\documentclass[12pt]{article} \usepackage{amsmath} \usepackage{amsfonts} \usepackage{epsfig} \usepackage{setspace} \usepackage{lscape} \usepackage{natbib} \usepackage{epigraph} \usepackage{verbatim} \usepackage{url} %\usepackage{harvard} \bibliographystyle{agsm} %\citationstyle{dcu} %\citationstyle{chicago} %\bibliographystyle{harvard} %\bibliographystyle{apalike} %\usepackage{rotating} \special{papersize=8.5in,11in} \addtolength{\oddsidemargin}{-.5in} \addtolength{\evensidemargin}{-.5in} \addtolength{\textwidth}{1in} \addtolength{\topmargin}{-.5in} \addtolength{\textheight}{1in} \makeatletter \def\@sect#1#2#3#4#5#6[#7]#8{\ifnum #2>\c@secnumdepth \let\@svsec\@empty\else \refstepcounter{#1}\edef\@svsec{\csname the#1\endcsname. \hskip 0.4em}\fi \@tempskipa #5\relax \ifdim \@tempskipa>\z@ \begingroup #6\relax \@hangfrom{\hskip #3\relax\@svsec}{\interlinepenalty \@M #8\par}% \endgroup \csname #1mark\endcsname{#7}\addcontentsline {toc}{#1}{\ifnum #2>\c@secnumdepth \else \protect\numberline{\csname the#1\endcsname}\fi #7}\else \def\@svsechd{#6\hskip #3\relax %% \relax added 2 May 90 \@svsec #8\csname #1mark\endcsname {#7}\addcontentsline {toc}{#1}{\ifnum #2>\c@secnumdepth \else \protect\numberline{\csname the#1\endcsname}\fi #7}}\fi \@xsect{#5}} \makeatother \makeatletter \renewcommand{\section}{\@startsection{section}{1}{0mm}{-\baselineskip}{0.25\baselineskip}{\centering\normalfont\normalsize\scshape}} \renewcommand{\subsection}{\@startsection{subsection}{2}{0mm}{-\baselineskip}{0.25\baselineskip}{\raggedright\normalfont\normalsize\scshape}} \renewcommand{\subsubsection}{\@startsection{subsubsection}{3}{0mm}{-\baselineskip}{0.25\baselineskip}{\raggedright\normalfont\small\scshape}} \def\@begintheorem#1#2{\trivlist \item[\hskip \labelsep{\bf #1\ #2:}]\it} \makeatother \makeatletter \def\monthname{\ifcase\month\or January\or February\or March\or April\or May\or June\or July\or August\or September\or October\or November\or December\fi} \makeatother \renewenvironment{epigraphwidth}{\setlength{13.5cm}} \renewcommand{\thesection}{\Roman{section}} \renewcommand{\thesubsection}{\Alph{subsection}} \renewenvironment{abstract} {\begin{center}\normalsize\textsc{}% \end{center}\begin{quote}\normalsize} {\end{quote}} \renewcommand{\appendix}{\footnotesize\parindent 0cm\setlength{\parskip}{\medskipamount}\setcounter{equation}{0}% \renewcommand{\theequation}{A.\arabic{equation}}} \newtheorem{theorem}{\small\sc Theorem}[section] \newtheorem{proposition}{\small\sc Proposition}[section] \newtheorem{assumption}{\small\sc Assumption}[section] \newtheorem{lemma}{\small\sc Lemma}[section] \newtheorem{corollary}{\small\sc Corollary}[section] \newcommand{\indep}{\perp\!\!\!\!\perp} \begin{document} \begin{titlepage} \vspace*{0.2cm} \setcounter{page}{0} %\vskip 30pt \vskip 10pt \begin{center}% {\Large \sc Political Investing: The Common Stock Investments of Members of Congress 2004-2008 \\ %\vskip 1.5em% \vspace{.5cm} \par}% \vskip 1em% {\large \lineskip .75em% \begin{tabular}[t]{c}% Andrew Eggers -- Yale University \\ Jens Hainmueller -- Massachusetts Institute of Technology \end{tabular}\par}% \vskip 1.5em% % {\small First version: May 2008\\ % This version: \monthname \ \number\year \par}% % \vskip 1.0em% {\monthname \, \number\day,\, \number\year} \par \vskip .2em% \end{center}\par \begin{abstract} We examine common stock investments made by members of Congress between 2004 and 2008. We find that that the Congressional stock portfolio in this period underperforms market indices by 2-3\% per year, suggesting that members of Congress are not the savvy insiders depicted in previous research but instead are quite ordinary in their mediocrity. Members of Congress are remarkable as investors, however, in two main respects: a) they invest disproportionately in companies to which they are politically connected, and b) their investments in companies headquartered in their own districts outperform the market by over 4\% per year. Our findings suggest that informational advantages enjoyed by members of Congress as investors arise primarily from their relationships with local companies, and that widespread concerns about corrupt and self-serving investing behavior in Congress have been misplaced. % question is whether those advantages are themselves corrupt % % %the degree to which they favor companies headquartered in their own districts %and the performance of these local investments % %% an extensive, newly-collected dataset of the investment portfolios of members of Congress between 2004 and 2008. %We find that members' politically-connected investments \--- those investments that they make in companies headquartered in their districts %and companies from which they receive campaign contributions \--- outperform the rest of their portfolios, and that %members invest disproportionately in these connected companies. Investments in local companies beat the market on average. Overall, however, %members of Congress are poor investors: % fairly dismal performance: %% in contrast to an earlier study asserting that members of Congress %%beat the market by 15\% per year, % we find that that the Congressional stock portfolio in this period underperforms market indices by 2-3\% per year. %% In contrast to an earlier study suggesting that members of Congress were making lucrative trades that %While recent public attention to Congressional investing has centered on the concern that members of Congress were growing rich by %trading on political information, our findings suggest that it is perhaps the ineptitude of Congressional investors, rather than their corruption, that is worthy of attention. \end{abstract} %invest in contributors and firms they regulate partly in order to cement political deals.% but not in trading on firms they oversee through committee assignments; meanwhile, members seem to \vspace*{0.25cm} \vfill \footnoterule {\footnotesize \noindent Andrew Eggers, Post-Doctoral Fellow, Leitner Program of Yale University. Email: andrew.eggers@yale.edu. Jens Hainmueller, Assistant Professor, MIT Department of Political Science. E-mail: jhainm@mit.edu. Authors are listed in alphabetical order and contributed equally. The authors recognize Harvard's Institute for Quantitative Social Science (IQSS), who generously provided funding for this project.\\ \indent We thank Adam Berinsky, Ryan Bubb, Justin Grimmer, Michael Hiscox, Gary King, Gabe Lenz, Ken Shepsle, Alberto Tomba, Jim Snyder, and seminar participants at Harvard, MIT, Princeton, and Stanford for helpful comments. For excellent research assistance we thank Thi Theu Dinh and Seth Dickinson. We would especially like to thank the Center for Responsive Politics for sharing data. The usual disclaimer applies. \vspace*{0.2cm} \noindent\footnotesize } \end{titlepage} \newpage \setcounter{page}{1} \addtolength{\baselineskip}{0.5\baselineskip} %\epigraph{\footnotesize ``Trying to take money out of politics is like trying to take jumping out of basketball.''}{\footnotesize Bill Bradley} %\epigraph{\footnotesize ``"} % %{\footnotesize } \section{Introduction} Recent research in political economy provides substantial reason to believe that members of Congress could be extraordinarily good investors. A long and growing list of political economy papers show that firm values are sensitive to political factors, both in the US \citep{roberts1990dst,jayachandran2006je,goldman2008dpc,goldman2008b} and in other countries \citep{fisman2001evp,johnson2003cac,khwaja2005lfp,faccio2006pcf,ferguson2008bhv}. If these studies do not greatly overstate the impact of politicians on stock prices, %and if politicians at least occasionally take advantage of their % inside knowledge of future political events, then members of Congress (who presumably know the effects of at least some of their actions before the market does) should be able to handsomely profit from information arbitrage. % handsomely profiting from their political positions. In addition, recent research in empirical finance \citep{cohen2008swi} suggests that mutual fund managers with more connections to senior corporate management perform better as investors. If valuable investment information comes in part simply from interaction with senior managers, one would expect members of Congress (who spend a large amount of time asking business leaders for money or being asked by them for legislation) to be capable of soundly outperforming the average investor. In this paper, we perform the most comprehensive study to date of the common stock investments of U.S. Congress. Using financial disclosures filed between 2004 and 2008, we reconstruct the daily holdings of the 453 members of the House and Senate who reported owning U.S. stocks in this period, and we use standard methods from empirical finance to assess the performance of these portfolios against market benchmarks. What we find is that, contrary to expectations of politicians as savvy insiders, members of Congress are in fact rather poor investors. We find that the Congressional portfolio underperformed the market index by 2-3\% per year (before expenses) during the period we examine. In dollar terms, \$100 invested in an index fund in January 2004 would have yielded \$80 by the end of 2008; the same \$100 invested dollar-for-dollar to mimic members of Congress would have yielded only about \$70. % As one indication of the poor overall performance of members of Congress as investors in the 2004\--2008 period, We find underperformance not just for the average member of Congress and the average dollar invested, but also separately for various subgroups including members of the House and Senate, Democrats and Republicans, members of power committees, and groups of members stratified by wealth, portfolio size, and turnover. Performance relative to the market was if anything slightly better in 2004-2006 than in 2007-2008, suggesting that on average members of Congress did not capitalize on the unusually active role of the government in the economy during the latter period. %While our findings of overall underperformance %Our findings suggest that members of Congress are on par with average individual investors, who have been shown to underperform the market on average %While surprising in the context of expectations of politicians as savvy and self-serving, our findings suggest line findings from other studies which suggest that average retail investors that move away from indexing tend to do poorly relative to the market %\citep{barber2000thy,} and even professional investors often fail to systematically earn excess returns \citep{carhart1997pmf}. Our finding that members of Congress on average underperform market indices % of overall underperformance contrasts sharply with the sole prior study of Congressional investments, \citet{ziobrowski2004arc}, who analyze stock trades reported by members of the Senate in the period 1993\--1998. % (their sample includes about 30 Senators on average). They conclude that a portfolio constructed from members' transactions would have beat market indices by as much as 12\% per year, a performance far better than that found for mutual fund managers \citep{carhart1997pmf}, hedge fund managers \citep{fung2008hedge}, and corporate insiders \citep{jeng2003eri}. \citet{ziobrowski2004arc}'s finding has been cited in numerous articles and reports emphasizing the ethical problems of Congressional investing,\footnote{Articles and broadcasts citing \citet{ziobrowski2004arc} include \emph{The New Yorker}'s {}``Financial Page'' of October 31, 2005; {}``An Ethics Quagmire: Senators Beat the Stock Market \--- and Get Rich \---- With Insider Information,'' \emph{Washington Spectator} January 1, 2006; ``Nieman Watchdog -- Questions the press should ask,'' March 10, 2006; R. Foster Winans, ``Let Everyone Use What Wall Street Knows," \emph{The New York Times}, March 13, 2007; NPR's \emph{Marketplace} on September 17, 2009 (\url{http://marketplace.publicradio.org/display/web/2009/09/17/pm-inside-dope/}); Brody Mullins and Jason Zweig, ``For Bill on Lawmaker Trading, Delay Is Long and Short of It", \emph{The Wall Street Journal}, May 5, 2010; ``Policy, portfolios and the investor lawmaker", \emph{The Washington Post}, November 23, 2009.} was featured in testimony before the House Financial Services Committee in July of 2009,\footnote{Available at \url{http://www.house.gov/apps/list/hearing/financialsvcs_dem/ziobrowski_testimony.pdf}; accessed Sept. 8, 2010.} and has provoked calls for legislation to prevent members of Congress from investing on the basis of their political knowledge.\footnote{The ``Stop Trading on Congressional Knowledge'' (STOCK) Act has been repeatedly introduced since 2006 by Reps Slaughter and Baird. For more on policy issues surrounding stock trading by members of Congress, see \cite{George2008} and \cite{jerke2010cashing}.} %While most of our analysis takes a more comprehensive We are unable to explain the difference between our results and those of \citet{ziobrowski2004arc}. While most of our findings are based on a more comprehensive view of members' investments than was used by \citet{ziobrowski2004arc} (namely, our main analysis is based on actual positions held by members rather than a portfolio constructed solely from trades), the discrepancy between our results and those of \citet{ziobrowski2004arc} persists when we perform their precise procedure using our data. Our best explanation for the difference is that the extraordinarily large abnormal returns in \citet{ziobrowski2004arc}'s study were the result of luck rather than investing skill, although we cannot rule out a change in the circumstances of Congressional investing (e.g. the composition of the Senate, the degree of scrutiny applied to members of Congress, or changes in market conditions) or simply computational error.\footnote{Professor Ziobrowski declined to share any of the data used in his study or even such basic information as the names of the members of the Senate whose investments they analyzed.} % or simply computational error %In order to account for the differences we asked Alan Ziobrowski for more detail on his data. Unfortunately, professor Ziobrowski is unwilling to share any details, including even basic information such as the names of the 30 Senators that are included in his sample.} %One possible reason for the difference between our overall findings and those of \citet{ziobrowski2004arc} is that we use all available information to reconstruct members' actual portfolios (holdings and transactions) while \citet{ziobrowski2004arc} analyze only their transactions (based on an arbitrary assumption about fixed holding periods). %If the portfolio contains many holdings that are not or only rarely traded, then any analysis soled based on transactions is likely to be uninformative about the overall portfolio performance. Transaction based analysis then primarily speaks to the canny timing of the trades. We find little evidence of abnormal returns when we replicate \citet{ziobrowski2004arc}'s transaction-based approach as closely as possible with our data. We therefore conclude that the performance of investors in the Senate has changed from the 1990s to today, and that this is due to either a change in the circumstances % (market conditions, the composition of the Senate, or Congressional ethics) or simply an end to an uncanny run of good luck.\footnote{In order to account for the differences we asked Alan Ziobrowski for more detail on his data. Unfortunately, professor Ziobrowski is unwilling to share any details, including even basic information such as the names of the 30 Senators that are included in his sample.} Our overall findings suggest that members of Congress are not the savvy and self-serving insiders portrayed by \citet{ziobrowski2004arc} but that, in terms of overall performance, they are little different from run-of-the-mill individual investors who have been shown to perform on average at or below market indices \citep{barber2000thy,barber2008just}. Whereas the large excess returns reported in \citet{ziobrowski2004arc} have provoked public criticisms of the ethics of members of Congress, the poor performance we report may if anything inspire concern about their judgment. Members of Congress could have avoided losing on average almost 10\% of their portfolios over the 2004\--2008 period (before expenses) merely by following widely-available investing advice and investing in a low\--cost index fund. Not only would this course of action have preserved a large chunk of their portfolios, it would also have shielded them from possible criticisms about financial conflicts of interest. Instead, members chose to hold individual stocks, %including (disproportionately) companies % with which they had political connections, and we find no evidence that the financial performance of these investments justified the political risk. When we dig a bit deeper into members' portfolios, however, it becomes clear that members of Congress differ from ordinary investors in important respects. First, we find that members appear to invest disproportionately in companies to which they are politically connected \--- those companies that are headquartered in their districts or contribute to their election campaigns. On average and controlling for firm and member characteristics, members invest about 16 times as much in companies located in their constituency as in other similar companies, and about 5 times as much in companies that contribute to their campaigns. % this is based on the column 1 regression Individual investors and mutual fund managers have been found in other studies to favor local companies in their portfolios as well, but the magnitude of local bias that we find is around twice as large as that found for individuals \citep{ivkovic2005local} and over 10 times as much as that found for mutual fund managers \citep{coval1999hbh}. The strength of the bias toward in-constituency firms, and the existence of bias toward contributors, provides our first clear evidence that members of Congress are political investors. One could imagine that members' bias %away from indexing toward companies to which they are politically connected might explain the poor average performance of the Congressional portfolio: if members invested so heavily in local companies and campaign contributors % companies to which they were politically connected not for financial reasons but in order to extract electoral or financial support (perhaps by bonding themselves to these companies' interests or signaling their aligned preferences), then we might expect these investments to perform poorly. Looking at the performance of members' connected portfolios leads us to our second finding distinguishing members of Congress from other investors. % We find that members' connected investments outperform the rest of their portfolios. We find that a portfolio of holdings where the company contributed money to the member's election campaigns performs as well as the market, as does a portfolio of holdings where the company lobbied the member's committee; most remarkably, a portfolio of holdings where the company is headquartered in the member's constituency robustly outperforms the market by about 4.5\% per year. % Again, this result distinguishes members of Congress from ordinary investors. Recent studies have found that neither individual investors \citep{seasholes2009there} nor mutual fund managers \citep{coval2001gii}\footnote{\citet{coval2001gii} find a local advantage before 1985 but not since.} enjoy a performance premium on their local investments, which suggests that members' local advantage reflects unusually valuable information about the economic circumstances of local companies. Our finding that the local investments of members of Congress outperformed the market in this period in fact provides one of very few examples in empirical finance of a substantial premium arising from investor knowledge. The overall picture we present is thus a somewhat subtle one: members of Congress seem to benefit as investors from knowledge of companies to which they are politically connected (and particularly those headquartered in their districts), but they seem not to invest enough in these connected companies to make up for the mediocre performance of the rest of their portfolios. %This finding suggest that Members of Congress are unusual compared to retail investors, who also tend to skew their stock holdings towards local companies but do not earn any excess returns on their local holdings (Seasholes and Zhu, 2009). While the home bias for individual investors is primarily driven by an increased familiarity with local companies, our analysis suggests that the home bias for Members of Congress may be driven largely by the fact that Members do have value-relevant information about local stocks. % that unusually good luck . %In contrast to earlier findings, our analysis suggests that, in terms of overall investment performance, %members of Congress are less like corporate insiders, who \citet{jeng2003eri} find earn large excess returns, %and more like average individuals, whose overconfidence, loss aversion, and failure to diversify have been documented and linked to poor performance in a %growing number of studies \citep{odean1999investors,barber2000thy,barber2007all,barber2008just,goetzmann2008equity}. %The fact that members invest disproportionately in companies to which they are politically connected, and that these investments do %somewhat better than the rest of their portfolios, does lend a whiff of corrupt opportunism to their investing performance, but %overall we are left with the impression that members of Congress are more remarkable for their mediocrity as investors. %%the overall impression we are left with is of a set of investors who are quite pedestrian % quotidian %%in their mediocrity. After describing our data in the next section, we assess the overall performance of the Congressional portfolio and subsets thereof, comparing this performance to that documented in other studies including \citet{ziobrowski2004arc}. We then divide the Congressional portfolio according to connections between companies and members and assess how much members invest in connected companies, how well these investments perform, and what that suggests about members' interactions with firms to which they are politically connected. % We then %place our results in the context of previous work on Congressional investing and other types of investors and consider possible explanations for the %performance of members' portfolios. We then conclude by weighing some implications of our findings. %The investment record of Members of Congress is not merely a story of overconfidence and mundane underperformance. % Connected portfolio performance; weights in the portfolio. Able to confirm that it's not just that these companies perform well. % %On the other hand, the unimpressive average performance %of members of Congress suggests that a set of powerful individuals with responsibility for American fiscal and economic policy have %largely failed to internalize the lessons of decades of empirical finance about the futility of stock-picking. Perhaps more damning, %while the overall performance of Congressional portfolios is poor, members appear to do better when they invest in companies %that are politically connected to them \--- companies that are headquartered in their constituency, companies whose PAC contributed money to their election %campaigns, and %% introduction needs to handle: %% policy relevance, previous findings, very briefly, description of what we do, implications. \begin{comment} Research in empirical finance tells us that average investors do poorly relative to the market \citep{barber2000thy} and even professional investors fail to systematically earn excess returns \citep{carhart1997pmf}. In this paper, we examine the portfolio choices and investment returns of members of Congress, a class of investors whose political roles put them in an informationally privileged position but also impose unusual constraints on their investments. In contrast to a previous paper analyzing Congressional investments using older and less complete data \citep{ziobrowski2004arc}, we do not find that members of Congress on average beat the market in the period 2004-2008. We do however find intriguing differences in returns across members and types of investments, which, combined with evidence that members tend to overweight local companies and campaign supporters in their portfolios, improves our understanding of how politicians trade off conflicting financial and political incentives while in office. %% AE: this needs to reflect more of the new findings -- the poor overall performance Why study the investments of members of Congress? %% AE: prob not nec to motivate in this way? Leaving aside the considerable public interest in the question of whether legislators personally benefit from their political positions (in violation of ethics rules), the investment choices and performance of members of Congress may provide political scientists with valuable indirect evidence about the way firms and legislators interact. Like all investors, members of Congress presumably invest in stocks in order to preserve and increase their wealth. But members of Congress differ from other investors in two principal ways that guide our investigation. First, they may possess unusually valuable market-relevant information about public companies and the regulations that affect them. Second, their political success depends to some extent on soliciting political support and campaign contributions from corporations and their stakeholders, and their investments (which after all are public) may play a role in establishing connections with public companies. In short, the investment choices and investment performance of members of Congress provide valuable clues about what kind of informational advantages members may possess and how they use those advantages, %the extent to which members possess and use informational advantages, as well as whether and in what ways they use their investments to cement political relationships with firms and their stakeholders. %the way in %which they trade off these advantages against political considerations including the desire to solicit corporate support. %This is work in progress, but Our findings indicate that members of Congress do not on average outperform market indices, in contrast to the sole previous study of Congressional stock investments \citep{ziobrowski2004arc}. But members appear to invest in a way that reflects political considerations, strongly overweighting local firms, firms that give campaign contributions, and firms with business before the member's own committees. The performance of these connected portfolios also diverges from market benchmarks: members' local investments outperform the market by about 4\% a year, indicating substantial informational advantages from proximity and relationships with supporters, while their investments in contributing companies underperform the market by about the same amount, suggesting that members invest in contributors in part to cement political relationships. \end{comment} %MORE HERE. %\section{Preliminary Discussion: What is Political about Politicians' Investments?} %Despite evidence that both amateur and professional investors do not systematically beat market indices, %typical investors underperform the market (as well as theoretical arguments about market efficiency), %recent research in political economy provides substantial reason to believe that members of Congress could be extraordinarily good investors. %A substantial and growing list of papers show that firm values are very sensitive to political factors: %\begin{itemize} %\item \citet{roberts1990dst} finds that the death of the ranking Democrat on the Senate Armed Service Committee resulted in lower stock value of firms located in the Senator's state and higher stock values of firms connected to the Senator who inherited his position on the committee. %\item \citet{jayachandran2006je} finds that the market value of Republican-connected firms dropped when Senator Jeffords unexpectedly departed the Republican Party in 2001, shifting the Senate majority to the Democrats. %\item \citet{goldman2008dpc} and \citet{goldman2008b} show that companies that announce the appointment of a politically-connected director experience a positive abnormal return and that politically connected firms are more likely to secure procurement contracts. %\footnote{In related work, \citet{agrawal2001sod} find that companies that have larger dealings with the government also have a larger number of board members with political experience and \citet{kroszner1998igc} demonstrate that interest group political action committees (PACs) donate more to politicians who are members of committees overseeing their industries.} %\end{itemize} %Comparable evidence abounds for other countries as well \citep{fisman2001evp,johnson2003cac,khwaja2005lfp,faccio2006pcf,ferguson2008bhv}. %The picture presented by all of these studies is that politicians can %significantly impact firm values. Presumably, politicians know about the impact of their own actions and those of other politicians with whom they work. If these %studies do not greatly overstate the impact of politicians on stock prices, an investment-minded member of Congress could handsomely profit from information arbitrage. %Politicians may enjoy additional informational advantages simply by being in close contact with corporate executives and industry lobbyists as part of their legislating and fundraising routines. %%While overall dismissive of the opportunity to achieve sustained excess returns, some r %Recent research in empirical finance suggests that mutual fund managers do better %when they invest in companies to which they are connected %through geographic proximity \citep{coval2001gii}\footnote{Although not in recent years; see below.} or personal ties to executives \citep{cohen2008swi}. Members of Congress necessarily have large personal networks %and frequent contact with corporate executives and lobbyists. Whether the firm approaches the legislator asking for policy favors or the legislator %approaches the firm asking for campaign donations, the firm may reveal information about its market prospects (either intentionally or unintentionally) that %the legislator could act on in her own investments. %% generally, there is a problem here of where to get into this theory. the standard thing would be intro brings up issues quickly, related lit, then "model" %%\begin{comment} %%The prospect that members of Congress may be good investors is bolstered by \citet{ziobrowski2004arc}, %%which concludes based on an analysis of transactions reported 1991-1996 that Senators experienced large abnormal returns. %%%based on analysis of transactions reported 1991-1996. %%As an indication of the uncanny timing exhibited by Senators in the period they consider, %%stocks that Senators sold outperformed the %%market by roughly 25 percent during the 12 months prior to the sell date and %%remained fairly flat thereafter, while stocks that they purchased %%beat the market by only 3 percent prior to the buying date and by %%almost 28 percent in the following year. %%%In subgroup analysis they fail to find a difference in the returns %%%of Democrats and Republicans, and if anything junior Senators %%%appear to be smarter traders. %%The Ziobrowski et al paper generated attention in the media% %%\footnote{The study was cited on the \emph{New Yorker}'s {}``Financial Page'' %%of October 31, 2005; it was described in a \emph{Washington Spectator %%}article on January 1, 2006, {}``An Ethics Quagmire: Senators Beat %%the Stock Market\textemdash{}and Get Rich\textemdash{}With Insider %%Information''; and it was featured on {}``Nieman Watchdog -- Questions %%the press should ask'' on March 10, 2006.% %%} and in Congress itself.\footnote{The ``Stop Trading on Congressional Knowledge'' (STOCK) Act was introduced in 2006 as H.R. 5015 by Reps Slaughter and Baird and reintroduced %%in 2007 (by the same members) as H.R. 2341. The bill proposed to prohibit %%members of Congress, congressional staffers, and members of the executive %%branch from trading on {}``material non-public information,'' defined %%as information members acquire as a result of their employment by %%the federal government. For more on policy issues surrounding stock %%trading by members of Congress, see \cite{George2008}.} Their explanation for Senators' %%outstanding performance is that they appear to be %%profiting from their privileged access to information and their influence over policy outcomes. %%\end{comment} % where should we talk about ZIobrowski? too many places as it is. depends on whether we want to have a related lit section or waht. %While members of Congress likely enjoy considerable information advantages because of their political power, %they also face a number of constraints arising from their need to appeal to political constituents. %We focus on three such constraints, which we will refer to as ``signaling," ``bonding," and ``ethics." %%In addition to the unusual advantages members of Congress likely enjoy because of their political power, %%there are also a number of constraints that they face arising from their need to appeal to political constituents. %%We focus on three such constraints, which we will refer to as ``signaling," ``bonding," and ``ethics." %% the signaling value of their portfolio choices and the % do not apply to other investors. %%We focus %%We point out two other ways in which investments might be related to politics: signaling and incentive alignment. %\subsubsection*{Signaling} %Members' investments are public (which is why this paper is possible) and occasionally subject to %journalistic scrutiny (e.g., \citet{boller1995}). To voters, firms, and other politicians, %a member's stock holdings may convey a signal about the member's policy preferences or ideology. %Suppose politicians agree on the expected return of a particular tobacco company's %stock but differ in their opposition to tobacco companies: some are pro-tobacco and some are anti-tobacco. %Anti-tobacco types experience a higher private cost of owning tobacco stock, perhaps because % they dislike the idea of indirectly supporting the firm or %of the cognitive dissonance of having a financial stake in a company they dislike. % or because (in the case of a politician) he is involved in a repeated game with anti-tobacco voters. %this is a bit murky? %Suppose that there is also a group of constituents who oppose the tobacco industry and want to elect an anti-tobacco politician. %%Constituents also have types and they want to vote for politicians of the same type. %%A group of pro-tobacco constituents wants to elect a pro-tobacco representative but cannot directly observe politicians' types. %If the private cost to anti-tobacco politicians of owning %tobacco stock is high enough, there may be a separating equilibrium where only pro-tobacco politicians choose %to own the stock and the anti-tobacco constituents vote for politicians who refuse to own tobacco stock.\footnote{In terms of Spence's educational signaling model, the constituents are the employers, the politicians are the workers, \emph{not} buying stock is the costly credential, and the anti-tobacco politicians are the high ability types for whom the credential is less costly.} %because they face lower private costs of doing so, and pro-tobacco constituents support stockholding politicians. %%\begin{comment} %%Signaling would suggest that politicians from different parties would own different portfolios. %%Consistent with this, our data shows that Republicans are much more likely to hold stock in Altria, Exxon Mobil, and Monsanto, among other companies; this is consistent with %%the fact that these firms are frequent targets of attack from environmental and public health interests of the political left. %%Table \ref{tab:partisan_companies} lists the top five Democratic and Republican companies, ranked according to the z-score on the log odds ratio comparing the number of Republicans and Democrats holding the stock \citep{monroe:fwl}. Portfolio choices would probably be less correlated with party if they remained private. %%%If portfolios were private, presumably politicians would be more likely to agree on %%%To put the ``signaling" idea a bit more prosaically, Democratic members would %%%probably face harsh criticism if they owned these stocks; if they were private citizens no one would know that they owned %%\end{comment} %In addition to signaling political ideology, investments might be thought to signal preferences about public service versus private gain: %by abstaining from using political knowledge to bank windfall profits, politicians may signal to voters that they prioritize public service. Consistent with the idea that members constrain themselves in order to send a signal to voters, \citet{ziobrowski2004arc} find smaller abnormal %returns in 1996 and 1997, which they suggest may be due to the unfavorable %media attention drawn to well-timed stocks trades among members by \cite{boller1995}. %To the extent that signaling is indeed a significant determinant of members' investment portfolios, we might expect %average returns to be modest, since basic portfolio theory tells us that restricting possible %investments (particularly based on non-financial considerations) cannot enhance returns. %CITE % %\subsubsection*{Bonding} %The second political constraint on members' investment decisions also comes from their relationship with voters and constituent firms. %%Politicians may also take on equity stakes specifically to align their policy preferences with the preferences of constituents. %As is widely discussed in the political economy literature, %CITES %politicians face a commitment problem with respect to voters and potential campaign donors. %Suppose that a firm is considering offering a campaign contribution to a politician, but is unsure of whether the politician %will pursue its interests in the legislature. This is clearly an incomplete contracting situation: it is impossible to write down all of %the ways in which the politician could serve or not serve the firm's interests, and at any rate courts would not enforce such a \emph{quid pro quo}. %The firm may then find it beneficial to require the politician to take an equity stake in the company, bringing the firm's and the politician's %policy interests into closer alignment.\footnote{It is standard for corporate directors to be required to %own large equity stakes in the companies on whose boards they serve in order to reduce slack in the shareholder/director relationship. Directors %are usually contractually required to hold the stock, which brings us back to the commitment problem here: it may not be time consistent for the politician to continue holding the company's stock once the firm's check has been cashed.} %CITE} %In short, members may take on equity in constituent companies in order %to make policy promises credible. % and possible a political exchange of support for representation. %If the politician is able to influence policy to help the firm in a way the market did not anticipate, we might expect these investments to be profitable; %in an inefficient market, though, there can be no such expectation: since the politician engages in the informal contract %principally to earn political returns, it may be that the \emph{ex ante} financial returns are nonexistent or negative. %\subsubsection*{Ethics} %The final political constraint is ethics regulation. Members of Congress face no special restrictions on their investment choices (other than the requirement to file annual disclosures), %but ethics rules state broadly that members should not financially profit from their political positions (\emph{Code of Conduct}, 2005). %\citep{code_of_conduct}. %A member of Congress who invested very aggressively might face ethics charges in addition to journalistic scrutiny \citep{boller1995}. \section{Data}% and Methods} Our study is based on common stock holdings and transactions reported by members of the U.S. Senate and House of Representatives between January 2004 and December 2008. As a result of the 1978 Ethics in Government Act, members of Congress are required to disclose their stock investments (as well as real estate and other investments, liabilities, and outside income and employment), as well as those of spouses and dependent children, in annual filings known as Financial Disclosure Reports.\footnote{Our analysis includes all holdings and trades reported by members, including those owned by spouses and dependent children. Members may also choose to create qualified blind trusts, which are managed on their behalf and whose holdings are unknown to the member. In our data 20 members report qualified blind trusts.} This paper is the product of using these reports to reconstruct members' actual portfolios %(a painstaking process described in detail below) and evaluating the performance of those portfolios using standard methods from empirical finance. % in order to evaluate portfolio performance. % of the aggregate Congressional portfolio and relevant subsets thereof. \subsection{Reconstructing Portfolios from Disclosure Forms} Members of Congress are required to submit Financial Disclosure Reports (FDRs) each spring, disclosing their year-end holdings as well as all transactions made during the year. Since 2004 the Center for Responsive Politics (\texttt{www.opensecrets.org}) has transcribed the FDRs, and since 2008 they have made this data freely available. We thus received the data as a pair of spreadsheets, one with a row for each of the 111,101 transactions recorded and another with a row for each of the 169,828 year-end holdings recorded. % A considerable amount of work remained, however, to reconstruct members' portfolios from this data. %After matching the named assets and transactions to publicly listed companies (and manually checking apparent financial stocks in order to identify and remove %savings accounts, IRAs, and insurance policies), we imputed The first task in converting this raw data to stock portfolios was to identify the companies in which members hold stocks. The FDRs do not identify holdings in standardized ways (e.g. an investment in Bank of America common stock may be described as ``Bank of America," ``Bank America Common Stock," ``Banc of America," or ``BOA"); we used search utilities provided by Google Finance and the Center for Research on Security Prices (CRSP) to link variously described assets to actual companies. Even more challenging, the descriptions may not precisely distinguish between stock holdings and other types of assets such as corporate bonds, mortgages, or bank accounts. To reduce the risk of misclassifying savings accounts as stock investments, we hand-checked the FDR for each apparent financial stock to determine whether other clues (such as columns reporting dividend or investment income) could distinguish stocks from other types of assets. The next task was to impute a dollar value for each holding and trade reported. The law requires only that %% AE: check that XXXX members report the value of their investments in broad value bands (e.g. \$15,000 \--- \$50,000) rather than exact dollar amounts.\footnote{Value band cutpoints are at \$1,000, \$15,000, \$50,000, \$100,000, \$250,000, \$500,000, \$1,000,000, \$5,000,000, \$10,000,000 and \$25,000,000, and a top category captures all investments of \$50,000,000 or more in value.} In order to impute precise values for investments reported in these bands, we took advantage of the fact that we do know the precise value of a sizable minority of reported investments \---- those cases in which a member submitted an annual statement from a bank or investment manager rather than filling out the official forms.\footnote{This information is available for about 25\% of the transactions in the dataset and about 8\% of the year-end holdings.} We used these investments to fit a distribution of precise values and, for each investment for which we know only the band, %used the distribution's band mean as value for imputation. an imputed investments in each band, we impute the expected value of the precise\--value distribution within that band.\footnote{This approach is inspired by the imputation method proposed in \cite{milyo1999electoral}.} %\footnote{By comparison, Ziobrowski et al impute the midpoint of each band, which is higher than our imputed value for every band except the lowest.} For the highest band (investments over \$50,000,000), of which there are fewer than 100 holdings and 5 trades in our estimation sample, we impute the value of \$50,000,000.%Because we have too little data to fit an accurate distribution, and in order to curb the influence of extremely large holdings. % \footnote{By comparison, Ziobrowski et al top-code at \$250,000, which as we show in the appendix impacts estimates of average returns (particularly value-weighted returns) but does not change the overall conclusions.} % (To follow previous work on Congressional investing, we focus only on stocks traded in the major U.S. exchanges: NYSE, NASDAQ, and AMEX.) %% does this deserve more entries? Having linked each holding and trade to a company and imputed dollar values, it remained to reconstruct the day-by-day stock portfolio. Our approach in reconstructing a portfolio from the FDRs was to start at the last day of each year, for which the reports provide the entire portfolio (i.e. the year-end holdings), and work backward to the beginning of the year, adjusting the portfolio each day to reflect purchases and sales as well as fluctuations in value due to security price changes. (In other words, each portfolio is rebalanced on a daily basis.) For example, suppose a member reported holding \$10,000 of stock in Company A at the end of the year and reported purchasing \$5,000 of stock in Company A on June 1. This member's portfolio on January 1 of that year is estimated by calculating what \$10,000 in Company A stock was worth on June 1 (based on the return between June 1 and the end of the year), subtracting \$5,000, and then calculating what that value was worth on January 1. In this way we calculate dollar value holdings for every member of every stock on each day between January 1, 2004 and December 31, 2008. \subsection{A Glimpse at Congressional Portfolios} \begin{comment} Figure \ref{fig:kerry} provides a glimpse of the reconstructed portfolio for one of the largest investors in Congress, John Kerry (D-MA). The top panel plots the daily total portfolio value as computed by our approach; the bottom panel records cumulative purchases and sales starting January 1, 2004. The overall trajectory of Kerry's portfolio value reflects the market's gradual rise until the middle of 2007 followed by decline culminating in the crash at the end of 2008. % the broad market in the period we examine: a gradual It also shows episodes where Kerry purchased or sold a large amount of stock: jumps in March and June of 2007, and drops in May of 2007 and March of 2008, reflect large net buys and sales (respectively), as seen in the lower panel. In addition, we see a sharp jump on January 1, 2006, and a sharp drop on January 1, 2008. In some cases, we see jumps between years like this because a member reports some holdings in one year that he neglects (presumably unintentionally) to report in another; in other cases, such jump might happen because of the limitations of imputing precise values based on value bands. (The apparent drop of about \$40,000,000 in Kerry's portfolio value from 2007 to 2008, for example, could happen if about ten holdings in Kerry's portfolio declined in value from just above \$5 million at the end of 2007 to just below \$5 million at the end of 2008; even if the actual value drop were fairly small, the jump from one band to another would create a large apparent drop at the beginning of 2008 because of the large difference between the imputed values for investments in those bands.) \end{comment} Our data covers FDRs from 650 members who served in the House and Senate between 2004 and 2008. Of these members, 453 reported holding a stock listed on NYSE, NASDAQ, or AMEX at some point during that period. Overall the dataset includes 30,859 reported end-of-year holdings and 48,325 reported transactions. A total of 2,581 companies are represented in the dataset; together these companies make up about 94\% of the total capitalization of these three exchanges over our sample period. % The holdings and trades we analyze include those reported by members but owned by spouses and dependent children. Table \ref{tab:summarystats} provides summary statistics describing the portfolios of the 453 members of Congress whose investments appear in our dataset. % or the annual averages of the Congressional portfolios of the 453 active members over the 2004-2008 period. The average annual portfolio size ranges from \$501 (for a member who reported a single stock in the lowest value band) to \$140 million, the average reported by Jane Harman.\footnote{The performance of Jane Harman's portfolio was unusually poor, largely due to a \$50 million position in Harman Industries that dropped about 1/3 in value in January of 2008. Because of the large size of her portfolio and the consequent large downward influence of her performance on aggregate excess returns, we exclude her from subsequent analyses unless otherwise noted. Estimates with Harman included (available from a web appendix) not surprisingly show lower returns, especially for the aggregate (value-weighted) estimates.} % has the largest portfolio size and her portfolio alpha return underperforms the market index by -1.15 per month (see Figure \ref{fig:member_returns} below) she exhibits an usually large downward influence on the aggregate congressional portfolio we construct below where portfolio returns are aggregated across members. Unless otherwise reported we therefore exclude Jane Harman from all analysis. In a web appendix we report all estimates with Harman included and the results look very similar, except that the overall aggregate performance is lower as expected. The bad performance of Harman's portfolio is largley explained by a \$50 million position (28\% of her portfolio) in Harman Industries which on January 14, 2008 dropped about 1/3 in value. See ``Harman Shares Tumble After Forecast," \emph{Reuters}, Jan 14, 2008.} % who is closely followed by John Kerry \$116 million John Kerry. The distribution of stock holdings is strongly skewed: the median member on average holds stocks worth about \$80,000 in 5 stocks, while the average member holds about \$1.6 million in 18 stocks. The right panel describes transactions reported by members in our dataset and aggregated to the yearly frequency. %shows the reported sell and buy transactions aggregated to an annual average for each member. This distribution is equally right-skewed: the average member buys and sells an average of 18 and 22 stocks per year (respectively), worth \$397,000 and \$611,000; the median member buys and sell 2 and 3 stocks worth \$16,000 and \$39,000. % (median) member buys on average about 18 (2) stocks worth about \$397,000 (\$16,000) and sells 22 (3) stocks worth about \$611,000 (\$39,000). The presence of a number of very large portfolios in the data suggests that conclusions about the performance of Congress as a whole will be sensitive to whether individual-level performances are weighted equally across members or by portfolio size; as described below, we do both. %% here we should ad numbers on average turnover probably %%The figure makes clear that the distribution of stock holdings in Congress is strongly skewed to the right. The majority of members in our dataset repor %%stock holdings of less than \$100,000, and 189 report holdings below \$50,000; on the top end, sixty-nine members report annual stock portfolios %worth more than \$1,000,000, and thirteen report portfolios worth more than \$10,000,000. The number of holdings reported per year is also skewed: half of the %members in our dataset report fewer than 5 holdings per year, while seventeen report more than 100. \section{Do Members Beat the Market?} We now turn to the task of assessing the performance of the common stock investments of members of Congress between 2004 and 2008. \subsection{Methods} % \subsection{Portfolio Returns from Portfolio Holdings} To compare Congressional stock portfolios to the market benchmark, we adopt the standard (e.g. \citet{barber2000thy}) calendar-time approach of regressing the time series of risk-adjusted portfolio returns (i.e. the return on the portfolio minus the risk-free rate) on a set of controls including the return on a market index; the intercept of this regression is interpreted as the ``abnormal return," or the average amount by which the portfolio outperforms the market index. %We construct daily portfolio returns following the standard methodology widely used in empirical finance (see for example in \cite{barber2000thy}). \begin{comment} A portfolio on trading day $t=1,...,T$ %The portfolio of a member or subset of members over time can be expressed as a vector $c \times \mathbf{w_t}$, where $c$ is the total portfolio value and $\mathbf{w_t}$ is a vector of non-negative holding weights (summing to one) whose length is equal to the universe of companies which we index by $j=1,...,J$. We have $T=1,259$ trading days in our five-year period and $J=2,581$ unique companies reported in the FDRs.\footnote{We assume that holdings on each day are beginning-of-day holdings and trades are made at the end of the day.} %Element $W_{dj}$ of which indicates the dollar value of the portfolio's holdings in company $j \in J$ on day $d$. Let $\mathbf{w}$ be the $T \times J$ matrix of portfolio weights over time. We obtained daily returns (including distributions) from CRSP for each of the 2,581 companies in our data set over the five-year period. These returns constitute a matrix $\mathbf{r}$ of the same dimensions as $\mathbf{w}$, a representative element $r_{tj}$ of which indicates the return of company $j$ on day $t$. Using this notation, the return on a portfolio over a set of trading days $\{1, 2, \dots T \}$ can be expressed as % on day $d$ \[ \mathbf{R}^p = \{\mathbf{w_1}\cdot\mathbf{r_1}', \mathbf{w_2}\cdot\mathbf{r_2}', \dots,\mathbf{w_T}\cdot\mathbf{r_T}'\},\] a vector whose representative element $R^p_t$ indicates the portfolio return on day $t$. % $p_d$ for a portfolio $\mathbf{w_d}$ on can be expressed as $\mathbf{w_d}\cdot\mathbf{r_d}$. % , where $\mathbf{W_d}$ and $\mathbf{R_d}$ are $|J|$-length vectors indicating portfolio values and returns, respectively, for each company on day $d$. \end{comment} % For the analysis that follows, w We consider two ways of characterizing the ``Congressional portfolio.'' One approach is to evaluate the return on what we call the aggregate Congressional portfolio, which is a portfolio that matches dollar-for-dollar every investment made by a member of Congress. \begin{comment} In terms of our notation, the return on the aggregate Congressional portfolio on a given day $t$ can be expressed \[R_t^\textrm{AGG} = \frac{\sum_{i=1}^N c^i \times \textbf{w}^i_t}{\sum_i{c_i}} \cdot \textbf{r}_t',\] i.e. as the return on the value-weighted average portfolio across members (who are indexed by $i$). \end{comment} The other approach is to evaluate the return on what we call the average Congressional portfolio, which is constructed using the average portfolio weight for each company across members at each point in time. %average return across members, which can be written %\[R_t^\textrm{AVG} = \frac{\sum_{i=1}^N \textbf{w}^i_t}{N} \cdot \textbf{r}_t',\] %i.e. as the return on the average portfolio across members. The average portfolio reproduces the investment return of the average member, while the aggregate portfolio reproduces the return of the average dollar invested by members of Congress. % Congressional portfolio more heavily weights the investment choices % what we call the average Congressional portfolio, which is a portfolio that and the other reflecting the average performance of Congressional portfolios (i.e. for a given time period $t$, $R_t = \frac{\sum_{i=1}^N \textbf{w}_i \cdot \textbf{r_t}} \frac{N} $ % \subsection{Calculating Abnormal Returns} For evaluating portfolio performance we calculate excess returns using two standard models widely used in empirical finance. In the CAPM model, the market index is the only control; i.e. we fit the regression \[ R_t - R^f_t = \alpha + \beta\big(R^m_t - R^f_t\big) + \epsilon, \] where $R_t$ is the return on the portfolio, $R^m_t$ is the return on a market index, and $R^f_t$ is the ``risk-free rate" or return on U.S. Treasury Bills (all in time period $t$); the intercept $\alpha$ is the estimate of the average excess portfolio return across time periods. In the Carhart Four-Factor model (an extension of the Fama-French Three Factor Model), we add controls that track the return on passive portfolios noted in the empirical finance literature for diverging the overall market; i.e. we fit the regression \[ R_t - R^f_t = \alpha + \beta_1\big(R^m_t - R^f_t\big) + \beta_2 \mathrm{SMB}_t + \beta_3 \mathrm{HML}_t + \beta_4 \mathrm{MOM}_t + \epsilon, \] where $\mathrm{SMB}_t$ is the return on a hedged portfolio that is long in small companies and short in big companies (``small-minus-big"), $\mathrm{HML}_t$ is the return on a hedged portfolio that is long in high book-to-market companies and short in low book-to-market companies (``high-minus-low"), and $\mathrm{MOM}_t$ \citep{carhart1997pmf} is the return on a hedged portfolio that is long in companies with the best performance in the previous year and short in the companies with the worst performance in the previous year. We obtained each control series from Kenneth R. French's website.\footnote{\url{http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html}} % The intercept of this regression, $\alpha$, is interpreted as the ``abnormal return" \--- the average amount by which the portfolio's return exceeds the market return in the period considered, after subtracting out the risk-free rate. %We focus on two models. In the CAPM model, the risk-adjusted return on the %value-weighted market portfolio ($R^m_t - R^m_t $) is the only control.\footnote{In our notation, $P^m_t$ can be expressed as $\mathbf{W}^m_t \cdot \mathbf{R_t} / \sum_\mathbf{W^m_t} $, where $W^m_jt$ is the market capitalization of company $j$ at time $t$.} %In the Carhart four-factor model, the controls include not just the risk-adjusted market return but also the return on %a zero-investment size portfolio ($SMB_t$), a zero-investment book-to-market portfolio ($HML_t$) and the Carhart (1997) momentum factor ($MOM_t$). Following standard procedure in the literature, we conduct our portfolio regressions at a monthly frequency; we thus aggregate daily portfolio returns to the monthly level before each regression. \citet{barber2000thy} show that ignoring intra-month timing of trades makes little difference in their overall return calculations, but we see no reason not to calculate daily returns first and then aggregate to the monthly level, % ignore fine-grained temporal information, particularly given the short time-frame in which information arbitrage would likely take place and the low computational cost of calculating daily returns. \begin{comment} As a check of the integrity of our approach, we produced our own value-weighted portfolio weight matrix $\mathbf{w}^m$ (using daily closing prices and shares outstanding from CRSP), % to construct a market capitalization-based weight matrix $\mathbf{w}^m$), computed the portfolio return series $\mathbf{R^m}$ from $\mathbf{w}^m$ and $\mathbf{r}$, aggregated daily returns to a monthly frequency, and confirmed that the abnormal return $\alpha$ on this portfolio was essentially zero in both CAPM and Carhart Four-Factor regressions. % (It was not precisely zero because our dataset includes only those firms that members reported holding, whereas the market index used in the CAPM and Carhart regressions includes all stocks on the NYSE, NASDAQ, and AMEX exchanges; the difference is negligible because the stocks held at some point by members of Congress in our period collectively make up a large proportion of the overall market.) % The stocks on our dataset make up approximately XXX percent of the total market value of these exchanges, which helps explain the similarity of the standard market index and the one produced we produced using market capitalizations of firms in our sample.) This test confirms the soundness of the machinery by which we produce daily portfolio returns from portfolio weights, aggregate them to the monthly level, and compare them to the market index. It also indicates that differences in performance we observe between Congressional portfolios and the market index are not due to differences between the universe of companies in the market index and the subset that were held at some point by members of Congress and thus appear in our data; members could have matched the standard market index almost exactly with a value-weighted portfolio of the 2,581 companies in our dataset. % \footnote{This is a reflection of the fact that the companies that appear in the CRSP index but not in our dataset tended to be small and thus had very little weight in the value-weighted index.} \end{comment} \subsection{Results: Overall Performance} Before looking at abnormal returns estimated by market models, we display in Figure \ref{fig:cumret} the cumulative raw returns for the Congressional portfolio over our period of study. The figure depicts the value over time of \$100 invested in each of three portfolios in January of 2004: the market index (blue line), the aggregate (i.e. value-weighted) Congressional portfolio (dotted green line), and the average (i.e. equal-weighted) Congressional portfolio (dotted orange line).\footnote{For each month, we compute each member's monthly portfolio return and aggregate/average across members; the figure depicts the compound return on this series of monthly returns.} The two Congressional portfolios clearly do considerably worse than the market index: %The investments of Members of Congress underperformed the market throughout the period of our study. % , and particularly in the crash of 2008. %The market index (blue line) rose steadily from the beginning of 2004 until mid-2007 before dropping \--- gradually at first and then, starting in September of 2008, precipitously \--- such that \$100 invested in a market index in January of 2004 would be worth about \$80 by the end of 2008, whereas invested in the %The same \$100 invested in the Congressional portfolio would be worth only about \$65, whether invested in the aggregate Congressional portfolio (dotted green line) or the average Congressional portfolio (dotted orange line) it would be worth only around \$70. The underperformance is clearly not limited to the stock market crash of late 2008; at the market peak in 2007 the Congressional portfolio was already about 10\% below the market on a cumulative basis since the start of 2004. Table \ref{tab:overallalpha} provides the results of our estimates of the abnormal return on the Congressional portfolio. Panel A shows that the average monthly excess return %our estimate of $\alpha$ (monthly percentage ``Excess Return'') for the aggregate Congressional portfolio is negative and significant at conventional levels in both the CAPM and Carhart 4-Factor specifications.\footnote{Notice that unless noted otherwise, we use White robust standard errors throughout. We have re-estimated all models with Newey-West standard errors (with 3 and 6 lags) and the results are similar, except that standard errors tend to be slightly lower as may be expected given their downward bias in finite samples (see \cite{petersen2009estimating}).} Panel B shows the same basic results for the average Congressional portfolio; the estimates are slightly less precise with p-values of .11 and .09 for the CAPM and 4-factor models respectively. The excess return estimates all fall between. % what is this now? For the CAPM, the magnitudes suggest that the aggregate Congressional portfolio underperforms the market by an average of about .27 percentage points per month, which annualizes to a yearly excess return of about -3.2\% with a .90 confidence interval of $[-4.8;-1.6]$ ; the average Congressional portfolio underperforms the market by an average of about .24 percentage points, which annualizes to a yearly excess return of about -2.9\% $[-5.7;.1]$. The corresponding annualized figures for the 4-Factor model are -3.1\% $[-4.6;-1.4]$ and -1.9\% $[-3.8;-.1]$. \subsection{Performance in Subgroups} Table \ref{tab:alphaforsubgroups} reports estimated excess returns for relevant subsets of Congress and indicates that the overall underperformance is very consistent across subgroups. House members seem to do slightly better than Senators according to the aggregate portfolio, but the opposite is the case for the average member portfolio. % the opposite is the case. Members on power committees in the House or Senate\footnote{We define ``power committees" in the House as Rules, Appropriations, Ways and Means, and Commerce; in the Senate they are Appropriations, Finance, and Commerce.} do slightly better than other members, but the differences are small. The estimated excess returns are also very similar for the 2004-2006 period, when the market was rising, and the 2007-2008 period, when the market fell and the government began to intervene more heavily in the economy. %indicating that performance suggesting that the increased political intervention in the economy in the wake of the crisis did not result in much different overall returns. Republicans do slightly better than Democrats although again the differences are not quite significant at conventional levels. Members with lower seniority seem to do slightly better than those with medium or high seniority. There are no consistent differences across the group of members when we stratify by net worth (high, medium, low) or portfolio size (high, medium, low). %Republicans seem to have performed slightly better than Democrats, members with less seniority (i.e. fewer years since they were first elected to Congress) seem to have done slightly better than more senior members, and performance seems to have been better in the first three years than in the volatile last two years of our period. %Our analysis of the average portfolio also indicates that Republicans performed somewhat better and that the first three years were considerably better than the last two, but it also suggests that the Senate performed better than the House and that seniority and excess return were not particularly related. Above all, the consistently negative results across subgroups displayed in Table \ref{tab:alphaforsubgroups} suggest that our overall findings are not the artifact of a few exceptionally poor investors in Congress but rather indicate a broader underperformance across members. Notably, only 1 out of the 72 point estimates for the excess return (18 subgroups, each estimated four ways) is positive, and that point estimate is effectively zero. % The only positive alpha return for the aggregate portfolio of low seniority members in the 4-factor model is effectively zero at 0.004 and not robust in the CAPM or the average member portfolio for this group. %the point %estimate on the excess return is not positive for a single one of the nine subgroups we consider, calculated four different ways. \subsection{Member-Level Performance} In Figure \ref{fig:member_returns} we display estimated excess returns for each member in our dataset: estimates of alpha from a separate Carhart four-factor regression for each member. (Names are plotted only for members with relatively high or low returns or portfolio values.) A box and whiskers plot on each axis depicts the marginal distributions (the line indicates the median, the edges of the box denote the interquartile range, and the whiskers indicate the 5th and 95th percentiles). Not surprisingly, the mean monthly excess return across members at -.15 is very close to the mean monthly return of the average portfolio (-.16), and it is similar to the mean monthly return of the aggregate portfolio (-.26). The marginal distribution of returns is fairly symmetric and clearly centered below zero (the median is at -.17). % again indicating that the underperformance is very consistent across members. \subsection{Performance in Context} %\subsection{``Beating the market" in context} % put it in context from the lit. don't be surprised. % Our analysis has compared the performance of Congressional stock portfolios to a market index, the standard benchmark in empirical finance. While our finding that Congressional stock portfolios underperformed the market may be somewhat surprising based on the popular perception of politicians as savvy, well-connected, and possibly corrupt, it is consistent with a long line of empirical work documenting that even supposed investment experts do not reliably outperform market indices. An early example is \citet{cowles1933can}, who found that stock market forecasts and recommendations made by financial service firms, fire insurance companies, and the editor of the \emph{Wall Street Journal} %in the late 1920s and early 1930s tended to perform no better than what would result from random chance. In fact, every set of recommendations he examined on average did slightly worse than the market. Much subsequent research in empirical finance has examined the performance of professional fund managers, with debate focusing on whether there is evidence of any mutual fund manager consistently beating the market. %after costs have been subtracted. Some papers fail to find any evidence of stock-picking ability among managers of active mutual funds \citep{gruber1996another}; other papers find evidence of individual ability among certain mutual fund managers \citep{carhart1997pmf} or even the average mutual fund manager % beat the market on average but that expenses eliminate any gains for investors \citep{grinblatt1989mutual} but note that high performance is fully offset by management expenses. Several papers in recent years have documented that the portfolios of individual investors generally perform poorly (see, for example, \citet{odean1999investors,barber2000thy,barber2007all,barber2008just,goetzmann2008equity}.) A particularly interesting example is provided by \citet{barber2008just}, who analyze all trades in Taiwan over the 1995-1999 period and document a large systematic transfer of wealth from generally inept individual investors to savvier institutional investors. Stocks sold by individuals in this sample subsequently perform better than the stocks they purchase, while the opposite is true for stocks traded by institutional investors. The results suggest that in general the stock market is a place where informed institutions take advantage of uninformed and overconfident individuals who would be better off by relying on simple indexing. It appears based on our findings that, despite the advantages of their professional situation and large network of connections, members of Congress fare no better on average than the average member of the latter category. To put our findings in perspective, we provide in Figure \ref{fig:benchmarks} a comparison of the excess return we find for members of Congress with similar findings for other subgroups of investors. Our finding suggests that members of Congress perform on par with individual investors and mutual fund managers, as measured in \citet{barber2000thy} and \citet{carhart1997pmf}, and below that of corporate insiders and hedge fund managers as found in \citet{jeng2003eri} and \citet{fung2008hedge}. %One other possibility that we consider below is that Congressional members perform poorly because they engage in political tradeoffs. For example, members may be more likely to invest in companies that contribute campaign contributions to advance some political gains at the expense of some financial loss. While at he face of it, the robustness of the underperformance across subgroups speaks against this idea, we will explicitly consider political connections between members and companies in the section further below. \subsection{Comparison to \citet{ziobrowski2004arc}} As is clear in Figure \ref{fig:benchmarks} and noted above, our finding of weak overall performance contrasts strongly with a previous widely-discussed study by \citet{ziobrowski2004arc}, who find abnormal returns among traders in the Senate in the 1990s that to our knowledge exceed those of any documented investor group. One possible explanation for this discrepancy is the difference in the type of data and methods of analysis employed: our analysis to this point has focused on the portfolio positions of members of the House and Senate, while \citet{ziobrowski2004arc}'s finding is based on an analysis of a hedged portfolio constructed solely from trades made by members of the Senate. To make the most direct possible comparison, we now apply the method described in \citet{ziobrowski2004arc} to our data, such that any remaining differences should be due to changes in circumstances between the period in which the Ziobrowski study was carried out and our own period of 2004-2008. %Not only do we have a broader set of Congressional investors and data from %a more recent period, we also focus on an analysis of actual (reconstructed) portfolios held by members rather than a portfolio constructed solely from %transactions. In order to assess why we find poor investment returns in Congress while \citet{ziobrowski2004arc} had found stellar performance, %we reproduced as closely as possible \citet{ziobrowski2004arc}'s analysis using our dataset. In particular, we ignore reported end-of-year holdings and construct three portfolios based on transactions only: a buy portfolio, which holds all stocks purchased by members of Congress for 255 days following the purchase date, a sell portfolio, which holds all stocks sold by members of Congress for 255 days following the sell date, and a hedged portfolio that holds %long the purchased stocks and sells short the sold stocks (buy less sell portfolio). Like \citet{ziobrowski2004arc}, we assign precise dollar values to trades using the midpoint of the value band specified on the FDR, with a top-code at \$250,000. After constructing the portfolio and calculating daily returns, we again aggregate up to the monthly level and estimate excess returns with the CAPM and Fama-French 3-Factor models.\footnote{The Fama-French model is the Carhart 4-Factor model without the momentum term.} Table \ref{tab:ziobrep} reports the results: estimated excess returns on the buy sample, the sell sample, and the hedged (long/short) portfolio under the CAPM and Fama-French model for all members, Senate, and House. The overall result provides no evidence of informed trading; none of the coefficients are statistically significant. In separate analysis, we construct the transaction-based portfolio using shorter holding periods; with some combinations of holding period, model, and subgroup we find evidence of good or bad trading acumen, but the overall results are consistent with the null of 0 excess returns. (The last line of Figure \ref{fig:benchmarks} graphically depicts our alpha estimate for aggregate hedged portfolio for the Senate, which can be compared with the \citet{ziobrowski2004arc} finding that appears on the top line.) Why do our results differ from those of \citet{ziobrowski2004arc}? We see four main possibilities. First, the investing acumen of the average member of Congress (including generic stock-picking ability and perhaps the ability to link political events to movements in securities prices) may have declined since the 1990s. Second, the readiness of members of Congress to openly take advantage of their specialized knowledge may have lessened, perhaps as a result of heightened scrutiny in part due to \citet{ziobrowski2004arc}; savvy members of Congress may have, for example, moved their assets to a qualified blind trust managed by a ``political intelligence" hedge fund. Third, market conditions may have changed in a way that disadvantaged congressional investors. It could be, for example, that the bull market of the 1990s provided more opportunities for members of Congress to benefit from stock tips (on IPOs, for instance) than did the relatively moribund and finally panic-stricken market of the period we examine (although the intensified involvement of the government in the financial sector in 2007 and 2008 would seem to provide opportunities for arbitrage). Finally, it may simply be a matter of chance. \citet{ziobrowski2004arc}'s analysis looked at an average of about thirty investors a year; small samples occasionally produce extraordinary results. %% AE: what is it exactly? XXX \section{Is the Congressional Portfolio Political?} %% need to rewrite this. Our evidence to this point %on the performance of the overall Congressional portfolio has suggested that members of Congress perform no better than the average individual investor. We now turn to a more disaggregated look at Congressional investments to assess the extent to which portfolio choices and performance measures reflect political factors linking members and companies. % Our approach is to look at %In order to assess the mix of informational advantages and political considerations that drive members' investing behavior (and particularly their portfolio bias toward local companies and contributors), we now turn to an examination of the returns of members' portfolios, paying particular attention to differences in the performance of investments made in connected and unconnected portfolios. \subsection{Connection Measures} We define three types of connections between politicians and companies in our dataset that reflect an attempt to capture important channels by which members an firms interact: %In order to assess to what extent members' portfolio choices and investment performance reflect political factors, we also collected data on connections between politicians and the companies that appear in our dataset. \begin{itemize} \item \textbf{Constituency}: We obtained the location of each company's headquarters from Compustat % Google Finance and assigned this address to a Congressional District using an API provided by \texttt{GovTrack.us}; this allows us to label whether each stock holding involved %was for a company in the owner's constituency.\footnote{For Senators, an investment is considered in-district if the company is headquartered in the Senator's state.} \item \textbf{Contributions}: We collected PAC contribution data from the FEC\footnote{Via \texttt{watchdog.net}.} and linked PACs to companies and their contributions to members (289,694 reports totalling \$466.5 million). % in total). %and identified PACs linked to the companies in our dataset by checking the name of the PAC; %%% XXX was there more to this? This allows us to record, for each stock holding, how much the company contributed to the owner's election campaigns between 2003 and 2008. \item \textbf{Committee Lobbying}: We collected data on lobbying from the Center for Responsive Politics (CRP) and linked companies to members according to the extent to which each company lobbied on legislation appearing before committees on which each member sits. In particular, for each lobbying disclosure form filed between 2003 and 2008 on behalf of a company in our dataset (238,040 reports totalling \$18.2 billion), we assessed whether any bills were mentioned under ``Specific Lobbying Issues" (as processed by CRP) and then distributed the value of the lobbying reported in that disclosure form among committees to which named bills were referred;\footnote{For example, if a report disclosing \$50,000 of lobbying expenditure by Halliburton mentioned one bill that was referred to the Agriculture Committee %and another that was referred to the Energy Committee, \$50,000 would be added to the total lobbying connection between Halliburton and every member who sits on the Agriculture Committee; if the same report mentioned two bills, one of which was referred to Agriculture and another of which was referred to Energy, then \$25,000 would be added to the total lobbying connection between Halliburton and every member who sits on the Agriculture Committee, and another \$25,000 would be added to the total lobbying connection between Halliburton and every member who sits on the Energy Committee.} this gives us an indication, for each stock holding, of how closely linked the company's lobbying priorities are to the owner's committee responsibilities. \end{itemize} \subsection{Portfolio Choice and Political Connections} To assess members' portfolio choices, we examine the weight that a member puts on a company in his portfolio as a function of the connections he has with the company. (See \citet{cohen2008swi} for another example of this kind of analysis.) In particular, we estimate a regression of the form \[ w_{ij}=\beta_0 + \beta_1 District_{ij} + \beta_2 Contrib_{ij} + \beta_3 Lobbying_{ij} + \alpha_i + \alpha_j + \varepsilon \] where $w_{ij}$ is the weight in basis points of company $j$ in member $i$'s portfolio (averaged across years for which we have the member's portfolio), $District_{ij}$ is an indicator variable that takes the value 1 if the company is in the member's district and 0 otherwise, $Contrib_{ij}$ is an indicator variable that takes the value 1 if the company's PAC contributed to the member in the period 2003-2008 and 0 otherwise, $Lobbying_{ij}$ is an indicator that takes the value 1 if the company lobbied legislation before the member's committee and 0 otherwise, and $\alpha_i$ and $\alpha_j$ are member and company fixed effects.\footnote{Overall about 3\% of all member-company holdings in our dataset are connected by district, 18\% are connected trough any contributions, and 55\% are connected through any lobbying.} Regression (1) from Table \ref{tab:uncondportfolio} reports the coefficients from this regression; the other regressions include interactions and assess other definitions of connectedness. %A key advantage of this approach is that it helps somewhat to disentangle the relationships among these different connections, as e.g. many contributors are likely also local companies and/or lobby on bills referred to a member's committees. Our finding that the average portfolio weight is higher for local companies and companies that contribute to the member indicates that members are in fact taking advantage of their informational advantage in investing in these companies %Another possibility to evaluate is that ethical considerations prevented members of Congress %from taking full advantage of their stock-picking ability and investment knowledge. %Considering that investing in a campaign contributor or local company might invite criticism on conflict of interest grounds, and considering that those investments generally performed better than the average Congressional investment, it seems possible that members shied away from these investments in order to avoid the appearance of profiting from their public position. In fact, we find the opposite, as documented in Table \ref{tab:uncondportfolio} and Figure \ref{fig:8exp}. %We carried out a regression that assesses how the average weight of a company in members' portfolio depends on the degree of political connections between the member and the company. . As Table \ref{tab:uncondportfolio} indicates, the average portfolio weight in the data is 3.83 basis points, meaning .0383 percent of the total portfolio. Regression (1) in Table \ref{tab:uncondportfolio} indicates that the portfolio weight is quite substantially higher when the company is in the member's district or if the company has contributed to the member's election campaigns. Regression (2) includes all of the interactions from Regression (1); both of the significant main effects from Regression (1) remain significant and all of the interactions are significant as well, indicating that multiple connections between a company and member greatly increases the predicted weight of that company's stock in the member's portfolio. (Figure \ref{fig:8exp} plots the estimated portfolio weights implied by Regression 2 for an average member and company according to whether the company is connected to the member by district, lobbying, or contributions.) Regressions 3-5 extend this analysis by using different measures of connection, based on a binary indicator for being above the median among a member's connected companies (3) or based on a measure using the company's share of all contributions or lobbying expenditures directed to the member or his committees (4 and 5). Because all of these regressions include member and firm fixed effects, we are confident that these findings reflect the effect of member-firm connections on portfolio weights rather than simply a correlation between member or firm characteristics and our measures of member-firm connections. Taken together these results suggest that there is a large political bias in members' portfolio choices: members place considerably larger bets in companies to which they are politically connected. The result is robust to using several additional definitions of connectedness, including different percentile- and rank-based cutoffs. We have also replicated the analysis conditioning only on stocks that members actively choose to hold and obtain very similar findings.\footnote{The results, which are displayed in Table B.1 in appendix B are very similar to the to the unconditional overweighting. That is, even comparing only among the stocks that members choose to actively hold, they place much larger bets on politically connected companies. For example, compared to an average weight of 298 basis points they place an additional 329 basis points on home district firms and an additional 50 basis points on firms that provide campaign contributions. The overweighting is similarly increasing in the strength and combinations of the connections.} %Taken together these findings suggest that there is a very strong political skew in the members' portfolio weights towards contributor and local companies. In the appendix we have replicated this same finding looking only at companies that are actively held by members. The results are presented in Table B.1 are substantively identical to the ones reported here. %The biases toward constituent and contributor companies that we observe are consistent with the idea that members invest based on information about firms, as well as the signaling and bonding stories: members probably know more about local and contributor companies, and they may be particularly interested in signaling and/or aligning incentives with them. One can imagine three possible explanations for the propensity of members to invest disproportionately in local and contributor companies. First, members may invest in these companies simply because they know them. This appears to be the case for average individual investors, who invest disproportionately in local stocks but do not seem to have any particular information advantage in choosing among them. %As an indication of the local bias on individual investors, The typical U.S. household has about 30\% of its portfolio invested in stocks headquartered within a 250 mile radius of the family home, while on average only 12\% of all firms (the market) are headquartered within the same radius (see Ivkovic and Weisbenner 2005; or Seasholes and Zhu 2009 for a recent review). But according to the most comprehensive study of local investing patterns (Seasholes and Zhu 2009), individual investors' local holdings do not seem to exhibit superior returns, suggesting that individuals choose these companies simply because of familiarity. % rather than because they know them, not because they possess market-relevant information about them. A second explanation is that members hold connected stocks for political reasons. Members may invest in companies headquartered in their districts, or companies from which they hope to receive campaign contributions, in order to make policy promises more credible: voters may be more likely to vote for a candidate, and corporate PACs may be more likely to contribute to a candidate, when the candidate has aligned his financial incentives with their own by buying stock and thus made it more likely that he will support legislation favorable to their interests.\footnote{This reasoning requires that it is somehow difficult for members to liquidate their stock holdings in connected companies, and that members do not face too much political risk from legislating in the interests of companies in which they are invested.} If connected investments are made for political rather than financial reasons, we would not expect to find them to perform well. % we may expect these investment to kind of political reason, we again have no reason to think that these investments would perform well. % financial interests % in %order to align their financial interests with those of their constituents; %voters may be more inclined to re-elect a member who they %similarly, %believe preferences. Similarly, %potential contributors may be more likely to support a politician if they believe that the member's %investment in their stock makes it more likely that she will legislate with their interests in mind. In short, members may invest in %constituent or contributor companies in order to make policy promises credible. A third explanation is that members hold connected stocks because they have valuable information about those companies based on interacting with them as constituents and donors. Many members of Congress entered politics from business or local office, and arrive in Washington with extensive personal and business connections to companies headquartered in their districts. Once in office, they interact with local businesses as important constituents and possible sources of campaign funding. % business careers or were otherwise actively involved %in the business community of their district, providing % Listed companies that are headquartered in a member's district %are often major employers who have extensive dealings with their Congressional representatives. Not only do these companies %often seek help from their representatives in pursuing legislative and regulatory matters, but members often seek the electoral %support of business leaders in their districts based on their local influence. Companies from which members seek financial support similarly are often closely connected to the member. These connections often involve regular interactions between corporate executives and members of Congress at social and fundraising events, as well as frequent meetings between company lobbyists and Congressional staff, all of which may provide opportunities for the member to collect market-relevant information about these connected companies. %In the case of local companies, it is likely that connections between some companies and members go back well before the member takes office, considering that many members arrive in Washington having been a prominent figure in local business and political circles. The idea that such interpersonal connections may bring market advantages has been reinforced by \citet{cohen2008swi}, who find that mutual fund managers make larger bets on companies to which they are connected through educational ties and are also more successful in these connected investments. %\citet{coval1999hbh, coval2001gii} similarly find that mutual fund managers on average exhibit a modest bias toward the stocks of local companies (with the average fund manager investing a little under 7\% of her assets in companies headquartered within 100 kilometers, even though only 6.16\% of the market is located within that area); for the period before 1985 they find that these local investments perform better than non-local investments, suggesting that mutual fund managers in that period enjoyed informational benefits from proximity. In order to distinguish among possible reasons for members' preference for the stocks of local companies and companies that contributed to their election campaigns, we now turn to evaluating the performance of members' connected investments. % % CEOs and members of Congress from the same region also likely travel in the same social and business circles. If this is the case we would expect members to do better when they invest in companies to which they are politically connected, particularly through either campaign contributions or through geographical proximity. %\citet{coval1999hbh, coval2001gii} find that mutual fund managers on average exhibit a modest bias toward the stocks of local (defined as headquartered within 100 kilometers) companies, with the average fund manager investing a little under 7\% of her assets locally, even though only 6.16\% of the market is located within her local area. % %This explanation may be particularly relevant for members of congress which tend to possess private knowledge about these companies' prospects based on interacting with them as constituents and donors. Listed companies that are headquartered in a member's district are often major employers who have extensive dealings with their representatives in Washington, DC. Not only do these companies often seek help from their representatives in pursuing legislative and regulatory matters, but members often seek the electoral support of business leaders in their districts based on their local influence. CEOs and members of Congress from the same region also likely travel in the same social and business circles. If this is the case we would expect members to do better when they invest in companies to which they are politically connected, particularly through either campaign contributions or through geographical proximity. %Research on mutual fund performance \citep{cohen2008swi} indicates that mutual fund managers are more successful when investing in companies with which they have more educational ties through senior management. find a similar benefit from investing in local firms for mutual fund managers. Consistent with \cite{cohen2008swi}, \citet{coval2001gii} find that mutual fund managers enjoy a modest information advantage with respect to local companies: between 1975 and 1984, the local component of their portfolios outperformed the nonlocal component by about 2\% per year. They find no evidence, however, that fund managers' local portfolios outperformed the market after 1985. %Examining brokerage accounts in the 1991-1996 period, \citet{zhu:lbi} also finds no evidence that investors with a stronger propensity to invest locally enjoy higher returns, and shows that investment behavior seems more driven by familiarity (either through proximity or advertising) than by responses to fundamental information. %In Finland, the median non-Helsinki headquartered % rm has 12% higher weight among investors in its municipality than it does among all %Finnish investors. Finally, individuals in mainland China invest 8% more in rms from %their province-of-residence than a market-capitalization portfolio would predict. \subsection{Portfolio Performance and Political Connections} For each type of connection, we divide the Congressional portfolio into two subportfolios, one in which the stocks are connected (e.g., where the company issuing the stock is headquartered in the member's constituency) and one where the stocks are not connected. In looking at stocks connected by lobbying and contributions, we consider two types of connected portfolios, one that includes investments where the company was connected to any degree to the member (i.e. whether the company provided any contributions or lobbying to that member), and another that includes investments where the company's connected lobbying expenditure (or contributions, depending on the variable) are above the median for companies connected to that member. We also consider proftolios of companies that are double-connected in the sense that they are linked to a member through two connections simultaneously (e.g. lobbying and contribution, in district and contributions, etc.). Figure \ref{fig:cumrets} shows the cumulative raw returns for a \$100 dollar position in each average member mimicking portfolio beginning in January 2004. The connected (unconnected) average member portfolios mimics the investments of all members of Congress in connected (unconnected) stocks. We find that for the lobbying connection there is no notable difference between the returns of the connected portfolio. For the contribution connection, however, we find that the connected portfolio does about as well as the market while the unconnected portfolio, especially towards the end of the period, exhibits the familiar underperformance. Most remarkable, the portfolio consisting of members' investments in local companies robustly beats the market across the entire period; the end of period value is \$98, compared to \$80 for a market index fund. Non-local stocks in contrast exhibit the familiar underperformance. When we consider two-way connections in which local companies are also contributors or lobby the member's committee, the premium appears to grow slightly. %Our measures of connectedness, explained in detail above, are ``Lobbying" (the extent to which a company lobbies legislation before a member's committees), ``Contributions" (the extent to which a company's PAC contributes to the member's election campaigns), and ``In District" (whether or not the company is headquartered in a Senate's state or Representative's district); As a more formal test we estimate excess returns for each of these portfolios separately, and estimate the alpha on a hedged portfolio that is long on connected stocks and short on unconnected stocks. (See \citet{cohen2008swi} for another paper in empirical finance taking the same approach to assessing the role of company-investor connections in portfolio performance.) % companies that were connected to any degree to the member did any lobbying of legislation before the member's committee (made any PAC contributions) and one which includes companies that were among the top half of companies that lobbied legislation before the member's committee (companies that contributed to the member's campaigns). Table \ref{tab:connectedunconnectedalpha} reports the alpha estimates from CAPM (Panel A) and the Carhart 4-Factor model (Panel B) for the aggregate (left side) and average (right side) Congressional portfolio. The remarkable finding is that for almost every definition of connections the connected portfolio outperforms the unconnected portfolio, such that the hedged portfolio shows a positive excess return.\footnote{The only two exceptions (of 32 comparisons) are ``Lobbying (Any)"/CAPM/aggregate and average.} The alpha returns are also statistically significantly different from zero in 13 cases. For the average member portfolios in the 4-factor model all but one hedged estimate (the one for lobbying any) is significant at conventional levels. The largest estimates occur for the contributions and the in-district connection. One could match the market index by matching dollar-for-dollar all investments made by members of Congress in this period in companies from which they received campaign contributions. %gave campaign contributions to this member, one would easily match the market index. One could beat the market by an annualized 4.8\% by matching dollar-for-dollar all investments made by members Congress in this period in companies headquartered in their constituencies. %in which the member represents the constituency in which the company is headquartered, one would beat the market index by a highly significant .4 monthly excess return (with annualizes to 4.8\%). Consistent with Figure \ref{fig:cumrets}, the estimated excess returns for local investments are even higher when we consider companies that are both in-district and also gave contributions or lobbied a member's committees. This suggests that members possess substantial information advantage about local companies, arguably the strongest member-company connection we consider here given that members presumably interact with local companies on a regular basis. The relative performance of the connected portfolio seems weaker in the case of lobbying than the other two definitions of connectedness. How robust is this finding? For each of the local connections, Figure \ref{fig:localpremium} provides box plots of the distribution of alpha estimates that are computed on a member-by-member basis for the member's connected, unconnected, and hedged portfolios. Clearly, for both the CAPM and the 4-Factor models the average member specific return robustly beats the market on the connected portfolio, and this premium increases in the two-way connections (the median alpha on the connected portfolios in the 4-factor models are, for example, .48, .66, and .66 for the in-district, in-district and contributions, and in-district and lobbying connection respectively). This strongly suggests that the abnormal returns we find for local investments are not driven by a few unusual members. %essentially match the market index (as indicated by the alpha estimates for ``In District"/connected/aggregate, under either CAPM or the Carhart 4-Factor model). But by mimicking dollar-for-dollar all \emph{other} investments made by Congress in this period (i.e. those in which the member invested in a company headquartered in another district or state), one would underperform the market by more than 4\% a year (as indicated by the alpha estimates for ``In District"/unconnected/average, under either CAPM or the Carhart 4-Factor model). In the case of ``In District" connections, the excess return on the hedged portfolio is not statistically distinguishable from zero, but it is significant under several of the specification for both ``Contributions (Any)" and ``In District." %other definitions and specifications, including ``Contributions (Any)"/CAPM/Aggregate, ``Contributions (Any)"/Carhart/Aggregate, ``Contributions (Any)"/Carhart/Average, ``In District"/CAPM/Average, and ``In District"/Carhart/Average. %The hedged portfolio is statistically significant for lobbying only looking at companies connected by lobbying above the median and using the 4-Factor model, whereas 7/8 of the %hedged estimates for the contributions-based connection are at least borderline statistically significant and both of the estimates for in-district connections and the average portfolio are strongly statistically significant. In fact, the only non-hedged portfolio with a significantly positive excess return is the ``In District," average Congressional portfolio: if one were to create a portfolio that mimics the average investment in which the company is headquartered in the member's district, %and created a portfolio based on the average portfolio weights assigned to these investments across members, %one could have beaten the market by over 4.5\% % 4.76 exactly %per year in this period. %Figure \ref{fig:cumrets} provides another way of evaluating the relationship between member-firm connections and portfolio performance. In each panel the blue line indicates the performance of the market index, the solid orange line indicates the performance of the connected portfolio, the dotted orange line indicates the performance of the unconnected portfolio, and the dotted green line indicates the performance of the hedged (connected minus unconnected) portfolio. The top two panels indicate that the lobbying-connected portfolio (aggregate for all members) did not substantially outperform the unconnected portfolio, whether connected companies are defined as those that did \emph{any} lobbying of bills before a member's committees (left panel) or \emph{above the median amount} of lobbying on bills before a member's committees (right panel). Because the connected and unconnected portfolios performed similarly in this case, the hedged portfolio stays fairly steady throughout the period. For contribution-based connections, by contrast, connected companies do almost as well as the market and unconnected companies do substantially worse. Finally, the bottom panel shows that the connected in-district portfolio outperforms the market for most of the period we examine, while the unconnected portfolio underperforms the market. %%% AE: is there something wrong with the hedged stuff? XXX We suspect that the advantage members of Congress enjoy from investing in local companies is based on the market-relevant information they acquire about local companies in the course of their interactions with those firms. It is also possible that the local advantage emerges from members' knowledge of legislation that will affect local companies (e.g. earmarks), but without further evidence we are inclined to think that members of Congress resemble mutual fund managers as depicted in \citet{cohen2008swi}, in the sense that they benefit from personal connections to local companies and their senior managers. % contrib The apparent advantage of investing in one's contributors (which shows up both in the aggregate and the average portfolio analyses) may be a combination of familiarity and policy knowledge. Many contributors are local companies, and those that are not typically contribute because they seek access and influence with influential members. Either way, the contribution signals a relationship in which a member hears periodically about a firm's concerns and opportunities and also is likely to be involved with legislation affecting the firm's fortunes. Note that market-relevant information that members glean from local and contributing firms could be of two types. It could be public but not widely-known information, such as an important detail reported in the previous day's quarterly report; assuming less than perfect market efficiency, making investment decisions based on this information would seem to pose no legal or even ethical problems. This information could also be private, ``insider" information; while trading on this information could be lucrative for the member and sharing it might be rewarding to corporate executives hoping to attain favorable legislative or regulatory treatment, both the member and the executive could be implicated by existing insider trading laws. Unfortunately, we have no way of distinguishing between these channels; future work might look at whether members appear to make well-timed trades that anticipate earnings announcements and other corporate communications. \begin{comment} \subsection{Company Level Connections} A possible alternative explanation of our finding that the portfolio of contributions-connected companies outperforms other investments is that companies that generally gave more campaign contributions outperformed other companies in this period. In other words, the contributions-connected portfolio may have performed better not because of the specific relationships between the member and her contributor, but simply because companies that contribute generally did better than those that did not, and our member-firm connections merely pick up this overall pattern. To address this alternative explanation, we conduct the same analysis but define the connected portfolio as the set of all investments made by members in companies that gave contributions to any member during our time period. (Investments in a particular firm are thus all defined as connected or not connected, depending on the firm's PAC contribution total.) The results are displayed in Table \ref{tab:companyconnected}. We find no difference in the performance of the connected and unconnected portfolios defined in this way, suggesting that the portfolio of investments where the PAC contributed to the member outperforms the unconnected portfolio because of the specific relationship between the member and the firm rather than firm characteristics. \end{comment} \subsection{Implications} What have we learned about the relationship between members of Congress and firms to which they are politically connected? The strongest results relate to local companies. Our findings indicate that members invest disproportionately in local companies, and that they do this in part because they (unlike average individual investors) % or even mutual fund managers) have information about these companies that allow them to earn excess returns on these investments. While we cannot be sure of the specific nature of this knowledge (e.g. whether it is a diffuse knowledge of local companies gained through extensive and long-standing interaction with local individuals or specific knowledge of upcoming events in business or government), the fact that members' local investments perform so well (and that the local advantage is so widespread) %% need to document this I guess does suggest that relationships between members of Congress and local companies are substantial and common. The evidence on members' investments in their contributors is similar but less strong. Members disproportionately invest in these companies, and investments in contributors outperform the rest of their portfolios, but both the bias toward contributors and the apparent informational advantage from these investments is smaller than in the case of local companies. It is possible that the performance of these investments is less stellar because the desire to cement political exchanges plays more of a role in members' portfolio choices, but it is impossible to say with much confidence. % (although to a lesser extent than local companies) The fact that members do not appear to do particularly well when investing in companies that lobby legislation before their committees provides some indication that members' policy information is not particularly useful as the basis for an investment strategy. Our analysis of portfolio choices also fails to find much evidence that members disproportionately invest in lobbying-connected companies, suggesting that members also do not expect to be able to profit from their connection to these companies. These results suggest that the main concern in most public discussion of \citet{ziobrowski2004arc}'s finding, as well as in the STOCK Act \--- members' use of information about pending legislative activity to enrich themselves \--- was not a major factor in members' investment performance in the 2004-2008 period.\footnote{It is possible that knowledge of more targeted legislation, such as earmark activity, accounts for the local and contributor premium.} We are hesitant to make very much of this finding, however, because we view the lobbying measure as the noisiest of our connection measures: we are unsure whether we obtain a null result because members do not take advantage of (or do not possess) market-relevant information about companies over which they have legislative power or because our measure does not faithfully capture legislative relationships. %% was "lobbying" not legislative %\subsection{Why poor overall performance?} % %There are many possible ways in which we could account for the discrepancy between %Congressional investment performance and the performance of the market index. %We carry out one simple decomposition to assess whether members of Congress were %generally tilted toward the wrong sectors of the economy. In specific, we decompose the discrepancy between the %performance of the Congressional portfolio and that of the market into the component that is due to a) the %discrepancy between sector-level weights in the Congressional portfolio and sector-level weights in the index vs %b) the discrepancy between within-sector stock weights in the Congressional portfolio and within-sector stock weights in the %index. We find (full results available on request) that sector weights in the Congressional portfolio are %generally close to sector weights in the index, and that a hypothetical portfolio in which the aggregate Congressional %portfolio was reweighted to match the index's sector-level weights would have underperformed by roughly the same amount as the actual %Congressional portfolio. (A hypothetical portfolio in which sector weights were kept at their actual level in the aggregate %Congressional portfolio but within-sector stock weights were set at the market index level performed as well as the market.) This exercise rules out the %possibility that members of Congress underperformed the market because they held the wrong sector weights. % %Our finding that the Congressional stock portfolio underperformed relative to the market in the period 2004\---2008 differs from the findings of \citet{ziobrowski2004arc}, the sole published paper %analyzing Congressional investments, which asserted based on an analysis of FDR's filed by Senators between 1993 and 1998 that %a portfolio mimicking Senatorial stock transactions could beat the market by the staggering amount of almost one percentage point per month.\footnote{Professor Ziobrowski testified before the House Financial Services Committee regarding these findings on July 13, 2009. Testimony available at \url{http://www.house.gov/apps/list/hearing/financialsvcs_dem/ziobrowski_testimony.pdf}, accessed Sept. 1, 2010.} %One obvious difference between the papers is the time period; %it may be that the volatile market conditions, increased public scrutiny, and new cast of characters in our period explains the fact that we do not find members of Congress earning outsized returns. The more important difference, in our view, is that our data provides a much more complete picture of the actual investments of members of Congress. %\citet{ziobrowski2004arc} bases its analysis solely on the transactions of Senators; the portfolio they analyze is constructed by assuming that each purchase made by a Senator is held in the portfolio for one year (and each sale is shorted for one year). The fact that this portfolio beats the market according to their analysis suggests that Senators' trades were well\-timed and is a useful finding. Building a synthetic portfolio based on transactions may not give a good indication of the performance of Congress's investments overall, however, especially given that turnover in our dataset is less than 2\% per year. %In Appendix XXXX, we perform \citet{ziobrowski2004arc}'s exact transaction-based analysis on our data. We find some evidence of opportune timing in transactions, but the scale is nowhere near that reported by \citet{ziobrowski2004arc}. We find no evidence of excess returns in a portfolio that mimics Congressional transactions in our period. %\subsection{Connections and Portfolio Weights} %%% this should be shorter XXX %Ethical questions aside, one piece of advice for members of Congress based on our findings would seem to be to invest more heavily in %companies to which they are connected: portfolios made up solely of stocks from companies in a member's district, companies making contributions to a member's election campaigns, and companies lobbying a member's committees perform better than portfolios made up of unconnected stocks, and in some cases better than the market. As it turns out, members of Congress appear to already %substantially favor these connected companies. %Table \ref{tab:uncondportfolio} and Figure \ref{fig:8exp} report the results of a %portfolio weight %regression in which each member's holdings of each company (averaged over the five years of our sample) is regressed on variables measuring the connections between that member and that company. (To control for firm- and member-level heterogeneity, we include fixed effects for both firms and members.) The results indicate that members of Congress put hundreds of times greater portfolio weight on connected companies than other companies, controlling for member and company characteristics. %As Table \ref{tab:uncondportfolio} indicates, the average portfolio weight in the data is 3.83 basis points, meaning .0383 percent. Regression (1) in Table \ref{tab:uncondportfolio} indicates that the portfolio weight is quite substantially higher when the company is in the member's district or if the company has contributed to the member's election campaigns. Regression (2) includes all of the interactions from Regression (1); both of the significant main effects from Regression (1) remain significant and all of the interactions are significant as well, indicating that multiple connections between a company and member increases the predicted weight of that company's stock in the member's portfolio. %Figure \ref{fig:8exp} plots the estimated portfolio weights implied by Regression (2) of Table \ref{tab:uncondportfolio} for an average member and company according to whether the company is connected to the member by district, lobbying, or contributions. One indication of the importance of member-firm connections in predicting portfolio weights is that, whereas a member's predicted portfolio weight for a company that is not connected to her in any way is around .04 percent, the predicted portfolio weight for a local company that contributes to her campaign and has lobbied legislation before her committees is around 1.5 percent \--- over 3,000 times as high. %Regression (3) of Table \ref{tab:uncondportfolio} indicates that the propensity of members to invest in companies that contribute to their campaigns %is increasing in the intensity of the contributions: the average portfolio weight for a company that provides any campaign contributions to a particular member is about .19\%, based on Regression (1), but %the average portfolio weight for a company that is among the top half of that member's contributors is is about .28\%, based on Regression (3). (The point estimate for companies that %lobby a member's committees particularly heavily is also higher, but neither estimate is significantly different from zero and they are not significantly different from each other.) %Regressions (4) and (5) of Table \ref{tab:uncondportfolio} reinforce these findings by looking at lobbying and contributions between firms and members as continuous rather than binary variables. Regression (4) indicates that the average member invests more heavily in firms that provide a greater share of her campaign contributions; Regression (5) suggests that this predilection is stronger when the firm is headquartered in the member's constituency. %\subsection{Commissions and Bid-Ask Spreads} %The foregoing analysis focuses only on gross returns and does not even consider commissions and bid-ask spreads, which are commonly estimated to amount to as much as 3\% of the value of a round-trip investment. \emph{Do BOTE on what this would amount to? Based on the Barber Odean stuff, and the transactions we have} %\emph{Or put in conclusion.} % \section{Conclusion} Our study of the investments of members of Congress has yielded two main findings that may appear somewhat at odds with one another. On one hand, our analysis indicates that members of Congress were mediocre investors during the 2004-2008 period that we examine, falling short of the market benchmark by 2-3\% per year. This finding contrasts with the sole previous study of Congressional investments, which found large excess returns in analysis of trades by Senators in the 1990s and has drawn considerable attention in the media and in Congress itself. On the other hand, we find that the politically-connected subset of members' portfolios outperformed the rest of their investments, and that members' investments in local companies handily outperformed the market. This finding is especially significant considering that there is no evidence of either individual investors or money managers outperforming the market in their local investments in recent decades, which suggests either that members of Congress have particularly strong local knowledge or that their valuable knowledge comes particularly from political interactions with constituents. % hmm a little unclear here We find the overall message to be consistent, however. Members of Congress are not investing geniuses. Most of what they know about political developments is probably quickly incorporated into asset prices, and many members likely recognize the possible political costs of trying to make money on whatever private political information they do possess. That their portfolios would perform about as well as the average individual investor is therefore not entirely surprising. The one area where member of Congress are on average perhaps the most unusual compared to ordinary investors is in their extensive connections to local business leaders, who seek out their assistance with legislation and whose assistance they seek out for reelection. Our findings suggest that it is on these local investments, rather than investments in companies affected by legislation for which they have responsibility, that members are able to excel. Unfortunately for them, these investments form a relatively small part of their overall portfolio and do not make up for the mediocre performance of their other investments. % more here? %\section{Discussion} %Our primary findings to this point have been a) that the common stock portfolios of members of Congress underperformed the market by a considerable degree between 2004 and 2008 (Table \ref{tab:overallalpha}), b) that this underperformance was fairly consistent across subgroups of Congress (Table \ref{tab:alphaforsubgroups}), and c) that members of Congress did better (though still no better than the market) when they invested in companies to which they were somehow connected \--- because the firm was headquartered in their district, lobbied legislation that appeared before the committees on which they sit, or provided money for their election campaigns (Table \ref{tab:connectedunconnectedalpha}). In this section we consider some implications and extensions of these findings. %\subsection{Investing in connected companies} %As described more fully above, Table \ref{tab:connectedunconnectedalpha} reports that members' connected portfolios appeared outperformed their unconnected portfolios %under three definitions of connections: companies that lobby the committees on which the member sits, companies whose PAC contribute to the %member's election campaigns, and companies headquartered in a member's constituency. %\section{Portfolio Choices} %\subsection{Measures of Connectedness} %We first examine portfolio choices of Members of Congress, focusing on the extent to which Members focus their investments on local firms, firms that provide campaign donations, and firms that are particularly affected by congressional legislation. In particular, we employ regression analysis to investigate how the weight that a Member puts on a company in his portfolio varies as a function of the connections he has with the firm. (See \citet{cohen2008swi} et al for another example of this kind of analysis.) %For each of the $453$ Members with stock holdings we construct their portfolio weights in basis points by computing the share of holdings of each firm relative to their total holdings averaged over the 2004-2008 period. We include all $2,617$ firms that are held by at least one Member in this period resulting in $1,185,501$ possible Member-firm holdings. Together these firms make up more than $94$ \% of the total market value in the entire universe of CRSP common stocks and thus provide an accurate coverage of the universe of firms among which Members are likely to chose their stocks allocations. %To compare allocations in stocks to which Members are connected politically, relative to stocks to which they are not connected we define three sets of measures of ``connected'' holdings. First, we classify stocks that are connected to Members through geographic proximity. It is well known in empirical finance that mutual fund managers and individual investors prefer to invest in local stocks since investors are more familiar with local firms (\cite{coval1999hbh,coval2001gii,zhu:lbi}). For Members we expect a similar local familiarity bias, but it may be even stronger since politicians have various additional dealings and frequent contact with local companies that ask their representative for policy favors. We define \emph{In State} as a binary indicator for stocks to which a Member is connected because the company has its headquarter in the Member's home state.\footnote{We extract the headquarter location for firms from Google Finance.} %Second, we classify stocks that are connected because the company PACs provided campaign donations to a Member.\footnote{PAC contributions data comes from the FEC via \texttt{watchdog.net}} We define the binary indicator \emph{Contributions} that codes a company and member as connected if, in the period 2004 to 2008, the company's PAC gave any contribution to the member. To capture the increasing degree of strength of this link we also code another binary indicator \emph{Contributions ($>$ p50)} that codes stocks of companies whose contributions exceeded the median amount of contributions given to each Member. Finally, \emph{Contributions Strength} measures the share of a company's contributions as a fraction of a Member's total contributions in basis points. %Third, we classify connections between firms and Members based on the Member's policy portfolio based on actual corporate lobbying. In specific, we tally up how much each company lobbied on bills referred to each committee, and define a company as connected to a member if the company lobbied a bill before one of the committees on which the member sat during 2004-2008.\footnote{In this approach we thus use bill referrals rather than statutory jurisdictions to define committee policy areas \citep{king1994ncc}, and we use bill lobbying rather than industrial classifications to determine which policy areas companies view as important to them. We considered an alternative coding based on a mapping between industries and committees based on the committees' stated jurisdictions, extending \citet{myers:jmp}'s mapping of House committees to two-digit SIC codes. (The approach of linking committees to industries through jurisdictions has previously been used by \citet{munger1989stt, endersby1992ila}). However, the industry classifications are far too coarse in some instances, making many companies appear connected to members when they are not, and in other cases clear connections are overlooked. For example, Northrop Grumman, a major defense contractor, falls under SIC code 38, ``Instruments and Related Products,'' along with photographic equipment companies like Kodak, Fuji, and Canon and a host of medical device companies. According to Myers' mapping, this industry comes under the jurisdiction of the armed services committee, but not the defense subcommittee of the appropriations committees. The problems with using statutory committee assignments were noted by \citet{king1994ncc}. In our view the lobbying/bill-referral approach gives the best representation of which members had a special role in shaping legislation that mattered to companies.} We again use three measures: \emph{Lobbying} is a binary indicator for companies that did any lobbying of this sort, \emph{Lobbying ($>$ p50)} codes companies whose lobbying exceeded the median amount of lobbying for each Member, and \emph{Lobbying Strength} measures the share of a company's lobbying as a fraction of a Member's total lobbying in basis points. %Overall, about $4$ \% of all stocks are coded as connected by the \emph{In State} metric, about $4$ \% of all stocks are coded as connected by the \emph{Contributions} metric, and about $18$ \% of all stocks are coded as connected by the \emph{Lobbying} metric. Apart from the connections outlined above, Members may choose stocks based on a number of other motivations (such as the general popularity of certain firms, the level of diversification, etc.). Since some of these factors may be correlated with our connection measures, we include a full set of Member and company fixed effects into the regression to difference out these two sources of unobserved heterogeneity. The model is therefore identified based on within-Member and within-company variation and we can rule out the possibility that the results are driven by unobserved factors that vary across Members and or firms.\footnote{Note that the use of both fixed effects extends the approach used by \citet{cohen2008swi} who include either firm or fund fixed effects but never both.} We cluster our standard errors by Members in order to account for the fact that a Member's investments may not be independent. %\subsection{Results} %Results from the portfolio choice analysis are presented in table \ref{tab:uncondportfolio}. We find a strong political skew in the Members' portfolio allocations. Looking at column 1, Members' invest an additional $15$ basis points if a companies is in their home state. Compared to the average weight of $3.8$ basis points this constitutes an increase of more than $400$ \%, a degree of overweighing that far exceeds previous estimates of local bias for other types of investors (NEED CITES HERE). As another benchmark, in their study of education connections \citet{cohen2008swi} find that fund managers place an additional 8 basis points on companies where a senior officer (CEO, CFO, or Chairman) attended the same school, with the same degree, and at a similar time as the fund manager.\footnote{Their table 2 model 8 with firm and quarter fixed effects (but not fund fixed effects).} %We find that Members' also heavily overweight companies that provide campaign contributions. Compared to non-contributors, they place an additional 14 basis points in companies with any PAC contributions (a $360$ \% increase over the mean weight). Moreover, as column 3 and 4 suggest this premium is increasing with the strength of the contribution connection. Members place an additional 24 basis points on companies that are among the top 50 percent of their contributors (column 3) and a 10 basis point increase in the relative share of contributions from a company results in a 0.5 basis point increase in the Member's portfolio weight (column 4). %Looking at the measures of lobbying connections, we find no bias towards firms that lobby committees that Members sat on. The points estimate for \emph{Lobbying} is almost zero and the same is true once we narrow in on companies whose lobbying exceeds the median amount lobbying for each Member (column 3) or use the relative measure of \emph{Lobbying Strength} (column 5). %Finally, in column 2 we consider whether the overweighing is increasing for companies that are connected through multiple connections by including one dummy variable for each possible combination of \emph{In State}, \emph{Contributions}, and \emph{Lobbying}. The estimates of the conditional average weights for each of the combinations are also displayed in Figure \ref{fig:8exp}. The overweighing is clearly increasing for multiple connections. Compared to unconnected companies, Member's place an additional 10 basis points in companies that are solely connected through state, and an additional 11 basis points in companies that are solely connected only through contributions. Companies that are both contributors and in the home state receive a striking 76 additional basis points on average. This premium increases further to 96 basis points if, in addition, the company is also connected through committee lobbying. This reinforcing effect of additional connections is confirmed column 5 where we interact our continuous measures of contributions and lobbying strength with the home state indicator. In both models the interaction terms enter positive and significant indicating that the portfolio weights are increasing in contributions and lobbying to a much stronger extent if the companies are located in a Members' home state. %\subsection{Portfolio weight regressions} %As an alternative way of assessing members' portfolio choices, A key advantage of this approach is that it helps somewhat to disentangle the relationships among these different connections, as e.g. many contributors are likely also local companies and/or lobby on bills referred to a member's committees. Note however that the question addressed is somewhat different from the one posed above: here we examine the weight a member puts on a company, conditional on holding it, as a function of the connections between the company and the member. (Above, the comparison was between the total portfolio weight given to a set of connected companies by a member and the weight given to that set by all members.) %Table \ref{tab:reghomebias} provides the regression results. The regression includes dummies for state connection, contribution connection, and committee lobbying connection (our preferred measure of policy oversight by the member of the firm\footnote{Although the approach of linking industries to committees based on committees' statutory jurisdictions has been used in many previous papers (see cites above), the industry classifications are far too coarse in some instances, making many companies appear connected to members when they are not, and in other cases clear connections are overlooked. For example, Northrop Grumman, a major defense contractor, falls under SIC code 38, ``Instruments and Related Products,'' along with photographic equipment companies like Kodak, Fuji, and Canon and a host of medical device companies. According to Myers' mapping, this industry comes under the jurisdiction of the armed services committee, but not the defense subcommittee of the appropriations committees. The problems with using statutory committee assignments were noted by \citet{king1994ncc}. In our view the lobbying/bill-referral approach gives the best representation of which members had a special role in shaping legislation that mattered to companies.} %) as well as fixed effects for member, company, and year.\footnote{Because of the fixed effects, we can rule out the possibility that our results are based on e.g. correlations between members' overall level of diversification and connectedness to companies, or correlations between companies' popularity among investors generally and their amount of lobbying or political contributions.} With the average portfolio weight in the sample being just over 5\%, the weight given to a connected company is a full 2 percentage points higher if the company is in-state, and about half a percentage point higher if the company is a contributor or lobbies bills before the member's committees. %It may seem inconsistent that committee lobbying produces a positive bias here (conditional on owning stock in the company) but no bias above (looking at the portfolio share of connected firms). The two findings are, however, consistent with a situation in which the choice not to hold some connected companies is balanced by the choice to take bigger stakes in the companies the members do hold. %\subsection{Home bias calculations} % %It is well known in empirical finance that investors tend to invest much more heavily in domestic stocks relative to foreign stocks than would be optimal from a diversification standpoint, a puzzle known as ``equity home bias" \citep{french1991ida}. %%Using data on equity holdings in 20 countries from a 1997 IMF survey, Li et al report that investors everywhere are biased strongly toward holding domestic equity. In the US, 85\% of equity held %% by domestic investors was domestic, at a time when (according to CAPM) it would have been optimal from a diversification standpoint to hold about 55\%. %An international survey of home bias in 2005 \citep{sercu:hbi} showed that 82.2\% of equity held in the US was domestic, even though the optimal proportion from a diversification standpoint would have been 40.5\%. (In most countries the bias tended to be much bigger, e.g. 69\% domestic equity holdings in Sweden compared to an optimal 1.3\%.) % update this based on the new paper %In our sample, the share of domestic stocks in the average member's stock portfolio is 90.5\%, suggesting that members of Congress are somewhat more heavily invested in US equities than the average investor.\footnote{The IMF data %I believe %uses aggregate equity stocks, and thus may not be a good comparison group for members of Congress.} The average home bias among members does not appear to differ by party or chamber. % %Following up on \citet{coval1999hbh,coval2001gii} and \citet{zhu:lbi}, who find that mutual fund managers and individual investors prefer to invest in local stocks, we ask whether members of Congress show a preference for stocks in their neighborhood. We proceed by comparing members' portfolios to each other: is the share of a member's portfolio invested in his state larger than the average share of assets invested in that state across all members? Put differently, consider matrix $\mathbf{S}$ with representative element $s_{i,j}$ indicating the share of member $i$'s portfolio invested in member $j$'s state. The diagonal elements where $i=j$ indicate the share of member $i$'s portfolio invested in companies from his own state (as does $s_{i,j}$ when members $i$ and $j$ share a state). We want to assess whether $s_{i,i}$ is large compared to the expected value of $s_{+, i}$, i.e. the share of the average portfolio invested in member $i$'s state. % %question here of whether it should be the value-weighted avg portfolio, or the whole %market in a sense. % %Following the literature on equity home bias, we calculate for each member $i$ the following measure of bias: %\begin{equation} %\tilde{b} = 1 - \frac{1 - s_{i,i}}{1 - s_{+,i}}, %\end{equation} %which is roughly equivalent to the difference between one's own share of in-state holdings and the average share of holdings in that state (as can be seen by rearranging the expression as $\frac{s_{ii}-s_{+,i}}{1-s_{+,i}}$).\footnote{The main difference between our measure of home bias and the standard approach is that we use the average portfolio among members rather than CAPM or market capitalization (i.e. all investors' assets) as the ``unbiased'' portfolio.} As shown on the first line of Table \ref{tab:homebias} (computed for members who report stock holdings), about 18\% of the average member's portfolio is made up of companies in her own state, while under 4\% of other members' portfolios are made up of companies in that state. The average home bias (calculated as above) is about .15. Based on a permutation test in which we randomly shuffled the rows of the $\mathbf{S}$ matrix and recomputed the average home bias each time, we can reject the null hypothesis of no connection between members' states and their portfolio choices with a p-value that is virtually zero. There is also a large bias when we narrow down to Congressional districts for House members: the average member invests about 5\% of his portfolio in companies headquartered in his district, while Congress as a whole invests only .2\% in the average district. While our measures are not exactly comparable, these figures suggest that members of Congress exhibit a ``local home bias" much larger than that reported for mutual fund managers by \citet{coval2001gii}. In that paper, 6.95\% of fund assets were invested within 100 km of the fund manager (an area over twice as large as the average Congressional district), compared to 6.16\% of the market being headquartered in that area (on average, across fund managers). % %We conducted the same test looking at two other types of connections between members and the companies in their portfolios. First we looked at PAC contributions and examined whether members tilted their portfolios toward companies whose PACs provided campaign donations.\footnote{PAC contributions data comes from the FEC via \texttt{watchdog.net}. We define a company and a member as connected if, in the period 2004 to 2007, the company's PAC gave any contribution to the member.} As shown in Table \ref{tab:homebias}, contributor companies account for 24\% of the average member's portfolio, while those same companies on average account for just 16.7\% of all members' portfolios. %% is devoted to companies that gave campaign contributions in this period, while 16.7\% of other members' portfolios on average are made up of those same companies. %The average ``home bias'' across members calculated this way is .09, and again our permutation tests suggests that we can soundly reject the null hypothesis of no connection between contributions and portfolio weights. The bias toward contributors becomes somewhat stronger (depending on how it is calculated) when we narrow down to look at the top 40 and top 20 contributors for each member. Remarkably, 10\% of the average portfolio is invested with the member's top 20 PAC contributors, while on average only 5\% of all assets are invested with that set of companies. % %Since most contributors are local, the contributor bias and district bias are likely somewhat confounded. As one approach to addressing this, we ignore in-state contributors and assess whether members invest more in their out-of-state contributors than does Congress as a whole. The seventh row of \ref{tab:homebias} shows that they do, at a small although still substantively and statistically significant level. (Below we employ a regression approach to partial out the different biases.) % % %Perhaps it should not be surprising that the jurisdiction-based mappings of industries to committees produces no evidence that members favor companies they regulate. %that we use is imperfect and may produce a weak or even nonexistent signal of committee connection: % %\begin{comment} %\subsection{Portfolio weight regressions} % %As an alternative way of assessing members' portfolio choices, we examine the weight that a member puts on a company in his portfolio as a function of the connections he has with the company. (See \citet{cohen2008swi} et al for another example of this kind of analysis.) A key advantage of this approach is that it helps somewhat to disentangle the relationships among these different connections, as e.g. many contributors are likely also local companies and/or lobby on bills referred to a member's committees. Note however that the question addressed is somewhat different from the one posed above: here we examine the weight a member puts on a company, conditional on holding it, as a function of the connections between the company and the member. (Above, the comparison was between the total portfolio weight given to a set of connected companies by a member and the weight given to that set by all members.) % %Table \ref{tab:reghomebias} provides the regression results. The regression includes dummies for state connection, contribution connection, and committee lobbying connection (our preferred measure of policy oversight by the member of the firm\footnote{Although the approach of linking industries to committees based on committees' statutory jurisdictions has been used in many previous papers (see cites above), the industry classifications are far too coarse in some instances, making many companies appear connected to members when they are not, and in other cases clear connections are overlooked. For example, Northrop Grumman, a major defense contractor, falls under SIC code 38, ``Instruments and Related Products,'' along with photographic equipment companies like Kodak, Fuji, and Canon and a host of medical device companies. According to Myers' mapping, this industry comes under the jurisdiction of the armed services committee, but not the defense subcommittee of the appropriations committees. The problems with using statutory committee assignments were noted by \citet{king1994ncc}. In our view the lobbying/bill-referral approach gives the best representation of which members had a special role in shaping legislation that mattered to companies.} %) as well as fixed effects for member, company, and year.\footnote{Because of the fixed effects, we can rule out the possibility that our results are based on e.g. correlations between members' overall level of diversification and connectedness to companies, or correlations between companies' popularity among investors generally and their amount of lobbying or political contributions.} With the average portfolio weight in the sample being just over 5\%, the weight given to a connected company is a full 2 percentage points higher if the company is in-state, and about half a percentage point higher if the company is a contributor or lobbies bills before the member's committees. % %It may seem inconsistent that committee lobbying produces a positive bias here (conditional on owning stock in the company) but no bias above (looking at the portfolio share of connected firms). The two findings are, however, consistent with a situation in which the choice not to hold some connected companies is balanced by the choice to take bigger stakes in the companies the members do hold. % %\subsection{Discussion} % %We view the finding that members of Congress invest disproportionately in companies to which they are politically connected as quite striking. The bias is strongest toward local companies; as noted above, mutual fund managers also appear to favor local companies but apparently not to the degree that members of Congress do. The portfolio weight assigned to contributor companies and committee-lobbying companies is smaller but suggests that members invest almost 10\% more in connected than in non-connected companies, conditional on state connection and member and firm fixed effects, which is quite a sizable difference.\footnote{For comparison, this %%magnitude of the pro-contributor and pro-committee-lobbying company %bias in portfolio weights is about the same as the bias toward connected companies found in \citet{cohen2008swi}, whose central finding is that mutual fund managers make bigger bets on connected companies. (Connected companies there are defined based on shared educational background between mutual fund managers and the firm's managers). Meanwhile, the unconditional home bias estimates in that paper actually show a bias \emph{away} from connected companies compared to the market as a whole, while here we find strong pro-connection bias by both measures.} % %The biases toward constituent and contributor companies that we observe are consistent with the idea that members invest based on information about firms, as well as the signaling and bonding stories: members probably know more about local and contributor companies, and they may be particularly interested in signaling and/or aligning incentives with them. %The bias away from companies whose industries match one's committee jurisdictions is not consistent with the idea that members have information about regulation and invest based on that information, but this is somewhat contradicted by a bias toward companies that lobby bills referred to one's committee. %At this point we are inclined to believe that the lobbying-based measure is a more faithful representation of how policy oversight works, but more analysis is needed. % %In order to assess the mix of informational advantages and political considerations that drive members' investing behavior (and particularly their portfolio bias toward local companies and contributors), we now turn to an examination of the returns of members' portfolios, paying particular attention to differences in the performance of investments made in connected and unconnected portfolios. % %\end{comment} % %%While aggregation of investments across members and over time likely obscures many important dynamics, we focus on a central finding: while members of Congress match the market overall, their local portfolios beat the market and their contributor portfolios underperform the market, suggesting that informational advantages and political considerations operate in different proportions in the two types of investments. % % %%\footnote{Note however that \citet{cohen2008swi} find that mutual fund managers (unconditionally) underweight connected companies, even though they enjoy a higher return on them.} but it could be explained by signaling or ethics. % %%\footnote{It is also possible that members tilt their investment portfolios away from their policymaking portfolios in order to diversify away risk. After retiring, members of Congress often land lobbying contracts or other employment related to their in-office policy specialty. Since their human capital is tied up in the industries they oversee through committee assignments, perhaps they seek to reduce their overall exposure to that industry.} % %%In much of the rest of the paper, we assert that members of Congress have valuable information about firms and markets and benefit (albeit modestly) from using it in their investing. The tendency of members to invest in constituent companies and contributor companies is consistent with this view: members likely know more about what is happening in their state and among their contributors because they have extra contact with these firms, both in discussions about policy and in campaign fundraising situations. Members may also use their portfolios to bond themselves to constituent and contributor interests; again, their tendency to invest in these companies in consistent with that. What could explain the tendency of members of Congress to tilt their investing portfolios away from their committee jurisdictions? % %%Second, ethics considerations might be responsible. As noted above, there are currently no rules against investing based on political knowledge, but Congressional ethics rules broadly prohibit profiting from one's political position. It is also possible that members are trying to avoid attracting unwelcome attention. Investing in local companies may be unlikely to invite a scandal but investing in a company to which one is closely linked through a committee assignment might. % % %\begin{comment} %\section{Event Study of Timing of Transactions} % %We now turn to the performance of members' investments. % %\subsection{Methodology} % %We use an event study approach to examine whether members have well-timed stock transactions. The basic idea is to calculate, for each trading day around a transaction (e.g. {-2,-1,0,1,2}) the average return for the traded stocks, and to see whether stocks on average rose or fell before and after the member chose to sell or buy. Let $t$ be an event-day indicator that ranges from $t=(-255,-254,...,255)$ with $t=0$ denoting the day at which a member sold or bought a particular stock. Let $i=(1,...,N)$ be an indicator of the traded stocks in a particular sample. For several samples of buy and sell transactions, we compute the cumulative abnormal return (CAR) on each event-day. First, we compute the daily average abnormal return for the sample transactions as %\[ %\overline{AR}_t = \frac{\sum_i^N w_i (R_{i,t}-R_{m,t})}{\sum_i^N w_i } %\] %where $R_{i,t}$ is the return from sample transaction $i$ on the calendar day that corresponds to event day $t$, $R_{m,t}$ is the return on the CRSP value weighted market index, and $w_i$ is the trade weight of transaction $i$. We use transaction values as our trade weights. As noted above, we use midpoints of the ranges reported in the FDRs to obtain the transaction value unless the exact amount is reported. The cumulative abnormal return for a given day $t$ is then computed as %\[ %CAR_t = \sum_{T=-255}^t \overline{AR}_t %\] %To make the figures more easily interpretable, we normalize each CAR series by subtracting the value of $CAR_0$ so that the CAR is always zero at the trading day $t=0$. In addition to the value-weighted approach for the CARs described above we also compute an equal-member weighted CAR where the $\overline{AR}_t$ is first computed for the transactions of each member separately and then averaged across members. Intuitively, the value weighted approach examines how the value-weighted average of all transactions performed relative to the market, while the equal-member weighted approach examines how the transactions of the average member performed relative to the market. % %Notice that we use the CAR analysis primarily as a descriptive tool to describe the timing and performance of the members transactions vis-a-vis the market. Further below we provide more formal tests based on the calendar time portfolio approach. % %\subsection{Results of CAR Analysis} % %\subsubsection{Overall CAR} %Figure \ref{fig:car_overall} shows the CAR plots for the buy and sell samples of all members (value weighted and equal-member weighted), as well as the subsamples of only the Democratic and Republican members. %Figure \ref{fig:car_overall_bestmembers} shows the CAR plots for best and worst five selling and buying members. % % %\subsubsection{State Connected versus Unconnected} %In this section we compare the timing of stock trades of companies headquartered in a member's state relative to stocks trades of companies that are not. Figure \ref{fig:car_overall_state_connected} shows the CAR plots for the buy and sell samples of selected subsamples: all members, Republicans, House Republicans, and Senate Democrats. For these subsets, %% We selected these samples because they best illustrate the pattern that in general %members tend to have better timed stock transactions with companies that are in their own state or district compared to out-of-district companies. The same pattern roughly holds for many other subsamples. Figure \ref{fig:car_overall_state_connected_best} shows similar CAR plots for four selected members. Later we conduct systematic analysis to examine these patterns in more detail. % % %\subsubsection{Contribution-Connected versus Unconnected} %In this section we compare the performance of stock trades of companies which contributed to a members financially, relative to stocks trades of companies that did not contribute. %Figure \ref{fig:car_overall_contrib_connected} shows the CAR plots for the buy and sell samples of selected subsamples: All members, Rebulicans, House Rebulicans, and Senate Democrats. The results suggest that members fare worse with stocks transactions of companies that contributed to the members compared to their transactions with companies that did not contribute to them. % % %\subsubsection{Committee Connected versus Unconnected} %TBA % % %\section{Analysis of Calendar Time Portfolio Returns} % % %\subsection{Methodology} % %We use a standard calendar time portfolio approach to examine the risk adjusted returns that members earned on their portfolios. For members $h\in(1,...,H)$ we observe their holdings at the end of each year as well as the transactions that occurred within each year. We use this information to construct a member's monthly portfolios returns $R_{h,t}$ for each of $t\in(1,...,T)$ months between January 2004 and January 2008. Let $i\in(1,...,N_t)$ be an indicator for stocks held by member $h$ in month $t$, each with a dollar amount of $w_i$. A member's monthly portfolios return is computed as the weighted average of the monthly returns of the portfolio's underlying stocks, $R_{h,t}=\frac{\sum_{i=1}^N w_{h,i} R_{i,t}}{\sum_{i=1}^N w_{i,t}}$. Weights are computed at the beginning of each month; we therefore assume (as is standard) that all transactions reported in a given month take place at the end of the month.\footnote{\citet{barber2000thy} show that these simplifying assumptions only cause minor differences in the return calculations even with high portfolio turnover. In our data, the turnover rates are low so our return calculations should only marginally be affected by ignoring the intra-month trading activity.} %%In what follows we often divide the portfolio into connected and unconnected components; For the analysis where we separate connected and unconnected stocks, we assign the stocks in each %member's portfolio into two portfolios, one for holdings in : connected and unconnected. Connected stocks are weighted by the member's dollar holdings in the connected portfolio, and %non-connected stocks are weighted by the member's dollar holdings in the non-connected portfolio. % %We compute two kinds of portfolios, corresponding to different ways of averaging returns across members. Value weighted calendar time portfolios $R_{p,t}$ are computed by averaging across members, weighting individual member portfolios by the member’s total dollar holdings in that month, i.e. $R_{p,t}=\frac{\sum_h=1^H w_{h,t} R_{h,t}}{\sum_h=1^H w_{h,t}}$ where $w_{h,t}=\sum_{i=1}^N w_{i,t}$ is the total value invested by member $h$ in month $t$. This approach corresponds to an investment strategy of investing in a portfolio that mimics dollar for dollar the aggregate Congressional portfolio. We also compute monthly equal-member weighted calendar time portfolios by taking a simple average across members portfolios for each calendar month, so that $w_{h,t}=1$ and every member is weighted equally regardless of how much she invested. %%We apply the same approach for the analysis that separates connected and unconnected portfolios. % %To test whether members outperform the market we regress the risk-adjusted calendar time portfolios on a standard set of controls. (This is known in empirical finance as the ``four-factor" model \-- a regression of an excesss return series on the monthly returns from the three \citet{fama1993crf} factors and Carhart’s (1997) momentum factor.\nocite{carhart1997pmf}.) The regression yields an intercept (commonly called ``alpha") that calculates the risk-adjusted monthly returns on the portfolio: %\[ %R_{p,t}-R_{f,t} = \alpha + \beta(R_{m,t}-R_{f,t})+s SMB_t + h HML_t + m MOM_t + e_t, %\] %where $R_{f,t}$ is the risk-free rate of return (ie. the return on a 1-month Treasury bill), $R_m,t$ is the normal market return, $SMB_t$ is the size premium (small minus big), $HML_t$ is the value premium (high minus low), $MOM_t$ is the momentum factor (prior high return minus prior low returns).\footnote{We are grateful to Kenneth R. French for providing the factor data in his data library at \url{http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html}} To compare the excess returns for connected and unconnected portfolios we also consider a hedged portfolio in which the dependent variable is the difference between the (risk free adjusted) return on the connected portfolio minus the return on the unconnected portfolio. This mimics an investment strategy of going long in the connected and short in the unconnected stocks. % % %\subsection{Results of Portfolio Tests} % %\subsubsection{All members} % %Table \ref{tab:alpha_all_members} shows the annualized alpha estimates for the value-weighted portfolios constructed from all members.\footnote{The member-equal weighted estimates are qualitatively similar and not shown here.} The first row shows the average excess returns for the portfolio of all stocks. We find that the Congressional portfolio performs about as well as the market; the excess return is an insignificant minus .4 percent. This finding stands in contrast to the results from \citeauthor{ziobrowski2004arc} who found that in the 1990's the stock transactions of Senators produced excess returns. Notice that in contrast to this earlier analysis, our returns estimates are based on real portfolios and therefore approximate the actual returns.\footnote{\cite{ziobrowski2004arc} create a buy and sell portfolio from the record of transactions by assuming that stocks are held for a certain amount of time after the transaction.} % %Rows 2-4 show the results when we split the portfolios into stocks of companies that are located in and out of a member's own state. We find that the in-state portfolio on average outperforms the market by about 3.3 percent annually. %The returns on the out-of-state portfolio slightly underperformed the market. The returns on the hedged portfolio suggest that on average the connected stocks outperform the unconnected stocks by about 4.2 percent annually. %As we saw earlier, members also tend to overweight in-state companies in their portfolios. Based on the fact that the in-state portfolio outperforms the market, this appears to be a wise strategy, suggesting that members are not only bonding themselves to local interests by investing locally but also picking winners. % %Rows 5-7 show the results when we split the portfolios into stocks of companies that contribute to a members' campaign funds and stocks of companies that do not contribute. Here a stock is coded as connected if the company is among the top 20 PAC contributors over the period 2004-2007.\footnote{Results for other definitions, such as the top 40 contributors or the top 1\% of contributors, are qualitatively similar.} We find that the portfolio of contribution-connected stocks on average underperform the market by about 4 percent annually. The returns on the unconnected portfolio is indistinguishable from the normal market return. The returns on the hedged portfolio suggest that on average the connected stocks underperform the unconnected stocks by about 3.6 percent annually. So despite the fact that members invest more heavily in companies that provide campaign contributions, these investments perform worse than those in companies that do not contribute. This is consistent with the idea that members invest in companies in part to seal political exchanges rather than to profit from an informational advantage. % %Rows 8-10 examine the differences in returns for the committee jurisdiction connection. We find no differences in the returns of the connected and the unconnected portfolio. As indicated above, this may be because our definition of committee-connected companies provides only a weak signal. If it is indeed true that members do not enjoy excess returns in trading committee-connected stocks, it would appear that members either do not have an informational advantage due to their regulatory power or they do not use it in investing. % %Rows 11-13 examine the differences in returns for the committee lobbying connection (ie companies that lobbied bills before a member's committees). Here we find that the connected portfolio underperforms the market by almost 2\% while the unconnected portfolio outperforms the market by about 1.3\% annually. (The first point estimate has a p-value of .07 and the second of .47.) The difference between them is about -3.3, with a p-value of .12. Thus while the estimate is somewhat imprecise, these connected investments also seem to underperform compared to unconnected investments, suggesting that whatever regulation-related market knowledge they have is not translated into abnormal investment returns. % % %\subsubsection{Members} % %In this section we examine the distribution of alpha returns across members. We compute annualized four-factor alphas for each of the 429 members who report at least two years of stock holdings using the calendar time portfolio approach outlined above. Figure \label{fig:member_returns} plots the return estimates against the average annual value of a member's investment. There is a significant variation in members' returns ranging from 45 \% annual excess returns for John Yarmuth (D-Ky) to -43 \% for Bob Inglis (R-SC). In line with the overall portfolio results shown above, the average returns across members is -2.9 \%. The returns are roughly normally distributed. On average members who invest more earn higher returns as indicated by the linear fit. % %What else accounts for the variation in member's returns? We regress the member's return on a set of explanatory variables including the member's party, the year first elected to Congress, a dummy for whether the member served in a leadership position (committee chairman or ranking member), a ``revolving door score" obtained from the CRSP which consists of the number of a member's staffers who either came to Capitol Hill after representing private interests or left the member's staff for a lobbying position. We also include a two dummy variables that capture allegations of unethical activities. The first dummy, Named Corrupt 1, is coded one for those members that have once or twice between mentioned in the CREW's list of the 20 Most Corrupt members of Congress over the last four years. The second dummy, Named Corrupt 2, is coded one for those members that have been named three or more times over the last four years. Table \ref{tab:member_return_regression} shows regression results.\footnote{Notice that this regression ignores the estimation uncertainty in the members' alpha returns.} % %We find that higher revolving door scores are associated with higher returns. Ten additional revolving door staffers are associated with about a 0.1 increase in the average annual returns. Seniority is negatively correlated with returns. Members that enter Congress one year later have 0.1 lower returns on average. Somewhat remarkably, we find that the 128 members with leadership position earn 2.5 higher annual excess returns on average. Similarly, those members named more than twice on the list of corrupt members on average earn about 4.8 higher returns. Notcie that only 4 members fall into this group, however. Finally, we find no significant different betwee the average returns of Republican and Democratic members. % % %\section{Related literature} % % %The empirical literature examining the investments of members of Congress consists of one published paper and one working paper. %\citet{ziobrowski2004arc} uses transactions reported in the 1990s to demonstrate that Senators experienced large abnormal returns. %As an indication of the uncanny timing exhibited by Senators in the period they consider, %stocks sold by Senators outperformed the %market by 25 percent during the 12 months prior to the sell date and %remained fairly flat thereafter, while stocks that they purchased %beat the market by only 3 percent prior to the buying date, but by %almost 28 percent in the year following the transaction. %In subgroup analysis they fail to find a difference in the returns %of Democrats and Republicans, and if anything junior Senators %appear to be smarter traders. Intriguingly, they find smaller abnormal %returns in 1996 and 1997, which they suggest may be due to the unfavorable %media attention drawn to well-timed stocks trades among members by %\cite{boller1995}. The Ziobrowski et al paper generated attention in the media% %\footnote{The study was cited on the \emph{New Yorker}'s {}``Financial Page'' %of October 31, 2005; it was described in a \emph{Washington Spectator %}article on January 1, 2006, {}``An Ethics Quagmire: Senators Beat %the Stock Market\textemdash{}and Get Rich\textemdash{}With Insider %Information''; and it was featured on {}``Nieman Watchdog -- Questions %the press should ask'' on March 10, 2006.% %} and in Congress itself.\footnote{The ``Stop Trading on Congressional Knowledge'' (STOCK) Act was introduced in 2006 as H.R. 5015 by Reps Slaughter and Baird and reintroduced %in 2007 (by the same members) as H.R. 2341. The bill proposed to prohibit %members of Congress, congressional staffers, and members of the executive %branch from trading on {}``material non-public information,'' defined %as information members acquire as a result of their employment by %the federal government. For more on policy issues surrounding stock %trading by members of Congress, see \cite{George2008}.} %%Our work uses \citet{ziobrowski2004arc} as a starting-off point, replicating some of his analyses but considerably increasing the %%amount of subgroup analysis (particularly on connections between members and companies). %%We have fewer years of data but disclosures from both the House and the Senate, and whereas their data included %%only transactions we also have reported end-of-year holdings, allowing us to reconstruct the entire stock portfolio at any time. % %Consistent with the idea that members face considerable constraints that offset their informational advantages, Lenz (2009) finds %that members of Congress did not experience any abnormal wealth gains between 1995 and 2005 compared to similar %subjects in the Panel Study of Income Dynamics (PSID). The finding extends to stock holdings as well. %MORE HERE FROM GABE % %%\subsection{Beating the market} % %A large literature in empirical finance examines the investment behavior and performance of %other groups of investors, providing useful benchmarks and techniques. %\cite{barber2000thy} study the stock portfolios of over %65,000 retail investors between 1991 and 1996, using data from a discount %brokerage. They find that individual investors on average underperform %the market (earning annual returns of 16.4 percent in a period when %the market returned 17.9 percent annually), and that more active traders %did significantly worse, largely due to trading fees. %%The thrust of this well-known paper is captured in its title: ``Trading is Hazardous to your Wealth.'' % %Towards the other extreme of investor sophistication, \cite{jeng2003eri} %examine trades reported by corporate executives (who are required %by insider trading rules to report sales and purchases they make of %their own company stock). As did \citet{ziobrowski2004arc}, \citeauthor{jeng2003eri} evaluate the timing of insider trades %by creating portfolios based on the reported trades: a ``buy'' %portfolio that reflects the stocks purchased by insiders and a ``sell'' %portfolio that reflects the stocks sold by insiders. They find that %the sell portfolio does just as well as the market but that the buy %portfolio beats the market by more than 6\% per year. % , a handsome return for any fund manager. % %\citet{cohen2008swi} examine the portfolio choices of mutual fund %managers in the context of their social connections with the managers %of public companies. They find that portfolio managers tend to invest %more intensively in companies to which they are connected by educational %ties (e.g., the portfolio manager attended the same business school %at the same time as the CEO) and that these ``connected'' investments %perform better than non-connected investments.\footnote{They find that managers place a disproportionate weight on connected companies conditional on investing in those companies; unconditionally, they tend to underweight connected companies. In other words, the portfolio weight a fund manager assigns to a particular company, conditional on holding it, is higher if the manager and firm management are connected, but connected companies as a group have a smaller share of fund managers' portfolios than their market capitalization would suggest. The combination of conditionally high but unconditionally low portfolio weights might make sense if fund managers are choosing carefully among the connected companies: they perform well on connected companies both by choosing some companies to hold and others to avoid. Still, \citeauthor{cohen2008swi} are unable to explain why fund managers would choose not to hold a larger stake in connected companies if they beat the market so handily on this sub-portfolio.} %% CHECK WHETHER IT'S REALLY MARKET CAP. %They further show that %returns are concentrated around news announcements by portfolio %companies, suggesting that social ties allow connected portfolio managers %to obtain market-relevant news from company executives in advance of %other investors. Intriguingly, they find that returns are higher the %closer the connection is: on average portfolio managers do better %when they invest in companies whose CEO attended the same university %at the same time they did, for example, than when the CEO attended %the same school at a different time. They interpret their findings %as evidence that valuable market information travels through social %networks. %%% AE: can we indicate the lack of fixed effects here? % %\citet{coval1999hbh, coval2001gii} also examine portfolio choices and performance by mutual fund managers. %The principal finding of \citet{coval1999hbh} is that mutual fund managers prefer to invest in companies that are headquartered closer to their homes. The authors calculate the distance from each fund manager to the top holdings in that manager's portfolio (weighted by the size of the holdings) and the distance from the fund manager to the entire market (weighted by market capitalization of the companies), and find that fund managers are about 180 kilometers closer to their portfolios than they are to the market, indicating a substantial ``domestic home bias." In their 2001 paper, they show that fund managers on average exhibit a modest bias toward the stocks of local (defined as headquartered within 100 kilometers) companies, with the average fund manager investing a little under 7\% of her assets locally, even though only 6.16\% of the market is located within her local area. %% [Zhu] and [Huberman 2001] report that US individual investors show a similar local bias. %Consistent with \cite{cohen2008swi}, they find that mutual fund managers enjoy a modest information advantage with respect to local companies: between 1975 and 1984, the local component of their portfolios outperformed the nonlocal component by about 2\% per year. They find no evidence, however, that fund managers' local portfolios outperformed the market after 1985. Examining brokerage accounts in the 1991-1996 period, \citet{zhu:lbi} also finds no evidence that investors with a stronger propensity to invest locally enjoy higher returns, and shows that investment behavior seems more driven by familiarity (either through proximity or advertising) than by responses to fundamental information. % % %%%% MORE DISCUSSION % %\end{comment} \newpage \begin{singlespacing} \bibliography{st} \end{singlespacing} \newpage \section*{Tables} % Some descriptive Stata \begin{table}[hbt!]\caption{\label{tab:summarystats} The common stock holdings and transactions of Members of Congress - Annual Averages 2004-2008} \begin{center} \begin{tabular}{l|rr|rrrr} \multicolumn{ 1}{c}{} & \multicolumn{ 2}{|c|}{Holdings} & \multicolumn{ 4}{c}{Annual Transactions} \\ %\cline{4-7} \hline \multicolumn{ 1}{c}{} & \multicolumn{ 2}{|c|}{} & \multicolumn{ 2}{c}{Buys} & \multicolumn{ 2}{c}{Sells} \\ %\cline{2-3} \multicolumn{ 1}{c|}{} & \$ Value & Number & \$ Value & Number & \$ Value & Number \\ \hline \hline Min & 501 & 1 & 0 & 0 & 0 & 0 \\ 25th Percentile & 17,337 & 1 & 0 & 0 & 11,010 & 1 \\ Median & 80,164 & 5 & 16,277 & 2 & 38,765 & 3 \\ 75th Percentile & 404,047 & 18 & 104,520 & 8 & 173,934 & 11 \\ Max & 140,767,979 & 331 & 32,253,189 & 424 & 47,615,848 & 479 \\ \hline Mean & 1,611,183 & 18 & 396,562 & 18 & 611,198 & 22 \\ \hline \multicolumn{7}{p{5.5in}}{\tiny {\it Note:} Summary statistics are annual (aggregated) averages across the 2004-2008 period based on end-of-year financial disclosure reports for 453 members of Congress that report common stocks between 2004 to 2008. Values are reported in bands and imputed based on a log-normal model that was fitted to each value band for the group of Members that report exact amounts within each band (see text for details).} \end{tabular} \end{center} \end{table} \begin{table}[hbt!]\caption{\label{tab:overallalpha} Percentage Monthly Abnormal Return for Aggregate Congressional Portfolio and the Average Congressional Member} \begin{center}\small \begin{tabular}{lcccccc} \hline & Excess & \multicolumn{ 4}{c}{Coefficient Estimate on:} & Adjusted \\ \cline{3-6} & Return & ($R_{m,t}-R_{f,t}$) & $SMB_t$ & $HML_t$ & $MOM_t$ & $R^2$ \\ \hline \multicolumn{ 7}{c}{} \\ \multicolumn{ 7}{c}{Panel A: Gross Percentage Monthly Returns for Aggregate Congressional Portfolio} \\ \multicolumn{ 7}{c}{} \\ \hline CAPM & -0.269 & 0.925 & & & & 0.96 \\ & (0.095) & (0.038) & & & & \\ Carhart 4-Factor & -0.239 & 0.920 & -0.040 & 0.076 & -0.065 & 0.96 \\ & (0.099) & (0.037) & (0.053) & (0.055) & (0.037) & \\ \hline \multicolumn{ 7}{c}{} \\ \multicolumn{ 7}{c}{Panel B: Gross Percentage Monthly Returns for the Average Member} \\ \multicolumn{ 7}{c}{} \\ \hline CAPM & -0.319 & 0.979 & & & & 0.96 \\ & (0.093) & (0.032) & & & & \\ Carhart 4-Factor & -0.263 & 0.933 & 0.081 & 0.090 & -0.125 & 0.98 \\ & (0.080) & (0.025) & (0.042) & (0.042) & (0.030) & \\ \hline \multicolumn{7}{p{5.5in}}{\tiny {\it Note:} Returns are based on end-of-year financial disclosure reports for 453 members of Congress that report common stocks between 2004 to 2008. Daily portfolio returns are computed based on reported holdings and transactions and then compounded to the monthly level (see text for details). N=60. Panel A presents results for the gross monthly return on a portfolio that mimics the aggregate investments of all members of Congress (value-weighted). Panel B presents results for the gross return on a portfolio that mimics the investment of the average Member of Congress (equal Member weighted). CAPM is the result from a time-series regression of the Member excess return on the market excess return ($R_{m,t}-R_{f,t}$). Carhart 4-factor is the result from a time-series regression of the Member excess return on the Fama and French (1993) mimicking portfolios (the market excess return, a zero-investment size portfolio ($SMB_t$), a zero-investment book-to-market portfolio ($HML_t$) and the \citet{carhart1997pmf} momentum factor ($MOM_t$). Robust standard errors are presented in parentheses.} \end{tabular} \end{center} \end{table} \clearpage \begin{table}[hbt!]\caption{\label{tab:alphaforsubgroups} Percentage Monthly Abnormal Return for Selected Subgroups} \begin{center}\small \begin{tabular}{l|cc|cc} & \multicolumn{ 2}{|c}{Aggregate Portfolio} & \multicolumn{ 2}{|c}{Average Member Portfolio} \\ & \multicolumn{ 2}{|c}{Excess Return} & \multicolumn{ 2}{|c}{ Excess Return} \\ \hline & \multicolumn{1}{p{1in}}{ \centering CAPM} & \multicolumn{1}{p{1in}|}{\centering 4-Factor} & \multicolumn{1}{p{1in}}{\centering CAPM} & \multicolumn{1}{p{1in}}{\centering 4-Factor} \\ \hline House & -0.22 & -0.183 & -0.27 & -0.181 \\ & (0.099) & (0.112) & (0.161) & (0.100) \\ Senate & -0.334 & -0.335 & -0.111 & -0.081 \\ & (0.122) & (0.122) & (0.107) & (0.108) \\ \hline Power Committee House & -0.176 & -0.093 & -0.254 & -0.134 \\ & (0.145) & (0.143) & (0.233) & (0.152) \\ Power Committee Senate & -0.293 & -0.248 & -0.097 & -0.069 \\ & (0.139) & (0.134) & (0.098) & (0.106) \\ No Power Committee & -0.283 & -0.325 & -0.272 & -0.214 \\ & (0.113) & (0.130) & (0.121) & (0.088) \\ \hline 2004-2006 & -0.173 & -0.250 & -0.049 & -0.099 \\ & (0.128) & (0.095) & (0.584) & (0.119) \\ 2007-2008 & -0.312 & -0.214 & -0.633 & -0.159 \\ & (0.025) & (0.199) & (0.021) & (0.221) \\ \hline Democrats & -0.342 & -0.301 & -0.274 & -0.181 \\ & (0.133) & (0.127) & (0.144) & (0.098) \\ Republicans & -0.17 & -0.185 & -0.206 & -0.138 \\ & (0.092) & (0.103) & (0.155) & (0.103) \\ \hline Seniority Low & -0.086 & 0.004 & -0.273 & -0.171 \\ & (0.110) & (0.106) & (0.147) & (0.115) \\ Seniority Medium & -0.568 & -0.62 & -0.198 & -0.156 \\ & (0.120) & (0.126) & (0.170) & (0.100) \\ Seniority High & -0.289 & -0.354 & -0.227 & -0.133 \\ & (0.124) & (0.116) & (0.156) & (0.110) \\ \hline Net Worth Low & -0.634 & -0.526 & -0.376 & -0.264 \\ & (0.184) & (0.158) & (0.219) & (0.150) \\ Net Worth Medium & -0.258 & -0.314 & -0.047 & -0.013 \\ & (0.104) & (0.093) & (0.112) & (0.092) \\ Net Worth High & -0.273 & -0.264 & -0.274 & -0.21 \\ & (0.092) & (0.091) & (0.141) & (0.086) \\ \hline Portfolio Size Low & -0.399 & -0.397 & -0.165 & -0.071 \\ & (0.169) & (0.168) & (0.227) & (0.145) \\ Portfolio Size Medium & -0.54 & -0.493 & -0.309 & -0.198 \\ & (0.198) & (0.152) & (0.175) & (0.126) \\ Portfolio Size High & -0.259 & -0.245 & -0.221 & -0.192 \\ & (0.080) & (0.080) & (0.072) & (0.052) \\ \hline\hline \multicolumn{5}{p{5.1in}}{\tiny {\it Note:} Alpha returns are reported for selected subgroups. Aggregate returns are for a portfolio that mimics the aggregate investments of all members of Congress that belong to a specific group or time period (value-weighted). Average Member returns are for a portfolio that mimics the investments of the average Member of Congress in a specific group or time period (equal Member weighted). CAPM is the result from a time-series regression of the Member excess return on the market excess return. Carhart 4-factor is the result from a time-series regression of the Member excess return on the Fama and French (1993) mimicking portfolios and the \citet{carhart1997pmf} momentum factor. Heteroskedasticity and autocorrelation consistent Newey-West standard errors are presented in parentheses. Heteroskedasticity and autocorrelation consistent Newey-West standard errors are presented in parentheses.} \end{tabular} \end{center} \end{table} \clearpage \begin{table}[hbt!]\caption{\label{tab:ziobrep} Returns on Transaction-Based Portfolio} \begin{center} \begin{tabular}{lccc} % \multicolumn{4}{c}{Table A.1: Returns on Transaction Based Portfolios}\\ \hline & \multicolumn{ 3}{c}{Portfolio} \\ & Buys & Sells & Long/Short \\ \hline \emph{All Members}: & & & \\ CAPM alpha & -0.127 & -0.186 & 0.059 \\ & (0.092) & (0.052) & (0.111) \\ Fama-French alpha & -0.114 & -0.211 & 0.096 \\ & (0.081) & (0.048) & (0.083) \\ \hline \emph{Senate}: & & & \\ CAPM alpha & -0.234 & -0.251 & 0.016 \\ & (0.106) & (0.089) & (0.144) \\ Fama-French alpha & -0.248 & -0.284 & 0.036 \\ & (0.117) & (0.074) & (0.136) \\ \hline \emph{House}: & & & \\ CAPM alpha & -0.083 & -0.103 & 0.020 \\ & (0.115) & (0.102) & (0.136) \\ Fama-French alpha & -0.050 & -0.117 & 0.067 \\ & (0.079) & (0.097) & (0.101) \\ \hline \hline \multicolumn{4}{p{3.5in}}{\tiny {\it Note:} Monthly alpha returns for calendar time portfolios that mimics the value-weighted investments in stocks bought or sold by members over the 2004-2008 period. Following the procedure described in Ziobrowski et al 2004, we use a fixed holding period of 255 days and impute dollar values using band midpoints and a maximum value of \$250,000. Long-short is the monthly average return of a zero cost portfolio that holds the portfolio of bought stocks and sells short the portfolio of sold stocks. CAPM alpha is the result from a time-series regression of the portfolio excess return (i.e. raw return minus risk-free rate) on the market excess return. Fama-French alpha is the result from a time-series regression of the portfolio excess return on the three Fama and French (1993) mimicking portfolios. Heteroskedasticity- and autocorrelation-consistent Newey-West standard errors are presented in parentheses.} \end{tabular} \end{center} \end{table} \clearpage \begin{table}[hbt!]\caption{\label{tab:uncondportfolio} Portfolio Weights as a Function of Member-Firm Connections} \begin{center}\small \begin{tabular}{l|ccccc} \hline Model & (1) & (2) & (3) & (4) & (5) \\ \hline Dependent Variable: & \multicolumn{ 5}{c}{Portfolio Weight} \\ Mean: & \multicolumn{ 5}{c}{3.86} \\ \hline In District & 48.08 & 42.93 & 47.67 & 47.88 & 37.94 \\ & (8.09) & (8.39) & (8.07) & (8.05) & (8.27) \\ Lobbying (Any) & 0.38 & 0.62 & & & \\ & (0.64) & (0.62) & & & \\ Contributions (Any) & 15.06 & 20.18 & & & \\ & (2.49) & (5.25) & & & \\ In District \& Lobbying (Any) & & 12.44 & & & \\ & & (2.71) & & & \\ In District \& Contributions (Any) & & 44.07 & & & \\ & & (19.57) & & & \\ Lobbying (Any) \& Contributions (Any) & & 32.87 & & & \\ & & (18.72) & & & \\ In District \& Contributions(Any) \& Lobbying (Any) & & 153.13 & & & \\ & & (42.92) & & & \\ Lobbying ($>$ p50) & & & 0.70 & & \\ & & & (1.23) & & \\ Contributions ($>$ p50) & & & 24.43 & & \\ & & & (4.35) & & \\ Lobbying Strength & & & & -0.00 & -0.01 \\ & & & & (0.03) & (0.03) \\ Contribution Strength & & & & 0.06 & 0.05 \\ & & & & (0.02) & (0.02) \\ Lobbying Strength $\cdot$ In District & & & & & 1.12 \\ & & & & & (0.89) \\ Contribution Strength $\cdot$ In District & & & & & 0.17 \\ \hline Members Fixed Effects & x & x & x & x & x \\ Firms Fixed Effects & x & x & x & x & x \\ \hline % N & 1,185,501 & 1,185,501 & 1,185,501 & 1,185,501 & 1,185,501 \\ N & \multicolumn{ 5}{c}{1,166,160} \\ \hline \hline \multicolumn{6}{p{5.5in}}{\tiny {\it Note:} Regression coefficients with standards errors (clustered by Members) in parenthesis. The dependent variable is the \emph{portfolio weight}, i.e. the share of holdings of a firm in a Member's portfolio (in basis points). Members' portfolios are computed as average holdings over the 2004-2008 period. \emph{In District} is a binary indicator for firms that are connected to a Member since they are located in a Member's home district. \emph{Lobbying (any)} is a binary indicator for firms that are connected to a Member since they lobbied a Committee on which the Member served. \emph{Contributions (any)} is a binary indicator for firms that are connected to a Member since they provided her with campaign contributions. \emph{Lobbying ($>$ p50)} and \emph{Contributions ($>$ p50 )} are binary indicators for firms that provided more than the median lobbying or contribution amount for each Member. \emph{Lobbying Strength} and \emph{Contribution Strength} measure a firm's share of lobbying or contributions relative to each Member's total lobbying or contributions (in basis points). All regressions include a full set of Members and firms fixed effects (coefficients not shown here).} \end{tabular} \end{center} \end{table} \clearpage \begin{table}[hbt!]\caption{\label{tab:connectedunconnectedalpha} Percentage Monthly Abnormal Return for Connected and Unconnected Stocks} \begin{center}\footnotesize \begin{tabular}{l|ccc|ccc} & \multicolumn{ 3}{c}{Aggregate Congressional Portfolio} & \multicolumn{ 3}{|c}{Average Member Portfolio} \\ \hline & \scriptsize{Connected} & \scriptsize{Unconnected} & \scriptsize{Long/Short} & \scriptsize{Connected} & \scriptsize{Unconnected} & \scriptsize{Long/Short} \\ \hline \multicolumn{ 7}{c}{} \\ \multicolumn{ 7}{c}{Panel A: Excess Returns from CAPM} \\ \multicolumn{ 7}{c}{} \\ \hline Lobbying (Any) & -0.277 & -0.234 & -0.043 & -0.244 & -0.196 & -0.048 \\ & (0.138) & (0.163) & (0.232) & (0.113) & (0.130) & (0.171) \\ Lobbying ($>p50$) & -0.21 & -0.305 & 0.094 & -0.241 & -0.265 & 0.024 \\ & (0.158) & (0.129) & (0.211) & (0.128) & (0.107) & (0.154) \\ Contributions (Any) & -0.083 & -0.311 & 0.228 & -0.147 & -0.28 & 0.133 \\ & (0.234) & (0.094) & (0.230) & (0.175) & (0.086) & (0.151) \\ Contributions ($>p50$) & -0.216 & -0.277 & 0.06 & -0.14 & -0.312 & 0.172 \\ & (0.284) & (0.093) & (0.270) & (0.176) & (0.090) & (0.140) \\ Lobbying \& Contributions & -0.055 & -0.314 & 0.258 & -0.141 & -0.265 & 0.124 \\ & (0.243) & (0.094) & (0.238) & (0.163) & (0.091) & (0.136) \\ In District& 0.07 & -0.283 & 0.353 & 0.354 & -0.335 & 0.688 \\ & (0.426) & (0.088) & (0.424) & (0.192) & (0.093) & (0.173) \\ District \& Contributions & 0.277 & -0.288 & 0.564 & 0.354 & -0.327 & 0.681 \\ & (0.663) & (0.093) & (0.653) & (0.190) & (0.094) & (0.169) \\ District \& Lobbying & 0.579 & -0.298 & 0.877 & 0.433 & -0.324 & 0.757 \\ & (0.486) & (0.092) & (0.475) & (0.192) & (0.091) & (0.155) \\ \hline \multicolumn{ 7}{c}{} \\ \multicolumn{ 7}{c}{Panel B: Excess Returns from Carhart 4-Factor} \\ \multicolumn{ 7}{c}{} \\ \hline Lobbying (Any) & -0.174 & -0.31 & 0.136 & -0.124 & -0.249 & 0.125 \\ & (0.124) & (0.180) & (0.225) & (0.095) & (0.094) & (0.129) \\ Lobbying ($>p50$) & -0.111 & -0.33 & 0.219 & -0.115 & -0.264 & 0.149 \\ & (0.151) & (0.148) & (0.220) & (0.110) & (0.076) & (0.126) \\ Contributions (Any) & 0.03 & -0.302 & 0.332 & -0.019 & -0.259 & 0.239 \\ & (0.235) & (0.100) & (0.233) & (0.137) & (0.074) & (0.112) \\ Contributions ($>p50$) & -0.083 & -0.259 & 0.176 & 0.016 & -0.277 & 0.293 \\ & (0.244) & (0.101) & (0.238) & (0.139) & (0.079) & (0.118) \\ Lobbying \& Contributions & 0.052 & -0.301 & 0.353 & -0.038 & -0.227 & 0.189 \\ & (0.252) & (0.100) & (0.249) & (0.139) & (0.078) & (0.117) \\ In District & 0.044 & -0.246 & 0.29 & 0.423 & -0.272 & 0.696 \\ & (0.429) & (0.094) & (0.424) & (0.152) & (0.086) & (0.168) \\ District \& Contributions & 0.403 & -0.261 & 0.664 & 0.5 & -0.274 & 0.774 \\ & (0.681) & (0.098) & (0.679) & (0.178) & (0.084) & (0.204) \\ District \& Lobbying & 0.721 & -0.272 & 0.993 & 0.529 & -0.268 & 0.797 \\ & (0.521) & (0.097) & (0.514) & (0.173) & (0.078) & (0.177) \\ \hline \hline \multicolumn{7}{p{6in}}{\tiny {\it Note:} \tiny {\it Note:} Monthly alpha returns for calendar time portfolios of investments in connected and unconnected stocks over the 2004-2008 period. Aggregate returns are for a portfolio that mimics the aggregate investments of all members of Congress (value-weighted) in either connected or unconnected stocks. Average Member returns are for a portfolio that mimics the investments of the average Member of Congress (equal Member weighted) in either connected or unconnected stocks. Long-short is the monthly average return of a zero cost portfolio that holds the portfolio of connected stocks and sells short the portfolio of nonconnected stocks. The connections are defined as follows. \emph{In District} connected firms are connected to a Member since they are located in a Member's home district. \emph{Lobbying (any)} connected firms are connected to a Member since they lobbied a Committee on which the Member served. \emph{Contributions (any)} connected firms are connected to a Member since they provided her with campaign contributions. \emph{Lobbying ($>$ p50)} and \emph{Contributions ($>$ p50 )} connected firms are connected since they provided more than the median lobbying or contribution amount for each Member. CAPM is the result from a time-series regression of the Member excess return on the market excess return. Carhart 4-factor is the result from a time-series regression of the Member excess return on the Fama and French (1993) mimicking portfolios and the \citet{carhart1997pmf} momentum factor. Heteroskedasticity and autocorrelation consistent Newey-West standard errors are presented in parentheses. Robust standard errors are presented in parentheses.} \end{tabular} \end{center} \end{table} \clearpage \begin{comment} \begin{table}[hbt!]\caption{\label{tab:companyconnected} Percentage Monthly Abnormal Return for Company Level Connected and Unconnected Stocks} \begin{center}\footnotesize \begin{tabular}{l|ccc|ccc} & \multicolumn{ 3}{c}{Aggregate Congressional Portfolio} & \multicolumn{ 3}{|c}{Average Member Portfolio} \\ \hline & \scriptsize{Connected} & \scriptsize{Unconnected} & \scriptsize{Long/Short} & \scriptsize{Connected} & \scriptsize{Unconnected} & \scriptsize{Long/Short} \\ \hline \multicolumn{ 7}{c}{} \\ \multicolumn{ 7}{c}{Panel A: Excess Returns from CAPM} \\ \multicolumn{ 7}{c}{} \\ \hline Lobbying (Any) & -0.291 & -0.152 & -0.138 & -0.247 & -0.06 & -0.187 \\ & (0.104) & (0.178) & (0.237) & (0.144) & (0.197) & (0.230) \\ Contributions (Any) & -0.295 & -0.23 & -0.064 & -0.282 & -0.062 & -0.219 \\ & (0.141) & (0.153) & (0.251) & (0.172) & (0.121) & (0.207) \\ \hline \multicolumn{ 7}{c}{} \\ \multicolumn{ 7}{c}{Panel B: Excess Returns from Carhart 4-Factor} \\ \multicolumn{ 7}{c}{} \\ \hline Lobbying (Any) & -0.236 & -0.325 & 0.089 & -0.126 & -0.154 & 0.028 \\ & (0.096) & (0.202) & (0.242) & (0.100) & (0.119) & (0.133) \\ Contributions (Any) & -0.214 & -0.313 & 0.1 & -0.108 & -0.17 & 0.062 \\ & (0.118) & (0.171) & (0.236) & (0.112) & (0.110) & (0.143) \\ \hline \hline \multicolumn{7}{p{6in}}{\tiny {\it Note:} \tiny {\it Note:} Monthly alpha returns for calendar time portfolios of investments in connected and unconnected stocks over the 2004-2008 period. Aggregate returns are for a portfolio that mimics the aggregate investments of all members of Congress (value-weighted) in either connected or unconnected stocks. Average member returns are for a portfolio that mimics the investments of the average Member of Congress (equal Member weighted) in either connected or unconnected stocks. Long-short is the monthly average return of a zero cost portfolio that holds the portfolio of connected stocks and sells short the portfolio of nonconnected stocks. In contrast to table \ref{tab:connectedunconnectedalpha} the connections here are defined only at the company, not the company-member levels, so for all members a company is coded as connected if it provided campaign contributions (or lobbying depending on the connection) to \emph{any} member in the sample. CAPM is the result from a time-series regression of the Member excess return on the market excess return. Carhart 4-factor is the result from a time-series regression of the Member excess return on the Fama and French (1993) mimicking portfolios and the \citet{carhart1997pmf} momentum factor. Heteroskedasticity and autocorrelation consistent Newey-West standard errors are presented in parentheses. Heteroskedasticity and autocorrelation consistent Newey-West standard errors are presented in parentheses.} \end{tabular} \end{center} \end{table} \clearpage \end{comment} \section*{Figures} %\begin{figure}[!hbt] %\caption{\label{fig:double_power_law} Distribution of Value of Holdings and Number of Holdings Across Members} %% \centering % \includegraphics[scale=.9]{figs/double_power_law.pdf} %\tiny{Note: Ranked value and number of holdings are shown (x-axis, bottom and top scale, respectively) for the 453 members in our dataset. For members who report holdings in more than one year, we use the average across years in which holdings are reported for that member.} %\end{figure} % %\clearpage % %\begin{figure}[!hbt] %\caption{\label{fig:kerry} Portfolio Value and Transactions for John Kerry} %% \centering % \includegraphics[scale=1]{figs/kerry_trajectory_and_net_purchases.pdf} %\tiny{Note: The top panel shows John Kerry's daily portfolio value as reported in his FDRs and imputed by our methods (described in the paper). Daily variation is due to both market fluctuations and Kerry's transactions; jumps at the end of year are due to reporting inconsistencies year-to-year and error in imputing precise values for each investment from value bands.} %\end{figure} % %\clearpage \begin{figure}[!hbt] \caption{\label{fig:cumret} Cumulative Monthly Return for Aggregate Congressional Portfolio and the Average Congressional Member} % \centering \includegraphics[scale=.7]{figs/cumret.pdf} \tiny{Note: Cumulative monthly return is shown for a \$100 dollar position in each mimicking portfolio beginning in January 2004. The aggregate congressional portfolio mimics the aggregate investments of all members of Congress (value-weighted). Portfolio of the average congressional Member is based on a portfolio that mimics the investment of the average Member of Congress (equal Member weighted). The market portfolio is the value weighted market portfolio from NYSE, AMEX, and NASDAQ as computed by the CRSP.} \end{figure} \clearpage \begin{landscape} \begin{figure}[!hbt] \caption{\label{fig:member_returns} Members' Monthly Excess Returns and Average Portfolio Size 2004-2008} %\centering \includegraphics[scale=.62]{figs/returnsall_and_size_4FF} \tiny{Note: Monthly alpha return is Carhart 4-factor alpha obtained from a calendar time portfolio regression of each Member's excess return on the Fama and French (1993) mimicking portfolios and the \citet{carhart1997pmf} momentum factor.} \end{figure} \end{landscape} \clearpage \begin{figure}[!hbt] \caption{\label{fig:benchmarks} Benchmark Estimates for Different Investor Groups} %\centering \includegraphics[scale=.5]{figs/benchthatshit} \tiny{Note: Point estimates for annual alpha returns with .90 confidence intervals for different investor groups compiled from different studies. Eggmueller refers to the current study. The last estimate refers to our replication of the Ziobrowksi et al. approach using our data for senators only.} \end{figure} \clearpage \begin{figure}[!hbt] \caption{\label{fig:8exp} Portfolio Weights as a Function of Member-Firm Connections} \centering \includegraphics[scale=.7]{figs/8exp.pdf} \end{figure} \clearpage \begin{landscape} \begin{figure}[!hbt] \caption{\label{fig:cumrets} Cumulative Monthly Return for Politically Connected and Unconnected Stocks} %\centering \includegraphics[scale=.7]{figs/allconnect} \tiny{Note: Cumulative monthly return is shown for a \$100 dollar position in each mimicking portfolio beginning in January 2004. The connected (unconnected) portfolios mimics the investments of all members of Congress in connected (unconnected) stocks. All portfolios shown are average member portfolios. The market portfolio is the value weighted market portfolio from NYSE, AMEX, and NASDAQ as computed by the CRSP. The connections are defined as follows. \emph{In District} connected firms are connected to a Member since they are located in a Member's home district. \emph{Lobbying (any)} connected firms are connected to a Member since they lobbied a Committee on which the Member served. \emph{Contributions (any)} connected firms are connected to a Member since they provided her with campaign contributions.} \end{figure} \end{landscape} \begin{figure}[!hbt] \caption{\label{fig:localpremium} Distribution of Member Specific Returns on Locally Connected and Unconnected Comapnies} \centering \includegraphics[scale=.5]{figs/local_premium.pdf} \tiny{Note: Box plots show the distribution of member specific monthly alpha estimates from a 4-factor Carhart model and a CAPM respectively for locally connected and unconnected companies as well as a zero cost portfolio that hold long the connected stocks and sells short the unconnected stocks. A company is locally connected if it is headquartered in a member's district.} \end{figure} \newpage \section*{Appendix A} \begin{figure}[!hbt] Figure A.1. Cumulative Abnormal Returns for Aggregate Congressional Portfolio and the Average Congressional Member\\ %\centering \includegraphics[scale=.45]{figs/all_members_ziob_mapping_holding_returns_tw} \includegraphics[scale=.45]{figs/all_members_ziob_mapping_holding_returnshed_tw} \includegraphics[scale=.45]{figs/all_members_ziob_mapping_holding_returns_ew} \includegraphics[scale=.45]{figs/all_members_ziob_mapping_holding_returnshed_ew} \tiny{Note: Cumulative Abnormal Returns (CARs) for aggregate congressional portfolio and the average congressional member. The aggregate portfolio mimics the aggregate trades of all members of Congress (value-weighted). The average congressional member portfolio mimics the trades of the average Member of Congress (equal Member weighted). The CARs are normalized such that the value is zero on the trading day zero. The left figures show the CARs for the buy and and sell portfolios. The right figures show the CARs for a zero cost portfolio that holds the portfolio of bought stocks and sells short the portfolio of sold stocks.} \end{figure} \clearpage \section*{Appendix B} \begin{table}[hbt!] \begin{center}\small \begin{tabular}{l|ccccc} \multicolumn{6}{c}{Table B.1: Portfolio Weights as a Function of Member-Firm Connections (Conditional on Holding)}\\ \hline Model & (1) & (2) & (3) & (4) & (5) \\ \hline Dependent Variable: & \multicolumn{ 5}{c}{Portfolio Weight} \\ Mean: & \multicolumn{ 5}{c}{298.52} \\ \hline In District & 329.36 & 200.17 & 327.40 & 321.19 & 266.31 \\ & (94.84) & (94.41) & (94.75) & (93.44) & (93.48) \\ Lobbying (Any) & 10.19 & 16.40 & & & \\ & (17.57) & (18.12) & & & \\ Contributions (Any) & 50.69 & 100.69 & & & \\ & (23.71) & (59.68) & & & \\ In District \& Lobbying (Any) & & 314.71 & & & \\ & & (239.70) & & & \\ In District \& Contributions (Any) & & 461.68 & & & \\ & & (230.98) & & & \\ Lobbying (Any) \& Contributions (Any) & & 50.19 & & & \\ & & (29.06) & & & \\ In District \& Contributions(Any) \& Lobbying (Any) & & 591.64 & & & \\ & & (232.57) & & & \\ Lobbying ($>$ p50) & & & -0.28 & & \\ & & & (21.72) & & \\ Contributions ($>$ p50) & & & 59.64 & & \\ & & & (32.79) & & \\ Lobbying Strength & & & & 0.01 & 0.01 \\ & & & & (0.03) & (0.03) \\ Contribution Strength & & & & 0.03 & 0.03 \\ & & & & (0.02) & (0.02) \\ Lobbying Strength $\cdot$ In District & & & & & 0.02 \\ & & & & & (0.13) \\ Contribution Strength $\cdot$ In District & & & & & 0.06 \\ & & & & & (0.08) \\ \hline Members Fixed Effects & x & x & x & x & x \\ Firms Fixed Effects & x & x & x & x & x \\ \hline N & \multicolumn{ 5}{c}{ 15,139} \\ % N & 15,211 & 15,211 & 15,211 & 15,211 & 15,211 \\ \hline \multicolumn{6}{p{5.5in}}{\tiny {\it Note:} Regression coefficients with standards errors (clustered by Members) in parenthesis. The dependent variable is the \emph{portfolio weight}, i.e. the share of holdings of a firm in a Member's portfolio (in basis points). Members' portfolios are computed as average holdings over the 2004-2008 period. \emph{In District} is a binary indicator for firms that are connected to a Member since they are located in a Member's district. \emph{Lobbying (any)} is a binary indicator for firms that are connected to a Member since they lobbied a Committee on which the Member served. \emph{Contributions (any)} is a binary indicator for firms that are connected to a Member since they provided her with campaign contributions. \emph{Lobbying ($>$ p50)} and \emph{Contributions ($>$ p50 )} are binary indicators for firms that provided more than the median lobbying or contribution amount for each Member. \emph{Lobbying Strength} and \emph{Contribution Strength} measure a firm's share of lobbying or contributions relative to each Member's total lobbying or contributions (in basis points). All regressions include a full set of Members and firms fixed effects (coefficients not shown here).} \end{tabular} \end{center} \end{table} \end{document}