtain this outcome. This can serve as a convenient definition of agency capture: the mere threat of lobbying causes the agency to accommodate industry wishes, so that in equilibrium the industry escapes effective government control without actually having to expend resources to do so. However, for greater ideal points of policy outcomes, x A > β/2α, the industry undertakes a lobbying effort and the effective administrative lobbying pressure is not sufficient to prevent 9

11 lobbying, 0 x C x x A policy conflict β 2α x lobbying, x A β 2α 0 x x C policy conflict x Figure 3: Ideal Points, Lobbying Pressure, and Policy Outcomes. regulation. Note that the agency s regulation is then i) increasing in its ideal point, ii) increasing in the interest group s lobbying cost, but iii) decreasing in the agency s lobbying burden. 2.3 Congress Policy Choice Congress, on the other hand, would like the agency to set policy so that the outcome, net of industry lobbying, is Congress ideal point: p A e (p A ) + ω = x C, which simplifies to p C = x C (1+α) α an amount of ω. For any given value of ω, then, Congress and the agency s ideal policies differ by ( ) ( 1 + α p C p A = x A x C β ). (2.4) α 2α This expression goes to zero when x A = x C + β 2α. Thus the ally principle fails to hold in our model: Congress prefers an agency not with its own ideal point, but one biased slightly against the industry, since policy outcomes are a convex combination of the agency s ideal point and the industry s desire for no regulation and lobbying pressure. In other words, lobbying by the interest group mitigates the preference conflict between Congress and the agency if x C < x A, as illustrated on the left in Figure 3, or increases the conflict if x C > x A, as illustrated on the right, over policy outcomes x. However, Congress is unable to observe the policy shock ω and can only determine the status quo policy and delegate discretion to the agency in order to affect policy outcomes to some extent Status Quo and Discretion Congress anticipating the agency s policy rule proposal and the interest group s pressure as well as a resulting policy conflict can set its optimal status quo policy p 0 and discretion limit d. The 10

12 policy outcomes given any status quo, discretion, and external shock are p 0 + d + ω x = x if R ω < x p 0 d if x p 0 d ω x p 0 + d (2.5) p 0 d + ω if x p 0 + d < ω R. Given these anticipated policy and outcome alternatives, Congress sets (p 0, d) to maximize its expected utility such that x d p0 EU C (p 0, d) = R R The optimal status quo policy follows from (p 0 + d + ω x C ) 2 x+d p0 ( x x C ) 2 dω dω (2.6) 2R x d p 0 2R (p 0 d + ω x C ) 2 x+d p 0 2R dω EU C = 2(d R)(p 0 x C ) = 0 p 0 = x C, (2.7) p 0 R In other words, the Congress optimal status quo policy is identical to its ideal point; and when Congress desire for greater regulation increases, then the status quo policy increases. Further, the optimal discretion limit follows from EU C d = (d R)2 + (p 0 x)(p 0 + x 2x C ) R = 0. (2.8) which can be written as d = R x A x C β 2α (2.9) We can see immediately that a higher policy uncertainty, a greater R implying a wider range of possible policy shocks, yields a greater optimal discretion level. A greater preference conflict over policy outcomes, x x c, has a negative effect on discretion which limits the interest group s influence and the agency s policy proposal. The preference conflict, as the difference in Congress ideal point and the agency s ideal point, x A x C, has ambiguous effects on discretion. For example, if the agency is relatively more industry friendly than Congress, x A < x C, then a further increase in the preference difference, meaning the agency moving relatively closer to the industry, results in less discretion. However, if 11

13 the agency is slightly more biased against the industry than Congress, x C < x A < x C + β 2α, then a further preference conflict increases actually discretion such that the interest group s pressure can offset the preference conflict. Finally, if the agency is too opposed against the industry and lobbying pressure cannot move the policy sufficiently close to Congress ideal point, x A > x C + β 2α, then here again, Congress places greater constraints on the agency. The more responsive the agency is to the effective administrative lobbying pressure, β/2α, Congresses increases or decreases discretion depending on the agency s preference relative to Congress ideal point. For example, if the agency is industry friendly or slightly more opposed than Congress, x A < x C + β 2α, then the more effective administrative lobbying is by the interest group, lower α or greater β, the smaller is discretion to prevent industry capture of the agency. However, if the agency is relatively more extremist against the industry than Congress, x A > x C + β 2α, then more effective administrative lobbying can move the agency s policy rule closer to Congress ideal point and Congress sets greater discretion. The same pattern holds for policy outcomes. Policy outcomes from Congress perspective improve with greater discretion whenever lobbying decreases the conflict between Congress and the agency. On the other hand, greater discretion would decrease policy outcomes if lobbying increases the conflict between both. 2.4 Summary of Hypotheses from Model We can summarize our analysis with the following testable hypotheses. 1. Lobbying efforts are decreasing as (a) agency s ideal point becomes closer to industry s one; (b) relative lobbying costs increase; (c) the magnitude of policy shocks decrease. 2. The agency chooses a lower policy level in response to greater external shocks and lobbying pressure by the industry. 3. The industry s lobbying effort is (a) decreasing in the agency s lobbying burden; (b) increasing in the agency s ideal point; 12

14 (c) decreasing in the industry s lobbying cost if the agency s ideal point is greater than the twice lobbying pressure, and vice versa. 4. If the lobbying burden for the agency is high relative to industry cost, then a mere threat of lobbying induces the agency to set policies preferred by the industry. 5. The policy conflict between Congress and the agency is increasing in the agency s lobbying burden but decreasing in the lobby s relative cost if Congress prefers a higher policy level. Hence, lobbying can amplify or mitigate conflict. 6. The ally principle does not hold and Congress prefers an agency that is slightly more biased against industry (as resistance against lobbying pressure). 7. Congress sets a greater status quo policy when its ideal point increases. 8. Congress delegates greater discretion when (a) there is more policy uncertainty for Congress; (b) there is less policy conflict between Congress and the pressured agency; (c) its ideal point is closer to the agency s ideal point; (d) the lobbying pressure (burden and cost) mitigates the preference conflict between Congress and agency. 9. Regulatory policy outcomes are higher when (a) industry s lobbying cost is high; (b) agency s lobbying burden is low; (c) agency s ideal point is high. 2.5 Extended Model: Interest Group s Legislative Lobbying Our extension of our delegation model with multi-tier lobbying follows immediately from our delegation model with administrative lobbying above by considering the possibility that the regulated industry may want to influence the decisions at the legislative stage by lobbying Congress directly too. Here we consider that the interest group may offer financial contributions and other resources to Congress and its members in exchange for legislator favors which influence Congress 13

15 I offers m(p 0, d) C chooses (p 0, d) Nature draws ω A proposes p A I chooses e p enacted, x realized Figure 4: Model with Legislative and Administrative Lobbying. choice of the status quo policy p 0 and the amount of delegated discretion d. To expand our model, we consider a legislative lobbying stage before Congress legislative choice of (p 0, d) at which the interest group offers a menu of campaign contributions, m = M(.), in exchange for Congress choice. Figure 4 illustrates the extended model and shows that the rest of the model is identical to the previous model. As before, we solve the model backwards and describe the subgame-perfect Bayesian-Nash equilibria. Except for the first stage we can apply the best-responses from our previous analysis. The interest group s payoff can be then described by the previous function and the monetary offering such that u I = x 2 αe 2 M(p 0, d), (2.10) where the interest group offers an optimal schedule M(p 0, d) to maximize its payoff. By choosing its optimal the schedule the interest groups the previously described best-response by the agency in announcing a policy rule, p A from (2.2), and its own best-response in choosing its administrative lobbying effort, e (p A ) from (2.1). Congress realizes utility from policy outcomes as before but may now get offered campaign contributions by the interest group in exchange for policy choices, M(p 0, d). The trade-off is then between policy utility and contributions and we describe the value that Congress puts on each by λ [0, 1] and (1 λ), respectively. Congress utility can be then described by u C = λ(x x C ) 2 + (1 λ)m(p 0, d), (2.11) where a greater λ implies a greater policy focus of Congress and a lower value a greater contribution focus. If λ = 1 or if the interest group would not be allowed to offer contributions, then Congress would maximize its expected utility of (2.6) and choose (p 0, d ) as described in (2.7) and (2.9). This expected utility is Congress expected reservation outcome and Congress could always choose to implement this legislation. However, if λ < 1, then the interest group could induce Congress with contribution offers m L to choose a different status quo, p L 0, a different amount of discretion, d L, or both. Obviously Congress expected utility should be at least the same as 14

16 without contributions i.e., EU C (p L 0, dl, m L ) EU C (p 0, d, 0) and would serve Congress as outside option and expected reservation utility. As it is common for the menu-auction approach (Bernheim and Whinston (1986)) we are solving for a jointly efficient solution that maximizes the interest group s expected utility and maximizing the weighted sum of expected utilities of Congress and the interest group with the relative weights of Congress policy focus, λ, and contribution focus, (1 λ). The optimal legislation (p L 0, dl ) maximizes then λeu C (p 0, d) + (1 λ)eu I (p 0, d). (2.12) Congress expected utility follows from (2.6) and the interest group s expected utility can be described by x d p0 EU I (p 0, d) = R R (p 0 + d + ω) 2 2R (p 0 d + ω) 2 x+d p 0 2R dω 1 + α α dω x+d p0 x 2 dω (2.13) x d p 0 2R with x = x A β 2α as described in (2.3) when the agency is expected to propose p A and the lobby exerts administrative lobbying effort e (p A ). The optimal status quo policy follows then from... = 2(d R)(p 0 λx C ) = 0 p L 0 = λx c, (2.14) p 0 R In other words, the optimal status quo policy is the identical to Congress ideal point weighted by its policy focus; and when Congress desire for contribution increases, lower λ, then the status quo policy decreases and moves closer to the industry s preferred policy. This also implies that if Congress would not value policy outcomes at all, λ = 0, then it would be captured by the interest group and prefer no regulation in exchange for small contributions as it would not be expensive for the interest group to induce p L 0 = 0. If Congress would not value contributions, λ = 1, then p 0 would be chosen as before as Congress ideal point x C. Further, the optimal discretion limit follows then from... d = 2αλx C x (1 + α λ) x 2 ( + α (d R) 2 + p 0 (p 0 2λx C ) ) αr = 0 (2.15) 15

17 Α=1, Β=5,Λ= Α=Β=1,Λ= xc Α=5, Β=1,Λ= xa xc (a) α = β = 1, λ = xa xc (b) α = 1, β = 5, λ =.5 xa 0 (c) α = 5, β = 1, λ =.5 Figure 5: Preference Conflict and Delegated Discretion with (dl R)2. with pl 0 = λxc such that L 2 (d R) = 1+α λ α x 2 + λ λx2c + 2xC x. (2.16) For a better illustration we consider the two extreme possibilities as benchmark and can state L d = R 1+α α xa β 2α R xa xc β 2α if λ = 0 = d. (2.17) if λ = 1 It follows immediately that dl (λ = 0) < dl (λ = 1) = λ if xc < xa β 2α. This implies that if Congress would be more industry-friendly than the agency s implemented policy rule and outcome, then the amount of discretion would increase in Congress policy focus. However, if Congress were more industry-friendly and its contribution focus would increase, decreasing λ, the delegated amount of discretion would decrease and Congress would constrain the regulatory agency more. Furthermore, the amount of delegated discretion is increasing in Congress ideal point, xc, and in the industry s administrative lobbying burden but decreasing in the agency s ideal point and agency s administrative lobbying burden. In Figure 5 we illustrate the delegation of discretion in more general terms and plot (dl R)2 as a function of xa and xc for given parameter values for α, β, and λ. The function implies a reverse relationship with dl meaning that low values of dl imply a greater value of (dl R)2 given dl < R. For an intermediate value of λ, when Congress values policy outcomes and contributions similarly, the delegated discretion is decreasing in the preference conflict, xa xc, as well as in the values of ideal points when both have similar preferences, moving along the diagonal line in the three figures. 16

18 3 Data To test our hypotheses, we create a new dataset comprising all federal laws and agency rules enacted from 1950 to the present that regulate the financial sector, defined as state- and federallychartered banks, bank holding companies, securities, commodities, and mortgage lending institutions. The major datasets collected are summarized in Table 1. Our dataset contains data that allow us to test the implications of our model as well as the extensions (italics in Table 1) of legislative lobbying, multiple agencies, and competing special interest groups representing various interests. Policy Making Stage Unit of Analysis Data Collected Administrative Lobbying Activity Legislation Law 1)Delegation; 2)Constraints; 3)Agency Location; 4)Regulatory Index; 5)Citations Agency Agency-Year 1)Budget; 2)Personnel & Skills; Resources Agency Rule Making Interest Group Lobbying Rule Promulgated Agency-Year Legislative Lobbying Activity Committee Bills associated with law Voting Campaign Contributions Bills associated with law 3)Workload Measures 1)Increase/decrease regulation; 2)Statutory Authority; 3)Sector(s) Impacted 1)Lobbying Activity by Firm; 2)Industry Lobbying Activity 1)Committee; 2)Hearings; 3)Member Profiles; 4)Witness Profiles; 5)Interest Groups 1)Vote number, date; 2)Amendments, Final Passage; 3)Member Votes Member-Year 1) Contributions by Firm; 2) Industry Contributions Sources CQ Almanac, Library of Congress Thomas, U.S. Statutes OMB; U.S. Budget; Agency annual reports; OPM Agency websites and annual reports; Federal Register U.S. Senate; Center for Responsive Politics CQ Almanac, Library of Congress Thomas, Policy Agendas Project, LexisNexis CQ Almanac, Library of Congress Thomas, Congressional Record, Vote View U.S. Senate; Center for Responsive Politics Table 1: Stages of the Legislative Process, Data and Sources. 3.1 Legislation We rely on a sample of financial regulation laws from 1998 to the present. 9 Our data are taken from O Halloran et al. (2016) as well as Groll et al. (2016) which provide a dataset of the universe of U.S. financial regulation with 112 laws from 1950 onward. 10 In our dataset, laws belong to one 9 Our analysis considers only laws that are enacted post-1998 because of the limitations of available reports on lobbying activities under Lobbying Disclosure Act of Their data are constructed in a three-sweep process: relevant laws using the Congressional Quarterly s policy trackers; Macey et al. (2001) s comprehensive survey of banking law; and the Library of Congress Thomas Legislation in Progress tracker. 17

19 Financial Bills Passed per Congress, Average Laws per Congress = 3.58 Average Under Unified = 3.54 Average Under Divided = Unified Divided Figure 6: Laws per Congress. or more of the following categories: depository institutions; securities; commodities; insurance; interest rate controls; consumer protection; mortgage lending/government sponsored enterprises; and state-federal issues. For each law, all agencies receiving delegated authority and the position of those agencies in the executive hierarchy are recorded. This allows us to calculate the average number of regulators per law and the degree of autonomy of these regulators, as measured by the average executive order of agencies receiving delegated authority. 11 We can also describe whether the law as a whole increased, decreased, or left unchanged the level of industry regulation. To separate out those pieces of legislation with historic significance, the citations for each law in both academic literature and popular press are counted. The distribution of laws by Congress is illustrated in Figure 6. The figure shows that, on average, Congress enacts approximately three and a half financial regulation laws each congressional term. The graph also denotes when the executive and legislative branches were controlled by the same political party (Unified) and when this was not the case (Divided). Periods of unified and divided government are shown in different colors. The graph again illustrates the significant changes in legislative activity in the aftermath of the financial crisis, with legislation spiking in the 110th Congress. 11 The executive order is defined as: Executive Office of the President (including the president)=1; cabinet=2; independent agencies=3; independent regulatory commissions=4; government corporations and federally-mandated private corporations=5. 18

20 Executive Discretion Regulatory Discretion Delegation, Constraints, and Agencies Receiving Authority Financial Regulation, Delegation/Constraint Index Constraints # of Agencies Delegation Number of Agencies Year Year Figure 7: Trends in Discretion as well as Constraints, Delegation, and Number of Agencies Coding Discretion One of the key variables to measure from the formal model is agency discretion, which in turn depends on both the amount of authority delegated and the administrative procedures that constrain the exercise of this authority. To measure delegation, each law was read independently, its provisions numbered, and all provisions that delegated substantive authority to the executive were identified. Delegation is defined as giving discretionary authority to an executive actor to move policy away from the status quo. For each law in the database, the delegation ratio measures the ratio of the number of provisions that delegate to the executive over the total number of provisions. For more details on the coding, calculation, and time-trends of delegated authority for the time of 1950 to 2010 see O Halloran et al. (2016) as well as Groll et al. (2016). Twelve possible types of administrative procedures identify the constraints associated with the delegation of authority and the number of times each constraint category appeared in the summary of each law. 12 In analyzing the data we investigate the feasibility of using factor analysis to ascertain the correlation matrix of constraint categories. As only one factor was significant, first dimension factor scores for each law were calculated, converted to the [0, 1] interval, and termed the constraint index. 13 We illustrate the evolution of regulatory discretion as well as the number of regulatory agen- 12 These constraints on discretion include: appointment power limits, time limits, spending limits, legislative action required, executive action required, reporting requirement, consultation requirement, rule-making requirements, public hearings, appeals procedures, direct oversight, and exemptions. Each of these categories is coded as constraints above and beyond those required by the 1946 Administrative Procedure Act. For a detailed description of these administrative constraints and their definition, see Epstein and O Halloran (1999). 13 Total discretion was defined as delegation minus constraints that is, the amount of unconstrained authority delegated to executive actors. For a given law, if the delegation ratio is D and the constraint index is C, both lying between 0 and 1, then discretion is defined as D (1 C). See Epstein and O Halloran (1999) for a complete discussion of this measure. 19

21 cies, the number of constraints, and the degree of delegation in Figure 7. The amount of authority delegated to oversee the financial sector has remained fairly constant over time, perhaps decreasing slightly in the past decade. The trends in Figure 7, though, are due mainly to a large and significant increase in the number of constraints placed on the regulators use of this authority. In addition, we find that the number of actors receiving authority has risen significantly over the time period studied, as also shown in Figure Committees and Voting - Extended Model For each piece of legislation, we traced out its path through Congress. Thus we first identified the committees that reported the bill or bills in both the House and Senate and recorded the members of those committees. We further collected data on hearings that the committees held while considering the bills, including witness profiles and the interest groups that appeared to testify for or against the proposed legislation. Once floor action started on the bills, we collected the roll call votes associated with them, including the votes of all individual representatives and senators. If the President announced a position on the roll call (usually only on final passage), we recorded this as well. 3.2 Agency Rule Making and Resources As noted above, our analysis goes past the agency setup phase, penetrating to the actual rule making process. We therefore collect a number of datasets concerning agency resources and rule making. First, we identified nine executive agencies that promulgate regulations affecting the financial sector: the Commodities and Exchange Commission (CFTC), Federal Deposit Insurance Corporation (FDIC), United States Securities and Exchange Commission (SEC), Office of the Comptroller of the Currency (OCC), Office of Thrift Supervision (OTS) (and its predecessor, the Federal Home Loan Bank Board), Federal Reserve, National Credit Union Administration (NCUA), and the Financial Crimes Enforcement Network (FINCEN). We recorded annual budget data for each agency, as well as the number of agency personnel. 14 For heads of agencies and members of commissions, we are recording the date of appointment and the president making the appointment. We are also estimating agency workload as the percent of regulated entities contacted each year, though the consistency of these data across agencies 14 In other agencies, outside contractors can form a significant portion of the workforce. The agencies listed here, though, use few if any contractors, making official personnel rolls a fairly accurate portrayal of manpower at hand. 20

22 and over time has yet to be established. The agency s resources and staffing should indicate to which extend an agency is resilient towards industry lobbying pressure. The greater an agency s resources the lower should be the burden from lobbying efforts, smaller β in our model; whereas low agency resources would imply a much greater burden to the agency from administrative lobbying, greater β, and a lower resilience towards their pressure which implies lower regulation standards. We are in the process of listing all regulations promulgated by each agency, as recorded in the Federal Register. For each rule, we are recording the public law pursuant to which the authority for the rule is derived and its impact on the industry (regulatory/ deregulatory/ neutral). For some agencies, we can also determine the segment of the financial industry towards which the rule is crafted (broker dealers, mortgage lenders, securities, commodities, insurance, etc.). For commissions, we are collecting the votes of individual commissioners on each rule. 3.3 Interest Group Influence We also collected lobbying activity by financial industry firms and agency contacts. To determine the resources available for lobbying, we collected data on industry profits each year, both for the financial sector as a whole and as a percent of total profits. We are currently in the process of gathering data on all lobbying activities by special interests in financial markets. Following our formal model we distinguish between interest groups lobbying at the administrative and at the legislative stage Administrative Lobbying: Federal Lobbying Reports For each agency, we are collecting all lobbying reports filed under the Lobbying Disclosure Act of These lobbying reports identify the representatives of special interests, their lobbyists, the amounts and efforts spent on activities, as well as the target agencies and issues. The reported activities are describing interest groups efforts rather than any transfers or promises to regulators and are therefore a measure for undertaken lobbying efforts that we analyzed in our formal model. These efforts are also not directly beneficial to regulators if the interactions are not providing new information but are rather unsolicited attempts of influence by special interests. In Figure 8 we report the trends in lobbying activities, focusing on financial regulation in the right panel. Overall lobbying has increased over time and the same is true for lobbying on financial 21

23 Dollars spent lobbying and number of lobbyists, Spending Year Tot Lobbying Spending ($Billions) Number of Lobbyists Lobbyists Lobbying Spending ($) on Finance, Insurance & Real Estate $500,000,000 $450,000,000 $400,000,000 $350,000,000 $300,000,000 $250,000,000 $200,000,000 $150,000,000 $100,000,000 $50,000,000 $ Insurance Securities & Investment Real Estate Commercial Banks Finance/Credit Companies Misc Finance Accountants Credit Unions Savings & Loans Figure 8: Trends in Lobbying Expenditures. regulation. The variation in lobbying pressure across regulatory agencies shown in Figure 1 also highlight that lobbying activity correlates with significant changes in the rule-making authority Legislative Lobbying: Federal Campaign Contributions - Extended Model Finally, for the extended model with interest group pressure at the Congress level we collected campaign contributions by interest groups representatives to individual member of Congress. These expenditures are transfers and promises to elected members who benefit from them positively. These contributions may or may not induce policy favors by Congress. As illustrated in our model the trade-off between regulatory policies and campaign contributions is determined by Congress policy focus relative to its contribution focus. We report in Figure 9 the times-series of federal campaign contributions across various donors and industries. Contributions vary across electoral cycles, with presidential elections associated with increases in total donations. The aftermath of the financial crisis also witnessed a significant increase in campaign contributions, reflecting greater lobbying pressure at the legislative stage. 4 Conclusion We explored the determinants of financial market regulation with a formal model of the policymaking process in which the legislature delegates authority to a government agency and special interests can lobby both the legislature and executive agency. We showed that the mere threat of administrative lobbying by the industry may be sufficient to induce the agency to set policies preferred by the industry. Our analysis also highlighted that policy conflict, the difference between 22

24 Finance, Insurance & Real Estate Contributions ($) $450,000,000 $400,000,000 $350,000,000 $300,000,000 $250,000,000 $200,000,000 $150,000,000 $100,000,000 $50,000,000 $ * 2006* 2008* 2010* Commercial Banks Total Securities & Investment Total Real Estate Total Insurance Total Finance / Credit Companies Total Accountants Total Savings & Loans Total Credit Unions Total Figure 9: Trends in Campaign Contributions. Congress preferred policy and the agency s implemented policy, is increasing in the agency s vulnerability to lobbying but decreasing in the interest group s lobbying cost when Congress prefers more extreme policies. Administrative lobbying either amplifies or mitigates the conflict between Congress and the agency. Relatedly, our formal findings showed that the ally principle does not hold and Congress prefers an agency that is slightly more biased against the industry. Congress delegates greater discretion to the agency when policy uncertainty is higher, when policy conflict between Congress and the agency is a lower, and when administrative lobbying mitigates the policy conflict between Congress and agency. Our current formal analysis will be completed by an empirical exploration testing our derived formal predictions. References Agarwal, S., A. Lucca, David Seru, and F. Trebbi (2014): Inconsistent Regulators: Evidence from Banking, Quarterly Journal of Economics, 129, Alonso, R. and N. Matouschek (2008): Optimal Delegation, Review of Economic Studies, 75, Becker, G. S. (1983): A Theory of Competition Among Pressure Groups for Political Influence, Quarterly Journal of Economics, 98,