THE QUARTERLY JOURNAL OF ECONOMICS - UCLA

THE

QUARTERLY JOURNALOF ECONOMICS

Vol CXX November 2005 Issue 4

A MEASURE OF MEDIA BIAS

TIM GROSECLOSE AND JEFFREY MILYO

We measure media bias by estimating ideological scores for several majormedia outlets To compute this we count the times that a particular media outletcites various think tanks and policy groups and then compare this with the timesthat members of Congress cite the same groups Our results show a strong liberalbias all of the news outlets we examine except Fox Newsrsquo Special Report and theWashington Times received scores to the left of the average member of CongressConsistent with claims made by conservative critics CBS Evening News and theNew York Times received scores far to the left of center The most centrist mediaoutlets were PBS NewsHour CNNrsquos Newsnight and ABCrsquos Good Morning Amer-ica among print outlets USA Today was closest to the center All of our findingsrefer strictly to news content that is we exclude editorials letters and the like

ldquoThe editors in Los Angeles killed the story They told Witcover that it didnrsquotlsquocome off rsquo and that it was an lsquoopinionrsquo story The solution was simple theytold him All he had to do was get other people to make the same points anddraw the same conclusions and then write the article in their wordsrdquo (empha-sis in original) Timothy Crouse Boys on the Bus [1973 p 116]

Do the major media outlets in the U S have a liberal biasFew questions evoke stronger opinions but so far the debate haslargely been one of anecdotes (ldquoHow can CBS News be balanced

We are grateful for the research assistance by Aviva Aminova Jose BustosAnya Byers Evan Davidson Kristina Doan Wesley Hussey David Lee PaulineMena Orges Obeqiri Byrne Offutt Matthew Patterson David Primo DarrylReeves Susie Rieniets Thomas Rosholt Michael Uy Diane Valos Michael Vis-conti Margaret Vo Rachel Ward and Andrew Wright Also we are grateful forcomments and suggestions by Matthew Baum Mark Crain Timothy GroelingFrances Groseclose Phillip Gussin James Hamilton Wesley Hussey Chap Law-son Steven Levitt Jeffrey Lewis Andrew Martin David Mayhew Jeffrey MinterMichael Munger David Primo Andrew Waddell Barry Weingast John Zallerand Jeffrey Zwiebel We also owe gratitude to the University of California at LosAngeles the University of Missouri Stanford University and the University ofChicago These universities paid our salaries funded our research assistants andpaid for services such as Lexis-Nexis which were necessary for our data collectionNo other organization or person helped to fund this research project

copy 2005 by the President and Fellows of Harvard College and the Massachusetts Institute ofTechnologyThe Quarterly Journal of Economics November 2005

1191

when it calls Steve Forbesrsquo tax plan lsquowackyrsquordquo) and untestedtheories (ldquoif the news industry is a competitive market then howcan media outlets be systematically biasedrdquo)

Few studies provide an objective measure of the slant ofnews and none has provided a way to link such a measure toideological measures of other political actors That is none of theexisting measures can say for example whether the New YorkTimes is more liberal than Senator Edward Kennedy or whetherFox News is more conservative than Senator Bill Frist We pro-vide such a measure Namely we compute an adjusted Americansfor Democratic Action (ADA) score for various news outlets in-cluding the New York Times the Washington Post USA Todaythe Drudge Report Fox Newsrsquo Special Report and all three net-worksrsquo nightly news shows

Our results show a strong liberal bias All of the news outletsexcept Fox Newsrsquo Special Report and the Washington Times re-ceived a score to the left of the average member of Congress Anda few outlets including the New York Times and CBS EveningNews were closer to the average Democrat in Congress than thecenter These findings refer strictly to the news stories of theoutlets That is we omitted editorials book reviews and letters tothe editor from our sample

To compute our measure we count the times that a mediaoutlet cites various think tanks and other policy groups1 Wecompare this with the times that members of Congress cite thesame think tanks in their speeches on the floor of the House andSenate By comparing the citation patterns we can construct anADA score for each media outlet

As a simplified example imagine that there were only twothink tanks and suppose that the New York Times cited the firstthink tank twice as often as the second Our method asks what isthe estimated ADA score of a member of Congress who exhibitsthe same frequency (21) in his or her speeches This is the scorethat our method would assign the New York Times

A feature of our method is that it does not require us to make asubjective assessment of how liberal or conservative a think tank isThat is for instance we do not need to read policy reports of the

1 Our sample includes policy groups that are not usually called think tankssuch as the NAACP NRA and Sierra Club To avoid using the more unwieldyphrase ldquothink tanks and other policy groupsrdquo we often use a shorthand versionldquothink tanksrdquo When we use the latter phrase we mean to include the othergroups such as the NAACP etc

1192 QUARTERLY JOURNAL OF ECONOMICS

think tank or analyze its position on various issues to determine itsideology Instead we simply observe the ADA scores of the membersof Congress who cite it This feature is important since an activecontroversy exists whether eg the Brookings Institution or theRAND Corporation is moderate left-wing or right-wing

I SOME PREVIOUS STUDIES OF MEDIA BIAS

Survey research has shown that an almost overwhelmingfraction of journalists are liberal For instance Povich [1996]reports that only 7 percent of all Washington correspondentsvoted for George H W Bush in 1992 compared with 37 percentof the American public2 Lichter Rothman and Lichter [1986]and Weaver and Wilhoit [1996] report similar findings for earlierelections More recently the New York Times reported that only8 percent of Washington correspondents thought George W Bushwould be a better president than John Kerry3 This compareswith 51 percent of all American voters David Brooks notes thatfor every journalist who contributed to George W Bushrsquos cam-paign another 93 contributed to Kerryrsquos campaign4

These statistics suggest that Washington correspondents asa group are more liberal than almost any congressional districtin the country For instance in the Ninth California districtwhich includes Berkeley 12 percent voted for Bush in 1992nearly double the rate of the correspondents In the Eighth Mas-sachusetts district which includes Cambridge 19 percent votedfor Bush approximately triple the rate of the correspondents5

Of course however just because a journalist has liberal orconservative views this does not mean that his or her reportingwill be slanted For instance as Jamieson [2000 p 188] notesldquoOne might hypothesize instead that reporters respond to thecues of those who pay their salaries and mask their own ideologi-

2 Eighty-nine percent of the Washington correspondents voted for Bill Clin-ton and two percent voted for Ross Perot

3 ldquoFinding Biases on the Busrdquo John Tierney New York Times August 12004 The article noted that journalists outside Washington were not as liberalTwenty-five percent of these journalists favored Bush over Kerry

4 ldquoRuling Class Warrdquo New York Times September 11 20045 Cambridge and Berkeleyrsquos preferences for Republican presidential candi-

dates have remained fairly constant since 1992 In the House district that con-tains Cambridge Bob Dole received 17 percent of the two-party vote in 1996 andGeorge W Bush received 19 percent in 2000 In the House district that containsBerkeley Bob Dole received 14 percent of the two-party vote and George W Bushreceived 13 percent

1193A MEASURE OF MEDIA BIAS

cal dispositions Another explanation would hold that norms ofjournalism including lsquoobjectivityrsquo and lsquobalancersquo blunt whateverbiases existrdquo Or as Crouse [1973] explains

It is an unwritten law of current political journalism that conservativeRepublican Presidential candidates usually receive gentler treatment fromthe press than do liberal Democrats Since most reporters are moderate orliberal Democrats themselves they try to offset their natural biases by goingout of their way to be fair to conservatives No candidate ever had a moreconsiderate press corps than Barry Goldwater in 1964 and four years laterthe campaign press gave every possible break to Richard Nixon Reporterssense a social barrier between themselves and most conservative candidatestheir relations are formal and meticulously polite But reporters tend toloosen up around liberal candidates and campaign staffs since they share thesame ideology they can joke with the staffers even needle them withoutbeing branded the ldquoenemyrdquo If a reporter has been trained in the traditionalldquoobjectiverdquo school of journalism this ideological and social closeness to thecandidate and the staff makes him feel guilty he begins to compensate themore he likes and agrees with the candidate personally the harder he judgeshim professionally Like a coach sizing up his own son in spring tryouts thereporter becomes doubly severe [pp 355ndash356]

However a strong form of the view that reporters offset orblunt their own ideological biases leads to a counterfactual im-plication Suppose that it is true that all reporters report objec-tively and their ideological views do not color their reporting Ifso then all news would have the same slant Moreover if onebelieves Crousersquos claim that reporters overcompensate in relationto their own ideology then a news outlet filled with conservativessuch as Fox News should have a more liberal slant than a newsoutlet filled with liberals such as the New York Times

Spatial models of firm location such as those by Hotelling[1929] or Mullainathan and Shleifer [2003] give theoretical rea-sons why the media should slant the news exactly as consumersdesire6 The idea is that if the media did not then an entrepre-neur could form a new outlet that does and he or she could earn

6 Some scholars claim that news outlets cater not to the desires of consum-ers but to the desires of advertisers Consequently since advertisers have pref-erences that are more pro-business or pro-free-market than the average con-sumer these scholars predict that news outlets will slant their coverage to theright of consumersrsquo preferences (eg see Parenti [1986] or Herman and Chomsky[1988]) While our work finds empirical problems with such predictions Sutter[2002] notes several theoretical problems Most important although an advertiserhas great incentive to pressure a news outlet to give favorable treatment to hisown product or his own business he has little incentive to pressure for favorabletreatment of business in general Although the total benefits of the latter type ofpressure may be large they are dispersed across a large number of businessesand the advertiser himself would receive only a tiny fraction of the benefits


greater-than-equilibrium profits possibly even driving the otheroutlets out of business This is a compelling argument and eventhe libertarian Cato Journal has published an article agreeingwith the view in this article Sutter [2001] notes that ldquoCharges ofa liberal bias essentially require the existence of a cartel [p 431]rdquo

However contrary to the prediction of the typical firm-loca-tion model we find a systematic liberal bias of the U S mediaThis is echoed by three other studiesmdashHamilton [2004] Lott andHassett [2004] and Sutter [2004] the only empirical studies ofmedia bias by economists of which we are aware

Although his primary focus is not on media bias in onesection of his book Hamilton [2004] analyzes Pew Center surveysof media bias The surveys showmdashunsurprisinglymdashthat conser-vatives tend to believe that there is a liberal bias in the mediawhile liberals tend to believe there is a conservative bias Whilemany would simply conclude that this is only evidence that ldquobiasis in the eyes of the beholderrdquo Hamilton makes the astute pointthat that individuals are more likely to perceive bias the furtherthe slant of the news is from their own position Since the samesurveys also show that conservatives tend to see a bias more thanliberals do this is evidence of a liberal bias

Lott and Hassett [2004] propose an innovative test for mediabias They record whether the headlines of various economic newsstories are positive or negative For instance on the day that theDepartment of Commerce reports that GDP grows by a largedegree a newspaper could instead report ldquoGDP Growth Less thanExpectedrdquo Lott and Hassett control for the actual economic fig-ures reported by the Department of Commerce and they includean independent variable that indicates the political party of thepresident Of the ten major newspapers that they examine theyfind that nine are more likely to report a negative headline if thepresident is Republican7

7 One of the most novel features of the Lott-Hassett paper is that to defineunbiased it constructs a baseline that can vary with exogenous factors In con-trast some studies define unbiased simply as some sort of version of ldquopresentingboth sides of the storyrdquo To see why the latter notion is inappropriate suppose thata newspaper devoted just as many stories describing the economy under PresidentClinton as good as it did describing the economy as bad By the latter notion thisnewspaper is unbiased However by Lott and Hassettrsquos notion the newspaper isunbiased only if the economy under Clinton was average If instead it was betterthan average then Lott and Hassett (as many would recognize as appropriateincluding us) would judge the newspaper to have a conservative bias Like Lottand Hassett our notion of bias also varies with exogenous factors For instancesuppose that after a series of events liberal (conservative) think tanks gain more


Sutter [2004] collects data on the geographic locations ofreaders of Time Newsweek and US News and World Report Heshows that as a region becomes more liberal (as indicated by itsvote share for President Clinton) its consumption of the threemajor national news magazines increases With a clever andrigorous theoretical model he shows that under some reasonableassumptions this empirical finding implies that the U S news-magazine industry taken as a whole is biased to the left

Notwithstanding these studies it is easy to find quotes fromprominent journalists and academics who claim that there is nosystematic liberal bias among the media in the United Statessome even claiming that there is a conservative bias The follow-ing are some examples

Our greatest accomplishment as a profession is the development since WorldWar II of a news reporting craft that is truly non-partisan and non-ideologicaland that strives to be independent of undue commercial or governmental influ-ence It is that legacy we must protect with our diligent stewardship To doso means we must be aware of the energetic effort that is now underway toconvince our readers that we are ideologues It is an exercise of in disinforma-tion of alarming proportions This attempt to convince the audience of theworldrsquos most ideology-free newspapers that theyrsquore being subjected to agenda-driven news reflecting a liberal bias I donrsquot believe our viewers and readers willbe in the long-run misled by those who advocate biased journalism8

when it comes to free publicity some of the major broadcast mediaare simply biased in favor of the Republicans while the rest tend to blurdifferences between the parties But thatrsquos the way it is Democrats shouldcomplain as loudly about the real conservative bias of the media as theRepublicans complain about its entirely mythical bias 9

The mainstream media does not have a liberal bias ABC CBSNBC CNN the New York Times The Washington Post Time Newsweek andthe restmdashat least try to be fair10

respect and credibility (say because they were better at predicting those events)which causes moderates in Congress to cite them more frequently By our notionfor a news outlet to remain unbiased it also must cite the liberal (conservative)think tanks more frequently The only other media-bias study of which we areaware that also constructs a baseline that controls for exogenous events isGroeling and Kernellrsquos [1998] study of presidential approval These researchersexamine the extent to which media outlets report increases and decreases in thepresidentrsquos approval while controlling for the actual increases and decreases inapproval (whether reported by the media or not) The focus of the paper howeveris on whether news outlets have a bias toward reporting good or bad news not onany liberal or conservative bias

8 New York Times Executive Editor Howell Raines accepting the ldquoGeorgeBeveridge Editor of the Year Awardrdquo at a National Press Foundation dinnershown live on C-SPAN2 February 20 2003

9 Paul Krugman ldquoInto the Wildernessrdquo New York Times November 8 200210 Al Franken [2003 p 3] Lies and the Lying Liars Who Tell Them A Fair

and Balanced Look at the Right


Irsquom going out telling the story that I think is the biggest story of ourtime how the right-wing media has become a partisan propaganda arm ofthe Republican National Committee We have an ideological press thatrsquosinterested in the election of Republicans and a mainstream press thatrsquosinterested in the bottom line Therefore we donrsquot have a vigilant indepen-dent press whose interest is the American people11

II DATA

The web site wwwwheretodoresearchcom lists 200 of themost prominent think tanks and policy groups in the UnitedStates Using the official web site of Congress httpthomaslocgov we and our research assistants searched the Congres-sional Record for instances where a member of Congress cited oneof these think tanks

We also recorded the average adjusted ADA score of themember who cited the think tank We use adjusted scores con-structed by Groseclose Levitt and Snyder [1999] because weneed the scores to be comparable across time and chambers12

Groseclose Levitt and Snyder use the 1980 House scale as theirbase year and chamber It is convenient for us to choose a scalethat gives centrist members of Congress a score of about 50 Forthis reason we converted scores to the 1999 House scale13

Along with direct quotes of think tanks we sometimes in-cluded sentences that were not direct quotes For instance manyof the citations were cases where a member of Congress notedldquoThis bill is supported by think tank Xrdquo Also members of Con-gress sometimes insert printed material into the CongressionalRecord such as a letter a newspaper article or a report If athink tank was cited in such material or if a think tank member

11 Bill Moyers quoted in ldquoBill Moyers Retiring from TV Journalismrdquo Fra-zier Moore Associated Press Online December 9 2004

12 Groseclose Levitt and Snyder [1999] argue that the underlying scales ofinterest group scores such as those compiled by the Americans for DemocraticAction can shift and stretch across years or across chambers This happensbecause the roll call votes that are used to construct the scores are not constantacross time nor across chambers They construct an index that allows one toconvert ADA scores to a common scale so that they can be compared across timeand chambers They call such scores adjusted ADA scores

13 Importantly we apply this conversion to congressional scores as well asmedia scores Since our method can only make relative assessments of the ideol-ogy of media outlets (eg how they compare with members of Congress or theaverage American voter) this transformation is benign Just as the averagetemperature in Boston is colder than the average temperature in Baltimoreregardless if one uses a Celsius scale or Fahrenheit scale all conclusions we drawin this paper are unaffected by the choice to use the 1999 House scale or the 1980House scale


wrote the material we treated it as if the member of Congresshad read the material in his or her speech

We did the same exercise for stories that media outlets re-port except with media outlets we did not record an ADA scoreInstead our method estimates such a score

Sometimes a legislator or journalist noted an action that athink tank had takenmdasheg that it raised a certain amount ofmoney initiated a boycott filed a lawsuit elected new officers orheld its annual convention We did not record such cases in ourdata set However sometimes in the process of describing suchactions the journalist or legislator would quote a member of thethink tank and the quote revealed the think tankrsquos views onnational policy or the quote stated a fact that is relevant tonational policy If so we would record that quote in our data setFor instance suppose that a reporter noted ldquoThe NAACP hasasked its members to boycott businesses in the state of SouthCarolina lsquoWe are initiating this boycott because we believe thatit is racist to fly the Confederate Flag on the state capitolrsquo aleader of the group notedrdquo In this instance we would count thesecond sentence that the reporter wrote but not the first

Also we omitted the instances where the member of Con-gress or journalist only cited the think tank so he or she couldcriticize it or explain why it was wrong About 5 percent of thecongressional citations and about 1 percent of the media citationsfell into this category

In the same spirit we omitted cases where a journalist orlegislator gave an ideological label to a think tank (eg ldquoEven theconservative Heritage Foundation favors this billrdquo) The idea isthat we only wanted cases where the legislator or journalist citedthe think tank as if it were a disinterested expert on the topic athand About 2 percent of the congressional citations and about 5percent of the media citations involved an ideological label14

14 In the Appendix we report the results when we do include citations thatinclude an ideological label When we include these data this does not cause asubstantial leftward or rightward movement in media scoresmdashthe average mediascore decreased by approximately 05 points ie it makes the media appearslightly more conservative The greater effect was to cause media outlets toappear more centrist For instance the New York Times and CBS Evening Newstended to give ideological labels to conservative think tanks more often than theydid to liberal think tanks As a consequence when we include the labeled obser-vations their scores respectively decreased (ie became more conservative) by38 and 16 points Meanwhile Fox Newsrsquo Special Report tended to do the oppositeWhen we included labeled observations its score increased (ie became moreliberal) by 18 points We think that such an asymmetric treatment of think tanks


For the congressional data we coded all citations that oc-curred during the period January 1 1993 to December 31 2002This covered the 103rd through 107th Congresses We used theperiod 1993 to 1999 to calculate the average adjusted ADA scorefor members of Congress15

As noted earlier our media data do not include editorialsletters to the editor or book reviews That is all of our resultsrefer only to the bias of the news of media There are severalreasons why we do not include editorials The primary one is thatthere is little controversy over the slant of editorial pages egfew would disagree that Wall Street Journal editorials are con-servative while New York Times editorials are liberal Howeverthere is a very large controversy about the slant of the news ofvarious media outlets A second reason involves the effect (if any)that the media have on individualsrsquo political views It is reason-able to believe that a biased outlet that pretends to be centrist hasmore of an effect on readersrsquo or viewersrsquo beliefs than say aneditorial page that does not pretend to be centrist A third reasoninvolves difficulties in coding the data Editorial and opinionwriters much more than news writers are sometimes sarcasticwhen they quote members of think tanks If our coders do notcatch the sarcasm they record the citation as a favorable one

(ie to give labels more often to one side) is itself a form of media bias This is whywe base our main conclusions on the nonlabeled data which accounts for this formof bias

15 Groseclose Levitt and Snyder [1999] have not computed adjusted scoresfor years after 1999 One consequence of this is that members who first enteredCongress in 2001 do not have adjusted scores Consequently we omitted theseobservations from our sample This omission causes little harm if any to ourestimation procedure First the citations of the new members comprised less thanone-half of 1 percent our sample Second the ideologies of the new members werefairly representative of the old members Third even if the new members were notrepresentative this fact alone would not cause a bias in our method To see thissuppose that these omitted members were disproportionately extreme liberals Toestimate ADA scores for a media outlet we need estimates of the citation behaviorof a range of members with ideologies near the ideology of the media outlet If wehad omitted some extreme liberal members of Congress this does not bias ourestimate of the citation pattern of the typical liberal it only makes it less precisesince we have less data for these members If on the other hand new membersbehaved differently from old members who have the same adjusted ADA scorethen this could cause a bias For instance suppose that new members with a 70adjusted ADA score tend to cite conservative think tanks more often than do oldmembers with a 70 score Then this would mean that Congressrsquos citation patternsare really more conservative than we have recorded This means the mediarsquoscitation patterns are really more liberal (relative to Congress) than they appear inour data set which would mean that the media is really more liberal than ourestimates indicate However we have no evidence to believe this (or the opposite)is the case And even if it were because the new members are such a small portionof the sample any bias should be small


This biases the results toward making the editorials appear morecentrist than they really are

In Table I we list the 50 groups from our list that weremost commonly cited by the media The first column lists theaverage ADA score of the legislator citing the think tankThese averages closely correspond to conventional wisdomabout the ideological positions of the groups For instance theHeritage Foundation and Christian Coalition with averagescores of 200 and 226 are near the conservative end theEconomic Policy Institute and the Childrenrsquos Defense Fund(803 and 820) are near the liberal end and the BrookingsInstitution and the World Wildlife Fund (533 and 504) are inthe middle of our mix of think tanks

While most of these averages closely agree with the conven-tional wisdom two cases are somewhat anomalous The first isthe ACLU The average score of legislators citing it was 498Later we shall provide reasons why it makes sense to define thepolitical center at 501 This suggests that the ACLU if anythingis a right-leaning organization The reason the ACLU has such alow score is that it opposed the McCain-Feingold Campaign Fi-nance bill and conservatives in Congress cited this often In factslightly more than one-eighth of all ACLU citations in Congresswere due to one person alone Mitch McConnell (R-KY) perhapsthe chief critic of McCain-Feingold If we omit McConnellrsquos cita-tions the ACLUrsquos average score increases to 559 Because of thisanomaly in the Appendix we report the results when we repeatall of our analyses but omit the ACLU data

The second apparent anomaly is the RAND Corporationwhich has a fairly liberal average score 604 We mentionedthis finding to some employees of RAND who told us they werenot surprised While RAND strives to be middle-of-the-roadideologically the more conservative scholars at RAND tend towork on military studies while the more liberal scholars tendto work on domestic studies Because the military studies aresometimes classified and often more technocratic than thedomestic studies the media and members of Congress tend tocite the domestic studies disproportionately As a consequenceRAND appears liberal when judged by these citations It isimportant to note that this factmdashthat the research at RAND ismore conservative than the numbers in Table I suggestmdashwillnot bias our results To see this think of RAND as two thinktanks RAND I the left-leaning think tank which produces the


TABLE ITHE 50 MOST-CITED THINK TANKS AND POLICY GROUPS

BY THE MEDIA IN OUR SAMPLE

Think tankpolicy group

Average scoreof legislatorswho cite the

group

Number ofcitations bylegislators

Number ofcitations by

media outlets

1 Brookings Institution 533 320 13922 American Civil Liberties Union 498 273 10733 NAACP 754 134 5594 Center for Strategic and International

Studies 463 79 4325 Amnesty International 574 394 4196 Council on Foreign Relations 602 45 4037 Sierra Club 687 376 3938 American Enterprise Institute 366 154 3829 RAND Corporation 604 352 350

10 National Rifle Association 459 143 33611 American Association of Retired Persons 660 411 33312 Carnegie Endowment for International

Peace 519 26 32813 Heritage Foundation 200 369 28814 Common Cause 690 222 28715 Center for Responsive Politics 669 75 26416 Consumer Federation of America 817 224 25617 Christian Coalition 226 141 22018 Cato Institute 363 224 19619 National Organization for Women 789 62 19520 Institute for International Economics 488 61 19421 Urban Institute 738 186 18722 Family Research Council 203 133 16023 Federation of American Scientists 675 36 13924 Economic Policy Institute 803 130 13825 Center on Budget and Policy Priorities 883 224 11526 National Right to Life Committee 216 81 10927 Electronic Privacy Information Center 574 19 10728 International Institute for Strategic

Studies 412 16 10429 World Wildlife Fund 504 130 10130 Cent for Strategic and Budgetary

Assessments 339 7 8931 National Abortion and Reproductive

Rights Action League 719 30 8832 Childrenrsquos Defense Fund 820 231 7833 Employee Benefit Research Institute 491 41 7834 Citizens Against Government Waste 363 367 7635 People for the American Way 761 63 7636 Environmental Defense Fund 669 137 7437 Economic Strategy Institute 719 26 7138 People for the Ethical Treatment of

Animals 734 5 7039 Americans for Tax Reform 187 211 6740 Citizens for Tax Justice 878 92 67


research that the media and members of Congress tend to citeand RAND II the conservative think tank which produces theresearch that they tend not to cite Our results exclude RANDII from the analysis This causes no more bias than excludingany other think tank that is rarely cited in Congress or themedia

The second and third columns respectively report the num-ber of congressional and media citations in our data These col-umns give some preliminary evidence that the media is liberalrelative to Congress To see this define as right-wing a think tankthat has an average score below 40 Next consider the ten most-cited think tanks by the media Only one right-wing think tankmakes this list the American Enterprise Institute In contrastconsider the ten most-cited think tanks by Congress (These arethe National Taxpayers Union AARP Amnesty InternationalSierra Club Heritage Foundation Citizens Against GovernmentWaste RAND Brookings NFIB and ACLU) Four of these areright-wing

For perspective in Table II we list the average adjusted ADAscore of some prominent members of Congress including somewell-known moderates These include the most conservativeDemocrat in our sample Nathan Deal (GA) and the most liberalRepublican in our sample Constance Morella (MD) AlthoughNathan Deal became a Republican in 1995 the score that we list

TABLE I(CONTINUED)



group



media outlets

41 National Federation of IndependentBusinesses 268 293 66

42 Hudson Institute 253 73 6443 National Taxpayers Union 343 566 6344 Stimson Center 636 26 6345 Center for Defense Information 790 28 6146 Handgun Control Inc 772 58 6147 Hoover Institution 365 35 6148 Nixon Center 217 6 6149 American Conservative Union 161 43 5650 Manhattan Institute 320 18 54


in the table is calculated only from his years as a Democrat16 Thetable also lists the average scores of the Republican and Demo-cratic parties17 To calculate average scores for each member wenote all of his or her scores for the seven-year period for which we

16 In fact for all members of Congress who switched parties we treatedthem as if they were two members one for when they were a Democrat and onefor when they were a Republican

17 The party averages reflect the midpoint of the House and Senate aver-ages Thus they give equal weight to each chamber not to each legislator sincethere are more House members than senators

TABLE IIAVERAGE ADJUSTED ADA SCORES OF LEGISLATORS

Legislator Average score

Maxine Waters (D-CA) 996Edward Kennedy (D-MA) 888John Kerry (D-MA) 876Average Democrat 843Tom Daschle (D-SD) 809Joe Lieberman (D-CT) 742Constance Morella (R-MD) 682Ernest Hollings (D-SC) 637John Breaux (D-LA) 595Christopher Shays (R-CT) 546Arlen Specter (R-PA) 513James Leach (R-IA) 503Howell Heflin (D-AL) 497Tom Campbell (R-CA) 486Sam Nunn (D-GA) 480Dave McCurdy (D-OK) 469Olympia Snowe (R-ME) 430Susan Collins (R-ME) 393Charlie Stenholm (D-TX) 361Rick Lazio (R-NY) 358Tom Ridge (R-PA) 267Nathan Deal (D-GA) 215Joe Scarborough (R-FL) 177Average Republican 161John McCain (R-AZ) 127Bill Frist (R-TN) 103Tom DeLay (R-TX) 47

The table lists average adjusted ADA scores The method for adjusting scores is described in GrosecloseLevitt and Snyder [1999] Scores listed are converted to the 1999 scale and are an average of each legislatorrsquosscores during the 1993ndash1999 period The one exception is Nathan Deal who switched parties in 1995 onlyhis score as a Democrat in 1994ndash1995 is included Deal is the most conservative Democrat over this timeperiod Constance Morella is the most liberal Republican


recorded adjusted scores (1993ndash1999) Then we calculated theaverage over this period

Because at times there is some subjectivity in coding ourdata when we hired our research assistants we asked for whomthey voted or would have voted if they were limited to choosing AlGore or George Bush We chose research assistants so that ap-proximately half our data was coded by Gore supporters and halfby Bush supporters

For each media outlet we selected an observation period thatwe estimated would yield at least 300 observations (citations)Because magazines television shows and radio shows produceless data per show or issue (eg a transcript for a 30-minutetelevision show contains only a small fraction of the sentencesthat are contained in a newspaper) with some outlets we beganwith the earliest date available in Lexis-Nexis We did this for (i)the three magazines that we analyze (ii) the five evening televi-sion news broadcasts that we analyze and (iii) the one radioprogram that we analyze18

III OUR DEFINITION OF BIAS

Before proceeding it is useful to clarify our definition of biasMost important the definition has nothing to do with the honestyor accuracy of the news outlet Instead our notion is more like ataste or preference For instance we estimate that the centristUnited States voter during the late 1990s had a left-right ideologyapproximately equal to that of Arlen Specter (R-PA) or Sam Nunn(D-GA) Meanwhile we estimate that the average New YorkTimes article is ideologically very similar to the average speech byJoe Lieberman (D-CT) Next since vote scores show Lieberman tobe more liberal than Specter or Nunn our method concludes thatthe New York Times has a liberal bias However in no way doesthis imply that the New York Times is inaccurate or dishonestmdashjust as the vote scores do not imply that Joe Lieberman is any lesshonest than Sam Nunn or Arlen Specter

In contrast other writers at least at times do define bias asa matter of accuracy or honesty We emphasize that our differ-ences with such writers are ones of semantics not substance Ifsay a reader insists that bias should refer to accuracy orhonesty then we urge him or her simply to substitute another

18 Table III in Section V lists the period of observation for each mediaoutlet


word wherever we write ldquobiasrdquo Perhaps ldquoslantrdquo is a goodalternative

However at the same time we argue that our notion of biasis meaningful and relevant and perhaps more meaningful andrelevant than the alternative notion The main reason we be-lieve is that only seldom do journalists make dishonest state-ments Cases such as Jayson Blair Stephen Glass or the falsifiedmemo at CBS are rare they make headlines when they do occurand much of the time they are orthogonal to any political bias

Instead for every sin of commission such as those by Glassor Blair we believe that there are hundreds and maybe thou-sands of sins of omissionmdashcases where a journalist chose facts orstories that only one side of the political spectrum is likely tomention For instance in a story printed on March 1 2002 theNew York Times reported that (i) the IRS increased its audit rateon the ldquoworking poorrdquo (a phrase that the article defines as anytaxpayer who claimed an earned income tax credit) while (ii) theagency decreased its audit rate on taxpayers who earn more than$100000 and (iii) more than half of all IRS audits involve theworking poor The article also notes that (iv) ldquoThe roughly 5percent of taxpayers who make more than $100000 have thegreatest opportunities to shortchange the government becausethey receive most of the nonwage incomerdquo

Most would agree that the article contains only true andaccurate statements however most would also agree that thestatements are more likely to be made by a liberal than a conser-vative Indeed the centrist and right-leaning news outlets by ourmeasure (the Washington Times Fox Newsrsquo Special Report theNewshour with Jim Lehrer ABCrsquos Good Morning America andCNNrsquos Newsnight with Aaron Brown) failed to mention any ofthese facts Meanwhile three of the outlets on the left side of ourspectrum (CBS Evening News USA Today and the [news pagesof the] Wall Street Journal) did mention at least one of the facts

Likewise on the opposite side of the political spectrum thereare true and accurate facts that conservatives are more likely tostate than liberals For instance on March 28 2002 the Wash-ington Times the most conservative outlet by our measure re-ported that Congress earmarked $304000 to restore opera housesin Connecticut Michigan and Washington19 Meanwhile none of

19 We assert that this statement is more likely to be made by a conservativebecause it suggests that government spending is filled with wasteful projectsThis conservatives often argue is a reason that government should lower taxes


the other outlets in our sample mentioned this fact Moreover theWashington Times article failed to mention facts that a liberalwould be more likely to note For instance it did not mention thatthe $304000 comprises a very tiny portion of the federal budget

We also believe that our notion of bias is the one that is morecommonly adopted by other authors For instance Lott and Has-sett [2004] do not assert that one headline in their data set is false(eg ldquoGDP Rises 5 Percentrdquo) while another headline is true (egldquoGDP Growth Less Than Expectedrdquo) Rather the choice of head-lines is more a question of taste or perhaps fairness than aquestion of accuracy or honesty Also much of Goldbergrsquos [2002]and Altermanrsquos [2003] complaints about media bias are that somestories receive scant attention from the press not that the storiesreceive inaccurate attention For instance Goldberg notes howfew stories the media devote to the problems faced by children ofdual-career parents On the opposite side Alterman notes howfew stories the media devote to corporate fraud Our notion of biasalso seems closely aligned to the notion described by Bozell andBaker [1990 p 3] ldquoBut though bias in the media exists it israrely a conscious attempt to distort the news It stems from thefact that most members of the media elite have little contact withconservatives and make little effort to understand the conserva-tive viewpoint Their friends are liberals what they read andhear is written by liberalsrdquo20

Similar to the facts and stories that journalists report thecitations that they gather from experts are also very rarely dis-honest or inaccurate Many and perhaps most simply indicatethe side of an issue that the expert or his or her organizationfavors For instance on April 27 2002 the New York Timesreported that Congress passed a $100 billion farm subsidies billthat also gave vouchers to the elderly to buy fresh fruits andvegetables ldquoThis is a terrific outcomemdashone of the most importantpieces of social welfare legislation this yearrdquo said Stacy Dean ofthe Center on Budget and Policy Priorities her only quote in thearticle In another instance on May 19 2001 CBS Evening Newsdescribed President Bushrsquos call for expanding nuclear power Itquoted the Sierra Clubrsquos Daniel Becker ldquo[S]witching from coal tonuclear power is like giving up smoking and taking up crackrdquo

20 We were directed to this passage by Sutterrsquos [2001] article which adoptsnearly the same definition of bias as we do


Most would agree that these statements are more normative thanpositive that is they are more an indication of the authorrsquospreferences than a fact or prediction

Similarly another large fraction of cases involve the organi-zationrsquos views of politicians For instance on March 29 2002 theWashington Times reported that the National Taxpayersrsquo Union(NTU) gave Hillary Clinton a score of 3 percent on its annualrating of Congress The story noted that the score according tothe NTU was ldquothe worst score for a Senate freshman in their firstyear in office that the NTU has ever recordedrdquo

Finally many other citations refer to facts that are generallybeyond dispute However like the facts that reporters themselvesnote these facts are ones that conservatives and liberals are notequally likely to state For instance on March 5 1992 CBSEvening News reported a fact that liberals are more likely to notethan conservatives ldquoThe United States now has greater dispari-ties of income than virtually any Western European countryrdquosaid Robert Greenstein of the Center on Budget and Policy Pri-orities Meanwhile on May 30 2003 CNNrsquos Newsnight withAaron Brown noted a fact that conservatives are more likely tostate than liberals In a story about the FCCrsquos decision to weakenregulations about media ownership it quoted Adam Thierer ofthe Cato Institute ldquo[L]etrsquos start by stepping back and taking alook at the landscape of today versus say 10 15 25 30 yearsago And by almost every measure that you can go by you can seethat there is more diversity more competition more choice forconsumers and citizens in these marketplacesrdquo21

21 Like us Mullainathan and Shleifer [2003] define bias as an instancewhere a journalist fails to report a relevant fact rather than chooses to report afalse fact However unlike us Mullainathan and Shleifer define bias as a questionof accuracy not a taste or preference More specifically their model assumes thatwith any potential news story there are a finite number of facts that apply to thestory By their definition a journalist is unbiased only if he or she reports all thesefacts (However given that there may be an unwieldy number of facts that thejournalist could mention it also seems consistent with the spirit of their definitionthat if the journalist merely selects facts randomly from this set or if he or shechooses a representative sample then this would also qualify as unbiased) As anexample suppose that out of the entire universe of facts about free trade most ofthe facts imply that free trade is good However suppose that liberals andmoderates in Congress are convinced that it is bad and hence in their speechesthey state more facts about its problems Under Mullainathan and Shleiferrsquosdefinition to be unbiased a journalist must state more facts about the advantagesof free trademdashwhereas under our definition a journalist must state more factsabout the disadvantages of free trade Again we emphasize that our differenceson this point are ones of semantics Each notion of bias is meaningful andrelevant And if a reader insists that ldquobiasrdquo should refer to one notion instead ofthe other we suggest that he or she substitute a different word for the other


IV A SIMPLE STRUCTURAL MODEL

Define xi as the average adjusted ADA score of the ith mem-ber of Congress Given that the member cites a think tank weassume that the utility that he or she receives from citing the jth thinktank is

(1) aj bjxi eij

The parameter bj indicates the ideology of the think tank Notethat if xi is large (ie the legislator is liberal) then the legislatorreceives more utility from citing the think tank if bj is large Theparameter aj represents a sort of ldquovalencerdquo factor (as politicalscientists use the term) for the think tank It captures nonideo-logical factors that lead legislators and journalists to cite thethink tank Such factors may include such things as a reputationfor high-quality and objective research which may be orthogonalto any ideological leanings of the think tank

We assume that eij is distributed according to a Weibulldistribution As shown by McFadden [1974] (also see Judge et al[1985 pp 770ndash772]) this implies that the probability that mem-ber i selects the jth think tank is

notion such as ldquoslantrdquo Further we suggest that Mullainathan and Shleiferrsquosnotion is an ideal that a journalist perhaps should pursue before our notionNevertheless we suggest a weakness of Mullainathan and Shleiferrsquos notion it isvery inconvenient for empirical work and perhaps completely infeasible Namelyit would be nearly impossiblemdashand at best a very subjective exercisemdashfor aresearcher to try to determine all the facts that are relevant for a given newsstory Likewise it would be very difficult and maybe impossible for a journalistto determine this set of facts To see this consider just a portion of the facts thatmay be relevant to a news story the citations from experts There are hundredsand maybe thousands of think tanks not to mention hundreds of academicdepartments At what point does the journalist decide that a think tank oracademic department is so obscure that it does not need to be contacted for acitation Further most think tanks and academic departments house dozens ofmembers This means that an unbiased journalist would have to speak to a hugenumber of potential experts Moreover even if the journalist could contact all ofthese experts a further problem is how long to talk to them At what point doesthe journalist stop gathering information from one particular expert before he orshe is considered unbiased Even if a journalist only needs to contact a represen-tative sample of these experts a problem still exists over defining the relevantuniverse of experts Again when is an expert so obscure that he or she should notbe included in the universe A similar problem involves the journalistrsquos choice ofstories to pursue A news outlet can choose from a hugemdashand possibly infinitemdashnumber of news stories Although Mullainathan and Shleiferrsquos model focuses onlyon the bias for a given story a relevant source of bias is the journalistrsquos choice ofstories to cover It would be very difficult for a researcher to construct a universeof stories from which journalists choose to cover For instance within this uni-verse what proportion should involve the problems of dual-career parents Whatproportion should involve corporate fraud


(2) expaj bjxi k1

J

expak bkxi

where J is the total number of think tanks in our sampleNote that this probability term is no different from the one wesee in a multinomial logit (where the only independent vari-able is xi)

Define cm as the estimated adjusted ADA score of the mthmedia outlet Similar to the members of Congress we assumethat the utility that it receives from citing the jth think tank is

(3) aj bjcm emj

We assume that emj is distributed according to a Weibull distri-bution This implies that the probability that media outlet mselects the jth think tank is

(4) expaj bjcm k1

J

expak bkcm

Although this term is similar to the term that appears in amultinomial logit we cannot use multinomial logit to estimatethe parameters The problem is that cm a parameter that weestimate appears where normally we would have an independentvariable Instead we construct a likelihood function from (2) and(4) and we use the ldquonlmrdquo (nonlinear maximization) command inR to obtain estimates of each aj bj and cm

Similar to a multinomial logit it is impossible to identifyeach aj and bj Consequently we arbitrarily choose one thinktank and set its values of aj and bj to zero It is convenient tochoose a think tank that is cited frequently Also to make mostestimates of the bjrsquos positive it is convenient to choose a thinktank that is conservative Consequently we chose the HeritageFoundation It is easy to prove that this choice does not affect ourestimates of cm That is if we had chosen a different think tankthen all estimates of cm would be unchanged

This identification problem is not just a technical point italso has an important substantive implication Our method doesnot need to determine any sort of assessment of the absoluteideological position of a think tank It only needs to assess therelative position In fact our method cannot assess absolute po-sitions As a concrete example consider the estimated bjrsquos for AEI


and the Brookings Institution These values are 026 and 038The fact that the Brookings estimate is larger means that Brook-ings is more liberal than AEI (More precisely it means that as alegislator or journalist becomes more liberal he or she prefersmore and more to cite Brookings than AEI) These estimates areconsistent with the claim that AEI is conservative (in an absolutesense) while Brookings is liberal But they are also consistentwith a claim eg that AEI is moderate-left while Brookings isfar-left (or also the possibility that AEI is far-right while Brook-ings is moderate-right) This is related to the fact that our modelcannot fully identify the bjrsquos that is we could add the sameconstant to each and the value of the likelihood function (andtherefore the estimates of the cmrsquos) would remain unchanged

One difficulty that arose in the estimation process is that ittakes an unwieldy amount of time to estimate all of the parame-ters If we had computed a separate aj and bj for each think tankin our sample then we estimate that our model would take overtwo weeks to converge and produce estimates22 Complicatingthis we compute estimates for approximately two dozen differentspecifications of our basic model (Most of these are to test restric-tions of parameters For example we run one specification wherethe New York Times and NPRrsquos Morning Edition are constrainedto have the same estimate of cm) Thus if we estimated the fullversion of the model for each specification our computer wouldtake approximately one year to produce all the estimates

Instead we collapsed data from many of the rarely citedthink tanks into six mega think tanks Specifically we estimateda separate aj and bj for the 44 think tanks that were most-citedby the media These comprised 856 percent of the total number ofmedia citations With the remaining think tanks we orderedthem left to right according to the average ADA score of thelegislators who cited them Let pmin and pmax be the minimumand maximum average scores for these think tanks To create themega think tanks we defined five cut points to separate themSpecifically we define cut point i as

(5) pi pmin i6pmax pmin

22 Originally we used Stata to try to compute estimates With this statisticalpackage we estimate that it would have taken eight weeks for our computer toconverge and produce estimates


In practice these five cut points were 2204 3610 5015 6421and 7827

The number of actual and mega think tanks to include (re-spectively 44 and 6) is a somewhat arbitrary choice We chose 50as the total number because we often used the mlogit procedurein Stata to compute seed values This procedure is limited to atmost 50 ldquochoicesrdquo which meant that we could estimate aj and brsquosfor at most 50 think tanks This still leaves an arbitrary choiceabout how many of the 50 think tanks should be actual thinktanks and how many should be mega think tanks We experi-mented with several different choices Some choices made themedia appear slightly more liberal than others We chose six asthe number of mega think tanks because it produced approxi-mately the average of the estimates In the Appendix we alsoreport results when instead we choose 2 3 4 5 7 or 8 as thenumber of mega think tanks

Our choice to use 50 as the total number of actual and mega thinktanks if anything appears to make the media appear more conserva-tive than they really are In the Appendix we report results wheninstead we chose 60 70 80 and 90 as the total number of actual andmega think tanks In general these choices cause the average estimateof cm to increase by approximately one or two points

V RESULTS

In Table III we list the estimates of cm the adjusted ADAscores for media outlets The ordering of the scores is largelyconsistent with conventional wisdom For instance the two mostconservative outlets are the Washington Times and Fox NewsrsquoSpecial Report two outlets that are often called conservative (egsee Alterman [2003]) Near the liberal end are CBS Evening Newsand the New York Times Again these are largely consistent withthe conventional wisdom For instance CBS Evening News wasthe target of a best-selling book by Goldberg [2002] a formerreporter who documents several instances of liberal bias at thenews show Further some previous scholarly work shows CBSEvening News to be the most liberal of the three network eveningnews shows Hamilton [2004] recorded the congressional roll callvotes that the Americans for Democratic Action chose for itsannual scorecard and he examined how often each network cov-ered the roll calls Between 1969 and 1998 CBS Evening News


consistently covered these roll calls more frequently than did theother two networks23

One surprise is the Wall Street Journal which we find as themost liberal of all twenty news outlets We should first remindreaders that this estimate (as well as all other newspaper esti-mates) refers only to the news of the Wall Street Journal weomitted all data that came from its editorial page If we included

23 However Hamilton also notes that CBS covered roll calls by the Ameri-can Conservative Union more frequently than the other two networks Neverthe-less one can compute differences in frequencies between roll calls from the ADAand ACU These differences show CBS to be more liberal than ABC and NBCThat is although all three networks covered ADA roll calls more frequently thanthey covered ACU roll calls CBS did this to a greater extent than the other twonetworks

TABLE IIIRESULTS OF MAXIMUM LIKELIHOOD ESTIMATION

Media outletPeriod of

observationEstimatedADA score

Standarderror

ABC Good Morning America 62797ndash 62603 561 32ABC World News Tonight 1194ndash 62603 610 17CBS Early Show 11199ndash 62603 666 40CBS Evening News 1190ndash 62603 737 16CNN NewsNight with Aaron Brown 11901ndash 2504 560 41Drudge Report 32602ndash 7104 604 31Fox Newsrsquo Special Report with Brit Hume 6198ndash 62603 397 19Los Angeles Times 62802ndash122902 700 22NBC Nightly News 1197ndash 62603 616 18NBC Today Show 62797ndash 62603 640 25New York Times 7101ndash 5102 737 16Newshour with Jim Lehrer 112999ndash 62603 558 23Newsweek 62795ndash 62603 663 18NPR Morning Edition 1192ndash 62603 663 10Time Magazine 8601ndash 62603 654 48US News and World Report 62795ndash 62603 658 18USA Today 1102ndash 9102 634 27Wall Street Journal 1102ndash 5102 851 39Washington Post 1102ndash 5102 666 25Washington Times 1102ndash 5102 354 27Average 626

The table gives our estimates of adjusted ADA scores for media outlets converted to the 1999 Housescale As a comparison 5006 is our estimate of the average American voter this is based upon the averageadjusted ADA scores of the House and Senate from 1995 to 1999 (Senate scores were population-weighted andincluded two extreme liberal phantom Senators for Washington DC) The average score for Republicans was161 and for Democrats 843 All data for the news outlets came from news content only (ie editorialsletters and book reviews were omitted)


data from the editorial page surely it would appear moreconservative

Second some anecdotal evidence agrees with the result Forinstance Irvine and Kincaid [2001] note that ldquoThe Journal hashad a long-standing separation between its conservative editorialpages and its liberal news pagesrdquo Sperry [2002] notes that thenews division of the Journal sometimes calls the editorial divisionldquoNazisrdquo ldquoFact isrdquo Sperry writes ldquothe Journalrsquos news and editorialdepartments are as politically polarized as North and SouthKoreardquo24

Third a recent poll from the Pew Research Center indicatesthat a greater percentage of Democrats 29 percent say they trustthe Journal than do Republicans 23 percent Importantly thequestion did not say ldquothe news division at the Wall Street Jour-nalrdquo If it had Democrats surely would have said they trusted theJournal even more and Republicans even less25

Finally and perhaps most important a scholarly studymdashbyLott and Hassett [2004]mdashgives evidence that is consistent withour result As far as we are aware this is the only other studythat examines the political bias of the news pages of the WallStreet Journal Of the ten major newspapers that it examines itestimates the Wall Street Journal as the second-most liberal26

Only Newsday is more liberal and the Journal is substantiallymore liberal than the New York Times Washington Post LosAngeles Times and USA Today

Another somewhat surprising result is our estimate of NPRrsquosMorning Edition Conservatives frequently list NPR as an egre-gious example of a liberal news outlet27 However by our esti-

24 Other anecdotes that Sperry documents are (i) a reporter Kent MacDou-gall who after leaving the Journal bragged that he used the ldquobourgeois pressrdquo tohelp ldquopopularize radical ideas with lengthy sympathetic profiles of Marxist econ-omistsrdquo (ii) another Journal reporter who after calling the Houston-based MMARGroup shady and reckless caused the Journal to lose a libel suit after jurorslearned that she misquoted several of her sources (iii) a third Journal reporterSusan Faludi (the famous feminist) characterized Safeway as practicing ldquorobberbaronrdquo style management practices

25 See httppeople-pressorgreportsdisplayphp3ReportID215 for a de-scription of the survey and its data See also Kurtz [2004] for a summary of thestudy

26 This comes from the estimates for the ldquoRepublicanrdquo coefficient that theylist in their Table 7 These estimates indicate the extent to which a newspaper ismore likely to use a negative headline for economic news when the president isRepublican

27 Sometimes even liberals consider NPR left-wing As Woodward notes inThe Agenda [1994 p 114] ldquo[Paul] Begala was steaming To him [OMB DirectorAlice] Rivlin symbolized all that was wrong with Clintonrsquos new team of Washing-


mate the outlet hardly differs from the average mainstream newsoutlet For instance its score is approximately equal to those ofTime Newsweek and US News and World Report and its scoreis slightly less than the Washington Postrsquos Further our estimateplaces it well to the right of the New York Times and also to theright of the average speech by Joe Lieberman These differencesare statistically significant28 We mentioned this finding to TerryAnderson an academic economist and Executive Director of thePolitical Economy Research Center which is among the list of thinktanks in our sample (The average score of legislators citing PERCwas 399 which places it as a moderate-right think tank approxi-mately as conservative as RAND is liberal) Anderson told usldquoWhen NPR interviewed us they were nothing but fair I think theconventional wisdom has overstated any liberal bias at NPRrdquo OurNPR estimate is also consistent with Hamiltonrsquos [2004 p 108]research on audience ideology of news outlets Hamilton finds thatthe average NPR listener holds approximately the same ideology asthe average network news viewer or the average viewer of morningnews shows such as Today or Good Morning America Indeed of theoutlets that he examines in this section of his book by this measureNPR is the ninth most liberal out of eighteen

Another result which appears anomalous is not so anomalousupon further examination This is the estimate for the DrudgeReport which at 604 places it approximately in the middle of ourmix of media outlets and approximately as liberal as a typicalSouthern Democrat such as John Breaux (D-LA) We should em-phasize that this estimate reflects both the news flashes that MattDrudge reports and the news stories to which his site links on otherweb sites In fact of the entire 311 think-tank citations we found inthe Drudge Report only five came from reports written by Matt

ton hands and represented the Volvo-driving National Public Radio-listeningwine-drinking liberalism that he felt had crippled the Democratic Party fordecadesrdquo

28 To test that NPR is to the right of Joe Lieberman we assume that wehave measured the ideological position of Lieberman without error Using thevalues in Table II and III the t-test for this hypothesis is t (742 663)10 79 This is significant at greater than 999 percent levels of confidence To testthat NPR is to the right of the New York Times we use a likelihood ratio test Thevalue of the log likelihood function when NPR and the New York Times areconstrained to have the same score is 7861664 The unconstrained value of thelog likelihood function is 7860935 The relevant value of the likelihood ratiotest is 2(7861664ndash7860935) This is distributed according to the Chi-Squaredistribution with one degree of freedom At confidence levels greater than 999percent we can reject the hypothesis that the two outlets have the same score


Drudge Thus for all intents and purposes our estimate for theDrudge Report refers only to the articles to which the Report links onother web sites Although the conventional wisdom often assertsthat the Drudge Report is relatively conservative we believe thatthe conventional wisdom would also assert thatmdashif confined only tothe news stories to which the Report links on other web sitesmdashthisset would have a slant approximately equal to the average slant ofall media outlets since after all it is comprised of stories from abroad mix of such outlets29

VI DIGRESSION DEFINING THE ldquoCENTERrdquo

While the main goal of our research is to provide a measurethat allows us to compare the ideological positions of mediaoutlets to political actors a separate goal is to express whethera news outlet is left or right of center To do the latter we mustdefine center This is a little more arbitrary than the firstexercise For instance the results of the previous section showthat the average New York Times article is approximately asliberal as the average Joe Lieberman (D-CT) speech WhileLieberman is left of center in the United States Senate manywould claim that compared with all persons in the entireworld he is centrist or even right-leaning And if the latter isonersquos criterion then nearly all of the media outlets that weexamine are right of center

However we are more interested in defining centrist byUnited States views rather than world views or say Europeanviews One reason is that the primary consumers for the twentynews outlets that we examine are in the United States If forexample we wish to test economic theories about whether UnitedStates news producers are adequately catering to the demands oftheir consumers then United States consumers are the ones onwhich we should focus A second reason is that the popular debateon media bias has focused on United States views not world

29 Of the reports written by Matt Drudge he cited the Brookings Institutiontwice (actually once but he listed the article for two days in a row) the ACLU onceTaxpayers for Common Sense once and Amnesty International once On June 222004 the Drudge Report listed a link to an earlier version of our paper Although thatversion mentioned many think tanks only one case would count as a citation This isthe paraphrased quote from RAND members stating that the media tends to cite itsmilitary studies less than its domestic studies (The above quote from PERC was notin the earlier version although it would also count as a citation) At any rate weinstructed our research assistants not to search our own paper for citations


views For instance in Goldbergrsquos [2002] insider account of CBSNews he only claims that CBS is more liberal than the averageAmerican not the average European or world citizen

Given this one definition of centrist is simply to use the meanor median ideological score of the United States House or SenateWe focus on mean scores since the median tends to be unstable30

This is due to the bimodal nature that ADA scores have followedin recent years For instance in 1999 only three senators out ofa total of 100 received a score between 33 and 67 In contrast 33senators would have received scores in this range if the scoreshad been distributed uniformly and the number would be evenlarger if scores had been distributed unimodally31

We are most interested in comparing news outlets to thecentrist voter who for a number of reasons might not have thesame ideology as the centrist member of Congress For instancebecause Washington DC is not represented in Congress andbecause DC residents tend to be more liberal than the rest of thecountry the centrist member of Congress should tend to be moreconservative than the centrist voter

Another problem which applies only to the Senate involvesthe fact that voters from small states are overrepresented Sincein recent years small states have tended to vote more conserva-tively than large states this would cause the centrist member ofthe Senate to be more conservative than the centrist voter

A third reason which applies only to the House is thatgerrymandered districts can skew the relationship between acentrist voter and a centrist member of the House For instancealthough the total votes for Al Gore and George W Bush favoredGore slightly the median House district slightly favored BushSpecifically if we exclude the District of Columbia (since it doesnot have a House member) Al Gore received 5019 percent of thetwo-party vote Yet in the median House district (judging byGore-Bush vote percentages) Al Gore received only 4896 percentof the two-party vote (Twelve districts had percentages betweenthe median and mean percentages) The fact that the latter num-ber is smaller than the former number means that House dis-

30 Nevertheless we still report how our results change if instead we usemedian statistics See footnotes 34 and 35

31 The year 1999 was somewhat but not very atypical During the rest ofthe 1990s on average 176 senators received scores between 33 and 67 approxi-mately half as many as would be expected if scores were distributed uniformlySee httpwwwadactionorgvotingrecordshtm for ADA scores of senators andHouse members


tricts are drawn to favor Republicans slightly Similar resultsoccurred in the 1996 election Bill Clinton received 5466 percentof the two-party vote Yet in the median House district he re-ceived 5354 percent

It is possible to overcome each of these problems to estimatean ADA score of the centrist voter in the United States First toaccount for the DC bias we can add phantom DC legislators tothe House and Senate Of course we necessarily do not know theADA scores of such legislators However it is reasonable to be-lieve that they would be fairly liberal since DC residents tend tovote overwhelmingly Democratic in presidential elections (Theyvoted 905 percent for Gore in 2000 and 906 percent for Kerry in2004) For each year we gave the phantom DC House memberand senators the highest respective House and Senate scores thatoccurred that year Of course actual DC legislators might not bequite so liberal However one of our main conclusions is that themedia are liberal compared with U S voters Consequently it isbetter to err on the side of making voters appear more liberalthan they really are than the opposite32

The second problem the small-state bias in the Senate canbe overcome simply by weighting each senatorrsquos score by thepopulation of his or her state The third problem gerrymandereddistricts in the House is overcome simply by the fact that we usemean scores instead of the median33

In Figure I we list the mean House and Senate scores overthe period 1947ndash1999 when we use this methodology (ie includ-

32 Another possible bias involves the fact that DC has slightly fewer peoplethan the average House district Using 2000 population estimates (source Alma-nac of American Politics [2002 edition]) DC had 572000 residents while theaverage House district in the country had 646000 We treat DC as one districtwhereas a more appropriate analysis would treat DC as 572646 of a districtAgain this will bias our results in the opposite direction of our main conclusionsSpecifically this will cause media outlets to appear more conservative than theyreally are

33 To see this imagine a state with three districts each with the samesymmetrical distribution of voters (Thus the median voter in each district has anideology identical to the median voter of the state) Now suppose that a Republi-can state legislature redraws districts so that Democratic voters are transferredfrom districts 1 and 2 to district 3 Suppose that Republican voters are transferredin the opposite direction Necessarily the increase in Democratic voters in district3 is twice the average increase in Republican voters in districts 1 and 2 Nextsuppose that the expected ideological score of a representative is a linear functionof the fraction of Democratic voters in his or her district Then it will necessarilybe the case that the expected average ideological score of the representatives inthis hypothetical state will be identical to the expected average before redistrict-ing However the same will not be true of the median score It will be expected todecrease (ie to become more conservative)


ing phantom DC legislators and weighting senatorsrsquo scores bythe population of their state) The focus of our results is for theperiod 1995ndash1999 We chose 1999 as the end year simply becausethis is the last year for which Groseclose Levitt and Snyder[1999] computed adjusted ADA scores However any conclusionsthat we make for this period should also hold for the 2000ndash2004period since in the latter period the House and Senate hadalmost identical party ratios We chose 1995 as the beginningyear because it is the first year after the historic 1994 electionswhere Republicans gained 52 House seats and 8 Senate seatsThis year it is reasonable to believe marks the beginning of aseparate era of American politics As a consequence if onewanted to test hypotheses about the typical United States voterof say 1999 then the years 1998 1997 1996 and 1995 wouldalso provide helpful data However prior years would not

Over this period the mean score of the Senate (after in-cluding phantom DC senators and weighting by state popu-lation) varied between 4928 and 5087 The mean of thesemeans was 4994 The similar figure for the House was 5018After rounding we use the midpoint of these numbers 501 as

FIGURE IWeighted-Average ADA Scores for the House and Senate 1947ndash1999


our estimate of the adjusted ADA score of the centrist UnitedStates voter34

A counterview is that the 1994 elections did not mark a newera Instead as some might argue these elections were an anom-aly and the congresses of the decade or so before the 1994elections are a more appropriate representation of voter senti-ment of the late 1990s and early 2000s Although we do not agreewe think it is a useful straw man Consequently we construct analternative measure based on the congresses that served between1975 and 1994 We chose 1975 because this was the first year ofthe ldquoWatergate babiesrdquo in Congress As Figure I shows this yearproduced a large liberal shift in Congress This period 1975ndash1994 also happens to be the most liberal twenty-year period forthe entire era that the ADA has been recording vote scores

The average ADA score of senators during the 1975ndash1994period (after including phantom DC senators and weightingaccording to state population) was 5351 The similar figure forthe House was 5458 After rounding we use the midpoint of

34 A clever alternative measure suggested to us by David Mayhew is to usea regression-based framework to estimate the expected ADA score of a legislatorwhose district is perfectly representative of the entire United States In the 2000presidential election Gore won 5027 percent of the two-party vote (includingDC) Suppose that we could construct a hypothetical congressional district withan identical Gore-vote percentage It is reasonable to believe that the expectedadjusted ADA score of the legislator from such a district is a good measure of theideology of the centrist United States voter and this appropriately adjusts for anybiases due to gerrymandered districts exclusion of DC voters and the small-state biases in the Senate To estimate this we regressed (i) the 1999 adjustedADA scores of members of Congress on (ii) Gorersquos percentage of the two-party votein the legislatorrsquos district In this regression we included observations from theSenate as well as the House (Remember that adjusted scores are constructed sothat they are comparable across chambers) The results of the regression wereADA Score 4648 191 Gore Vote This implies 4953 as the expected ADAscore of a district in which the Gore vote was 5027 percent We repeated thisanalysis using instead adjusted ADA scores from 1998 1997 1996 and 1995 Inthe latter three years we used the Clinton share of the two-party vote and we usedClintonrsquos national share 5474 percent as the share of the representative districtThese years give the following respective estimates of the ADA score of thecentrist U S voter 4883 4899 4724 and 4741 The average of these fivemeasures is 4840 Since this number is 17 points less than the mean-basedmeasure of the centrist voter (501) if one believes that it is the more appropriatemeasure then our main conclusions (based on the mean-based measure) arebiased rightwardmdashthat is the more appropriate conclusion would assert that themedia are an additional 17 points to the left of the centrist voter

Yet another measure is based on median scores of the House and Senate Theaverage Senate median over the five years was 5819 while the average Housemedian was 4061 (Again both these figures include phantom DC legislatorsand the Senate score is weighted by state population) The midpoint is 494 whichis 07 points more conservative than our mean-based measure If one believes thatthis is the more appropriate measure of centrist then once again this impliesthat our media estimates are biased in the direction of making them moreconservative than they really are


these two scores to define 540 as the centrist United States voterduring 1975ndash199435

VII FURTHER RESULTS HOW CLOSE ARE MEDIA OUTLETS

TO THE CENTER

Next we compute the difference of a media outletrsquos scorefrom 501 to judge how centrist it is We list these results in TableIV Most striking is that all but two of the outlets we examine areleft of center Even more striking is that if we use the more liberaldefinition of center (540)mdashthe one constructed from congres-sional scores from 1975ndash1994mdashit is still the case that eighteen oftwenty outlets are left of center

35 If instead we use medians the figure is 549

TABLE IVRANKINGS BASED ON DISTANCE FROM CENTER

Rank Media outletEstimatedADA score

1 Newshour with Jim Lehrer 5582 CNN NewsNight with Aaron Brown 5603 ABC Good Morning America 5614 Drudge Report 6045 Fox Newsrsquo Special Report with Brit Hume 3976 ABC World News Tonight 6107 NBC Nightly News 6168 USA Today 6349 NBC Today Show 640

10 Washington Times 35411 Time Magazine 65412 US News and World Report 65813 NPR Morning Edition 66314 Newsweek 66315 CBS Early Show 66616 Washington Post 66617 Los Angeles Times 70018 CBS Evening News 73719 New York Times 73720 Wall Street Journal 851

The table gives our methodrsquos rankings of the most to least centrist news outlet The rankings are basedon the distance of the outletrsquos estimated ADA score (from Table III) to 5006 our estimate of the averageUnited States voterrsquos ADA score


The first second and third most centrist outlets are respec-tively Newshour with Jim Lehrer CNNrsquos Newsnight with AaronBrown and ABCrsquos Good Morning America The scores of News-night and Good Morning America were not statistically differentfrom the center 501 Although the point estimate of Newshourwas more centrist than the other two outlets its difference fromthe center is statistically significant The reason is that its mar-gin of error is smaller than the other two which is due primarilyto the fact that we collected more observations for this outletInterestingly in the four presidential and vice-presidential de-bates of the 2004 election three of the four moderators wereselected from these three outlets The fourth moderator BobSchieffer worked at an outlet that we did not examine CBSrsquosFace the Nation

The fourth and fifth most centrist outlets are the DrudgeReport and Fox Newsrsquo Special Report with Brit Hume Theirscores are significantly different from the center at a 95 percentsignificance level Nevertheless the top five outlets in Table IVare in a statistical dead heat for most centrist Even at an 80percent level of significance none of these outlets can be calledmore centrist than any of the others

The sixth and seventh most centrist outlets are ABC WorldNews Tonight and NBC Nightly News These outlets are almost ina statistical tie with the five most centrist outlets For instanceeach has a score that is significantly different from Newshourrsquos atthe 90 percent confidence level but not at the 95 percent confi-dence level The eighth most centrist outlet USA Today receiveda score that is significantly different from Newshourrsquos at the 95percent confidence level

Fox Newsrsquo Special Report is approximately one point morecentrist than ABCrsquos World News Tonight (with Peter Jennings)and NBCrsquos Nightly News (with Tom Brokaw) In neither case isthe difference statistically significant Given that Special Reportis one hour long and the other two shows are a half-hour long ourmeasure implies that if a viewer watched all three shows eachnight he or she would receive a nearly perfectly balanced versionof the news (In fact it would be slanted slightly left by 04ADA points)

Special Report is approximately thirteen points more centristthan CBS Evening News (with Dan Rather) This difference issignificant at the 99 percent confidence level Also at 99 percentconfidence levels we can conclude that NBC Nightly News


and ABC World News Tonight are more centrist than CBSEvening News

The most centrist newspaper in our sample is USA TodayHowever its distance from the center is not significantly differentfrom the distances of the Washington Times or the WashingtonPost Interestingly our measure implies that if one spent anequal amount of time reading the Washington Times and Wash-ington Post he or she would receive a nearly perfectly balancedversion of the news (It would be slanted left by only 09ADA points)

If instead we use the 541 as our measure of centrist (whichis based on congressional scores of the 1975ndash1994 period) therankings change but not greatly The most substantial is the FoxNewsrsquo Special Report which drops from fifth to fifteenth mostcentrist The Washington Times also changes significantly Itdrops from tenth to seventeenth most centrist

Another implication of the scores concerns the New YorkTimes Although some claim that the liberal bias of the New YorkTimes is balanced by the conservative bias of other outlets suchas the Washington Times or Fox Newsrsquo Special Report this is notquite true The New York Times is slightly more than twice as farfrom the center as Special Report Consequently to gain a bal-anced perspective a news consumer would need to spend twice asmuch time watching Special Report as he or she spends readingthe New York Times Alternatively to gain a balanced perspec-tive a reader would need to spend 50 percent more time readingthe Washington Times than the New York Times

VIII POTENTIAL BIASES

A frequent concern of our method is a form of the followingclaim ldquoThe sample of think tanks has a rightward (leftward) tiltrather than an ideological balance For example the sample doesnot include Public Citizen and many other ldquoNaderrdquo groups (Forexample the sample includes National Association of Manufac-turers the Conference of Catholic Bishops or any number ofother groups) Consequently this will bias estimates to the right(left)rdquo However the claim is not true and here is the intuition ifthe sample of think tanks were say disproportionately conser-vative this of course would cause media outlets to cite conser-vative think tanks more frequently (as a proportion of citationsthat we record in our sample) This might seem to cause the


media to appear more conservative However at the same time itcauses members of Congress to appear more conservative Ourmethod only measures the degree to which media is liberal orconservative relative to Congress Since it is unclear how such adisproportionate sample would affect the relative degree to whichthe media cite conservative (or liberal) think tanks there is no apriori reason for this to cause a bias

In fact a similar concern could be leveled against any regres-sion analysis As a simple example consider a researcher whoregresses the arm lengths of subjects on their heights Supposeinstead of choosing a balance of short and tall subjects he or shechooses a disproportionate number of tall subjects This will notaffect his or her findings about the relationship between heightand arm length That is he or she will find that arm length isapproximately half the subjectrsquos height and this estimate ldquohalfrdquowould be the same (in expectation) whether he or she choosesmany or few tall subjects For similar reasons to achieve unbi-ased estimates in a regression econometrics textbooks place norestrictions on the distribution of independent variables Theyonly place restrictions upon eg the correlation of the indepen-dent variables and the error term

Another frequent concern of our method takes a form of thefollowing claim ldquoMost of the congressional data came from yearsin which the Republicans were the majority party Since themajority can control debate time this will cause the sample tohave a disproportionate number of citations by Republicans Inturn this will cause media outlets to appear to be more liberalthan they really arerdquo First it is not true that the majority partygives itself a disproportionate amount of debate time Instead theusual convention is that it is divided equally between proponentsand opponents on an issue This means that the majority partyactually gives itself less than the proportionate share Howeverthis convention is countered by two other factors which tend togive the majority and minority party their proportionate share ofspeech time i) many of the speeches in the Congressional Recordare not part of the debate on a particular bill or amendment butare from ldquospecial ordersrdquo (generally in the evening after thechamber has adjourned from official business) or ldquoone minutesrdquo(generally in the morning before the chamber has convened forofficial business) For these types of speeches there are no restric-tions of party balance and for the most part any legislator whoshows up at the chamber is allowed to make such a speech


ii) Members often place printed material ldquointo the Recordrdquo Weincluded such printed material as a part of any memberrsquos speechIn general there are no restrictions on the amount of materialthat a legislator can place into the Record (or whether he or shecan do this) Thus eg if a legislator has run out of time to makehis or her speech he or she can request that the remainder beplaced in written form ldquointo the Recordrdquo

But even if the majority party were given more (or less) thanits proportionate share of speech time this would not bias ourestimates With each media outlet our method seeks the legisla-tor who has a citation pattern that is most similar to that outletFor instance suppose that the New York Times cites liberal thinktanks about twice as often as conservative think tanks Suppose(as we actually find) that Joe Lieberman is the legislator who hasthe mix of citations most similar to the New York Times that issuppose that he also tends to cite liberal think tanks twice asoften as conservative think tanks Now consider a congressionalrules change that cuts the speech time of Democrats in halfAlthough this will affect the number of total citations that Lieber-man makes it will not affect the proportion of citations that hemakes to liberal and conservative think tanks Hence our methodwould still give the New York Times an ADA score equal to JoeLiebermanrsquos36

More problematic is a concern that congressional citationsand media citations do not follow the same data-generating pro-cess For instance suppose that a factor besides ideology affectsthe probability that a legislator or reporter will cite a think tankand suppose that this factor affects reporters and legislatorsdifferently Indeed Lott and Hassett [2004] have invoked a formof this claim to argue that our results are biased toward mak-ing the media appear more conservative than they really areThey note

For example Lott [2003 Chapter 2] shows that the New York Timesrsquo storieson gun regulations consistently interview academics who favor gun controlbut use gun dealers or the National Rifle Association to provide the other side In this case this bias makes [Groseclose and Milyorsquos measure of] the NewYork Times look more conservative than is likely accurate [p 8]

36 Another concern is that although Republicans and Democrats are givendebate time nearly proportional to their number of seats one group might citethink tanks more frequently than the other The above reasoning also explainswhy this will not cause a bias to our method


However it is possible and perhaps likely that members ofCongress practice the same tendency that Lott and Hassett [2004]have identified with reporters that is to cite academics when theymake an antigun argument and to cite say the NRA when theymake a progun argument If so then our method will have no biasOn the other hand if members of Congress do not practice the sametendency as journalists then this can cause a bias to our methodBut even here it is not clear in which direction it will occur Forinstance it is possible that members of Congress have a greater(lesser) tendency than journalists to cite such academics If so thenthis will cause our method to make media outlets appear moreliberal (conservative) than they really are

In fact the criticism we have heard most frequently is a form ofthis concern but it is usually stated in a way that suggests the bias isin the opposite direction Here is a typical variant ldquoIt is possible that (i)journalists care about the lsquoqualityrsquo of a think tank more than legislatorsdo (eg suppose that journalists prefer to cite a think tank with areputation for serious scholarship instead of another group that isknown more for its activism) and (ii) the liberal think tanks in thesample tend to be of higher quality than the conservative think tanksrdquoIf statements (i) and (ii) are true then our method will indeed makemedia outlets appear more liberal than they really are That is themedia will cite liberal think tanks more not because they prefer to citeliberal think tanks but because they prefer to cite high-quality thinktanks On the other hand if one statement is true and the other is falsethen our method will make media outlets appear more conservativethan they really are For example suppose that journalists care aboutquality more than legislators but suppose that the conservative groupsin our sample tend to be of higher quality than the liberal groups Thenthe media will tend to cite the conservative groups disproportionatelybut not because the media are conservative rather because they havea taste for quality) Finally if neither statement is true then ourmethod will make media outlets appear more liberal than they reallyare Note that there are four possibilities by which statements (i) and(ii) can be true or false Two lead to a liberal bias and two lead to aconservative bias

This criticism in fact is similar to an omitted-variable bias thatcan plague any regression Like the regression case however if theomitted variable (eg the quality of the think tank) is not correlatedwith the independent variable of interest (eg the ideology of thethink tank) then this will not cause a bias In the Appendix weexamine this criticism further by introducing three variables that


measure the extent to which a think tankrsquos main goals are scholarlyones as opposed to activist ones That is these variables are possiblemeasures of the ldquoqualityrdquo of a think tank When we include thesemeasures as controls in our likelihood function our estimated ADAratings do not change significantly For example when we includethe measures the average score of the twenty news outlets that weexamine shifts less than three points Further we cannot reject thehypothesis that the new estimates are identical to the estimates thatwe obtain when we do not include the controls

Finally some anecdotal evidence provides a compelling ar-gument that our method is not biased Note that none of theissues discussed above suggest a problem with the way ourmethod ranks media outlets Now suppose that there is no prob-lem with the rankings yet our method is plagued with a signifi-cant bias that systematically causes media outlets to appear moreliberal (conservative) than they really are If so then this meansthat the three outlets we find to be most centrist (Newshour withJim Lehrer Good Morning America and Newsnight with AaronBrown) are actually conservative (liberal) But if this is true whydid John Kerryrsquos (George W Bushrsquos) campaign agree to allowthree of the four debate moderators to come from these outlets

IX DISCUSSION IMPLICATIONS FOR THE INDUSTRIAL

ORGANIZATION OF THE NEWS INDUSTRY

At least four broad empirical regularities emerge from ourresults In this section we document these regularities and ana-lyze their significance for some theories about the industrialorganization of the news industry

First we find a systematic tendency for the United Statesmedia outlets to slant the news to the left As mentioned earlier thisis inconsistent with basic spatial models of firm location such asHotellingrsquos [1929] and others In such models if an equilibriumexists then there is always an equilibrium in which the median firmlocates at the ideal location of the median consumer which ourresults clearly do not support

Some scholars have extended the basic spatial model to pro-vide a theory why the media could be systematically biased Forinstance Hamilton [2004] notes that news producers may preferto cater to some consumers more than others In particularHamilton notes that young females tend to be one of the mostmarginal groups of news consumers (ie they are the most will-


ing to switch to activities besides reading or watching the news)Further this group often makes the consumption decisions forthe household For these two reasons advertisers are willing topay more to outlets that reach this group Since young femalestend to be more liberal on average a news outlet may want toslant its coverage to the left Thus according to Hamiltonrsquos the-ory United States news outlets slant their coverage leftward notin spite of consumer demand but because of it37

A more compelling explanation for the liberal slant of newsoutlets in our view involves production factors not demand factorsAs Sutter [2001] has noted journalists might systematically have ataste to slant their stories to the left Indeed this is consistent withthe survey evidence that we noted earlier As a consequence ldquoIf themajority of journalists have left-of-center views liberal news mightcost less to supply than unbiased news [p 444]rdquo Baron [2005]constructs a rigorous mathematical model along these lines In hismodel journalists are driven not just by money but also a desire toinfluence their readers or viewers Baron shows that profit-maxi-mizing firms may choose to allow reporters to slant their stories andconsequently in equilibrium the media will have a systematic bias38

A second empirical regularity is that the media outlets thatwe examine are fairly centrist relative to members of CongressFor instance as Figure II shows all outlets but one have ADAscores between the average Democrat and average Republican inCongress In contrast it is reasonable to believe that at least halfthe voters consider themselves more extreme than the partyaverages39 If so then a basic spatial model where firms are

37 Sutter [2001] similarly notes that demand factors may be the source ofliberal bias in the newspaper industry Specifically he notes that liberals mayhave a higher demand for newspapers than conservatives and he cites somesuggestive evidence by Goff and Tollison [1990] which shows that as the voters ina state become more liberal newspaper circulation in the state increases

38 Perhaps even more interesting in Baronrsquos model news consumers inequilibrium can be influenced in the direction of the bias of the news outletdespite the fact that they understand the equilibrium of the game and thepotential incentives of journalists to slant the news

39 A simple model supports this assertion Suppose that in every congres-sional district voters have ideal positions that are uniformly distributed between1 and 1 where 1 represents the most liberal voter and 1 represents the mostconservative voter Assume that a voter is a Democrat if and only if his or her idealposition is less than 0 Four candidates two Republican and two Democratsimultaneously choose positions in this space Next they compete in two primaryelections where the Republican voters choose between the two Republican can-didates and likewise for the Democratic primary Each voter votes for the candi-date who is nearest his or her ideological position and if two candidates areequidistant then the voter flips a coin (This assumption implies that voters aremyopic in the primary election If instead the voters were fully rational then it


can easily be shown that the candidates will choose even more centrist positionswhich means that even more voters will consider themselves more extreme thanthe party averages) Assume that candidates maximize the votes that they receivein the general election (ie the votes they receive in the primary election are onlya means to winning votes in the general election) Then this setup implies that inequilibrium both Democratic candidates will locate at 5 and both Republicancandidates will locate at 5 Each winner of the primary has a 50 percent chanceat winning the general election Once this is repeated across many districts thenthe expected number of voters who consider themselves more extreme than theparty averages will be 50 percent

FIGURE IIAdjusted ADA Scores of Selected Politicians and Media Outlets


constrained to charge the same exogenous price implies thatapproximately half the media outlets should choose a slant out-side the party averages40 Clearly our results do not support thisprediction

Moreover when we add price competition to the basicspatial model then as Mullainathan and Shleifer [2003] showeven fewer media outlets should be centrist Specifically theirtwo-firm model predicts that both media firms should chooseslants that are outside the preferred slants of all consumersThe intuition is that in the first round when firms chooselocations they want to differentiate their products significantly soin the next round they will have less incentive to compete on priceGiven this theoretical result it is puzzling that media outlets in theUnited States are not more heterogeneous We suspect that onceagain the reason may lie with production factors For instance onepossibility may involve the sources for news storiesmdashwhat one couldconsider as the raw materials of the news industry If a news outletis too extreme many of the newsmakers may refuse to grant inter-views to the reporters

A third empirical regularity involves the question whetherreporters will be faithful agents of the owners of the firms forwhich they work That is will the slant of their news storiesreflect their own ideological preferences or the firmrsquos ownersThe conventional wisdom at least among left-wing commenta-tors is that the latter is true For instance Alterman [2003]titles a chapter of his book ldquoYoursquore Only as Liberal as the ManWho Owns Yourdquo A weaker assertion is that the particularnews outlet will be a faithful agent of the firm that owns itHowever our results provide some weak evidence that this isnot true For instance although Time magazine and CNNrsquosNewsnight are owned by the same firm (Time Warner) theirADA scores differ substantially by 94 points41 Further al-

40 For instance suppose that consumers are distributed uniformly between1 and 1 Suppose that there are twenty news outlets and suppose that consum-ers choose the outlet that is closest to them It is easy to show that an equilibriumis for two firms to locate at 9 two firms to locate at 7 and two firms tolocate at 9

41 This difference however is not statistically significant at the 95 percentconfidence level A likelihood ratio test constraining Time and Newsnight to havethe same score gives a log-likelihood function that is 11 units greater than theunconstrained function This value multiplied by two follows a Chi-Squaredistribution with one degree of freedom The result 22 is almost significant atthe 90 percent confidence level but not quite (The latter has a criterion of 271)We obtained similar results when we tested the joint hypothesis that (i) News-


most half of the other outlets have scores between those ofNewsnight and Time

A fourth regularity concerns the question whether one shouldexpect a government-funded news outlet to be more liberal thana privately funded outlet ldquoRadical democraticrdquo media scholarsMcChesney and Scott [2004] claim that it will For instance theynote ldquo[Commercial journalism] has more often served the minor-ity interests of dominant political military and business con-cerns than it has the majority interests of disadvantaged socialclasses [2004 p 4]rdquo And conservatives who frequently complainthat NPR is far left also seem to agree However our results do notsupport such claims If anything the government-funded outlets inour sample (NPRrsquos Morning Edition and Newshour with JimLehrer) have a slightly lower average ADA score (610) than theprivate outlets in our sample (628)42 Related some claim that afree-market system of news will produce less diversity of news thana government-run system However again our results do not sup-port such a claim The variance of the ADA scores of the privatelyrun outlets is substantially higher (1313) than the variance of thetwo government-funded outlets that we examine (551)

In interpreting some of the above regularities especially per-haps the latter two we advise caution For instance with regard toour comparisons of government-funded versus privately fundednews outlets we should emphasize that our sample of government-funded outlets is small (only two) and our total sample of newsoutlets might not be representative of all news outlets

night and Time have identical scores and that (ii) all three network morning newsshows have scores identical to their respective evening news shows A likelihoodratio test gives a value of 804 which follows a Chi-Square distribution with fourdegrees of freedom The value is significant at the 90 percent confidence level(criterion 778) but not at the 95 percent confidence level (criterion 949) Ourhunch is that with more data we could show conclusively that at least sometimesdifferent news outlets at the same firm produce significantly different slants Wesuspect that consistent with Baronrsquos [2005] model editors and producers likereporters are given considerable slack and that they are willing to sacrificesalary in order to be given such slack

42 This result is broadly consistent with Djankov McLiesh Nenova andShleiferrsquos [2003] notion of the public choice theory of media ownership This theoryasserts that a government-owned media will slant news in such a way to aidincumbent politicians If so some reasonable theories (eg Black [1958]) suggestthat the slant should conform to the median view of the incumbent politicians Weindeed find that the slant of the government-funded outlets in the United Stateson average is fairly close to the median politiciansrsquo view In fact it is closer to themedian view than the average of the privately funded outlets that we examineSee Lott [1999] for an examination of a similar public-choice theory applied to themedia and the education system in a country


Related in our attempts to explain these patterns we in no wayclaim to have provided the last word on a satisfactory theory Nor dowe claim to have performed an exhaustive review of potential the-ories in the literature Rather the main goal of our research issimply to demonstrate that it is possible to create an objectivemeasure of the slant of the news Once this is done as we hope wehave demonstrated in this section it is easy to raise a host oftheoretical issues to which such a measure can be applied

APPENDIX

We believe that the most appropriate model specification isthe one that we used to generate Table III However in thisAppendix we consider alternative specifications

Recall that we excluded observations in which the journalistor legislator gave an ideological label to the think tank or policygroup The first column of Table V lists ADA estimates wheninstead we include these observations while maintaining all the

TABLE VESTIMATED ADA SCORES FOR ALTERNATIVE SPECIFICATIONS

Media outlet 1 2 3 4 5 6 7 8 9

ABC Good Morning America 567 560 550 560 593 595 562 555 454ABC World News Tonight 614 613 609 620 616 624 609 598 587CBS Early Show 675 671 641 675 678 683 660 649 568CBS Evening News 721 740 740 746 732 741 728 717 696CNN NewsNight with Aaron Brown 558 558 548 580 560 564 555 533 517Drudge Report 553 606 590 625 608 621 602 581 560Fox News Special Report 415 390 388 412 405 406 398 388 334Los Angeles Times 678 704 694 717 705 709 693 685 658NBC Nightly News 621 617 631 630 613 623 612 602 609NBC Today Show 640 648 647 652 651 661 638 629 559New York Times 699 749 726 743 739 747 733 716 708Newshour with Jim Lehrer 551 560 544 570 558 559 560 536 509Newsweek 657 667 645 670 669 675 657 644 689NPR Morning Edition 656 669 662 674 661 671 661 646 592Time Magazine 682 655 624 662 643 654 642 633 647US News and World Report 652 658 653 670 658 664 648 636 657USA Today 617 632 625 637 628 639 624 604 669Wall Street Journal 861 851 858 862 855 864 848 825 821Washington Post 647 670 655 674 668 672 667 643 567Washington Times 357 338 344 362 353 362 348 329 480Average of all 20 outlets 621 628 619 637 630 637 622 607 594


other assumptions that we used to create Table III eg that weuse 44 actual think tanks and 6 mega think tanks etc

In column 2 we report the results when we exclude citationsof the ACLU (while we maintain all the other model specificationswe used to construct Table III including the decision to omitlabeled observations)

In columns 3 to 8 we report the results when instead of using44 actual think tanks and 6 mega think tanks we use 48 (respec-tively 47 46 45 43 and 42) actual and 2 (respectively 3 4 5 7and 8) mega think tanks

In column 9 we use sentences as the level of observationinstead of citations One problem with this specification is thatthe data are very lumpy that is some quotes contain an inordi-nate number of sentences which cause some anomalies Oneanomaly is that some relatively obscure think tanks become someof the most-cited under this specification For instance the Alexisde Tocqueville Institute which most readers would agree is not oneof the most well-known and prominent think tanks is the thirteenthmost-cited think tank by members of Congress when we use sen-tences as the level of observation It is the fifty-eighth most-citedhowever when we use citations as the level of observation43 Arelated problem is that these data are serially correlated That is forinstance if a given observation for the New York Times is a citationto the Brookings Institution then the probability is high that thenext observation will also be a citation to the same think tank (sincethe average citation contains more than one sentence) However thelikelihood function that we use assumes that the observations arenot serially correlated Finally related to these problems the esti-mates from this specification sometimes are in stark disagreementwith common wisdom For instance the estimates imply that theWashington Times is more liberal than Good Morning America Forthese reasons we base our conclusions on the estimates that usecitations as the level of observation rather than sentences

In columns 1 to 4 of Table VI we report the results wheninstead of using 44 actual think tanks and 6 mega think tanks we

43 Nunberg [2004] in a critique of an earlier version of our paper deservescredit for first noting the problems with the sentence-level data involving theAlexis de Tocqueville Institute Our earlier version gave approximately equalfocus to (i) estimates using citations as the level of observation and (ii) estimatesusing sentences as the level of observation Partly due to his critique the currentversion no longer focuses on sentences as observations We did not have the sameagreement with the rest of his criticisms however See Groseclose and Milyo[2004] for a response to his essay


use 54 (respectively 64 74 and 84) actual think tanks and 6mega think tanks That is we let the total number of think tanksthat we use change to 60 70 80 and 90

Columns 5ndash9 of Table VI address the concern that our mainanalysis does not control for the ldquoqualityrdquo of a think tank or policygroup To account for this possibility we constructed three variablesthat indicate whether a think tank or policy group is more likely toproduce quality scholarship The first variable closed membershipis coded as a 0 if the web site of the group asks visitors to join thegroup For instance more activist groupsmdashsuch as the NAACP NRAand ACLUmdashhave links on their web site that give instructions for avisitor to join the group while the more scholarly groupsmdashsuch asthe Brookings Institution the RAND Corporation the Urban Insti-tute and the Hoover Institutionmdashdo not Another variable staffcalled fellows is coded as 1 if any staff members on the grouprsquos website are given one of the following titles fellow (including researchfellow or senior fellow) researcher economist or analyst

Both variables seem to capture the conventional wisdom aboutwhich think tanks are known for quality scholarship For instance

TABLE VIESTIMATED ADA SCORES FOR ALTERNATIVE SPECIFICATIONS

Media outlet 1 2 3 4 5 6 7 8 9

ABC Good Morning America 569 599 602 603 632 609 625 639 617ABC World News Tonight 616 624 629 629 617 588 606 621 593CBS Early Show 671 689 689 690 660 630 645 661 631CBS Evening News 740 743 750 750 776 742 763 786 753CNN NewsNight with Aaron Brown 562 566 573 573 552 524 533 550 522Drudge Report 602 611 610 607 631 606 622 636 610Fox News Special Report 417 422 425 423 405 387 386 400 381Los Angeles Times 695 700 701 698 714 682 699 709 680NBC Nightly News 633 634 636 635 639 614 625 646 621NBC Today Show 652 666 664 666 673 641 660 680 649New York Times 741 750 753 749 757 727 745 763 734Newshour with Jim Lehrer 581 585 590 593 603 564 583 598 560Newsweek 669 676 684 680 687 650 673 682 649NPR Morning Edition 672 679 683 682 689 656 676 693 661Time Magazine 656 657 659 657 649 615 639 646 617US News and World Report 661 672 684 685 697 663 681 699 667USA Today 633 645 649 650 695 657 680 691 656Wall Street Journal 849 866 869 868 868 834 852 872 838Washington Post 662 668 669 668 689 660 676 698 668Washington Times 354 358 361 351 412 400 393 408 394Average of 20 outlets 632 641 644 643 652 622 638 654 625


of the top-25 most-cited groups in Table I the following had bothclosed membership and staff called fellows Brookings Center forStrategic and International Studies Council on Foreign RelationsAEI RAND Carnegie Endowment for Intl Peace Cato Institute forInternational Economics Urban Institute Family Research Coun-cil and Center on Budget and Policy Priorities Meanwhile thefollowing groups which most would agree are more commonlyknown for activism than high-quality scholarship had neitherclosed membership nor staff called fellows ACLU NAACP SierraClub NRA AARP Common Cause Christian Coalition NOW andFederation of American Scientists44

The third variable that we constructed is off K street It iscoded as a 1 if and only if the headquarters of the think tank orpolicy group is not located on Washington DCrsquos K Street thefamous street for lobbying firms45

Recall that in the estimation process for Table III we esti-mated individual ajrsquos and bjrsquos only for the 44 think tanks that themedia cited most All other think tanks were placed into one of sixmega think tanks It is not clear how one should code the qualityvariables for the mega think tanks For example should a megathink tank be coded as one if most of the actual think tanks thatcomprise it have closed membership Alternatively should itreceive the average closed-membership score of the think tanksthat comprise it If so should such an average be weighted by thenumber of times that the media cite the actual think tanksShould instead such weights include the number of times thatlegislators cite it Another complicating factor is that a few of themore minor think tanks no longer have web sites which made itimpossible for us to code the quality variables for them Insteadwe altered our analysis so that we only used data from the top 50most-cited think tanks and we did not include any mega thinktanks in this analysis These think tanks comprised approxi-mately 88 percent of the media citations in our total sample

So that we are comparing apples with apples we constructbaseline estimates for comparing the effect of the quality vari-ables These estimates listed in column 5 of Table VI use data

44 Despite its name the Federation of American Scientists is more of alobbying group than a scholarly think tank Indeed like most other well-knownlobbying groups its address is on K Street in Washington DC

45 Only four of the 50 most-cited groups had an address on the street Thesewere Center for Strategic and International Studies Federation of AmericanScientists Employee Benefit Research Institute and People for the AmericanWay


only from the top 50 most-cited think tanks and do not exploit anyquality variables as controls Note that this specification causesthe media to appear more liberal than our main analysis com-pared with the estimates of Table III the average media score isapproximately 26 points higher

Next we incorporate quality variables in the likelihood func-tion In Table VI column 6 we use a likelihood function thatassumes that the probability that media outlet i will cite thinktank j is

(5) expaj bjci d1 closed membershipj

k1

J

expak bkci d1 closed membershipk

The likelihood function still uses (2) as the probability that amember of Congress cites think tank j ie it sets d1 to zero forthe congressional observations Thus d1 measures the extent towhich a media outlet is more likely than a legislator to cite athink tank with closed membership

Columns 7 and 8 of the table give estimates when we substi-tute staff called fellows and off k street for closed membership in(5) Column 9 of the table gives estimates when we include allthree of the control variables in the likelihood function As thesecolumns show when we include the quality variables this causesthe media scores to appear slightly more conservative Howeverthe change has very little substantive significance For instancein none of the three specifications does the average score changeby more than three ADA points Further the change is less thanthe effect of using only data from the top 50 think tanks That iswhen we compare these estimates with those in Table III we seethat if we (i) use data only from the top 50 most-cited think tanksand (ii) include quality variables then the net effect of these twodecisions is to make the media appear more liberal

The change from including the quality variables also hasvery little if any statistical significance For instance with eachspecification listed in columns 6ndash9 we reestimated the likelihoodfunction while constraining the media estimates to the valueslisted in column 5 (while allowing the estimates for the qualityvariables to reach their optimum values) Using a likelihood ratiotest even at p-values of 30 percent we could never reject the null


hypothesis that the quality variables cause no change to theestimated ADA scores

DEPARTMENT OF POLITICAL SCIENCE UNIVERSITY OF CALIFORNIA AT LOS ANGELES

DEPARTMENT OF ECONOMICS AND TRUMAN SCHOOL OF PUBLIC AFFAIRS UNIVERSITY

OF MISSOURI

REFERENCES

Alterman Eric What Liberal Media The Truth about Bias and the News (NewYork Basic Books 2003)

Baron David ldquoPersistent Media Biasrdquo Journal of Public Economics LXXXIXforthcoming (2005)

Black Duncan The Theory of Committees and Elections (London CambridgeUniversity Press 1958)

Bozell L B and B H Baker Thatrsquos the Way It Isnrsquot A Reference Guide to MediaBias (Alexandria VA Media Research Center 1990)

Crouse Timothy Boys on the Bus (New York Ballantine Books 1973)Djankov Simeon Caralee McLiesh Tatiana Nenova and Andrei Shleifer ldquoWho

Owns the Mediardquo Journal of Law and Economics XLVI (2003) 341ndash381Franken Al Lies and the Lying Liars Who Tell Them A Fair and Balanced Look

at the Right (New York Dutton 2003)Goff Brian and Robert Tollison ldquoWhy Is the Media so Liberalrdquo Journal of Public

Finance and Public Choice I (1990) 13ndash21Goldberg Bernard Bias A CBS Insider Exposes How the Media Distort the News

(Washington DC Regnery 2002)Groeling Tim and Samuel Kernell ldquoIs Network News Coverage of the President

Biasedrdquo Journal of Politics LX (1998) 1063ndash1087Groseclose Tim Steven D Levitt and James M Snyder Jr ldquoComparing Interest

Group Scores across Time and Chambers Adjusted ADA Scores for the U SCongressrdquo American Political Science Review XCIII (1999) 33ndash50

Groseclose Tim and Jeffrey Milyo ldquoResponse to lsquo ldquoLiberal Biasrdquo Noch Einmalrsquo rdquoLanguage Log viewed December 20 2004 httpitrecisupennedumyllanguagelogarchives001301html

Hamilton James All the News Thatrsquos Fit to Sell How the Market TransformsInformation into News (Princeton NJ Princeton University Press 2004)

Herman Edward S and Noam Chomsky Manufacturing Consent The PoliticalEconomy of the Mass Media (New York Pantheon Books 1988)

Hotelling Harold ldquoStability in Competitionrdquo Economic Journal XXXIX (1929)41ndash57

Irvine Reed and Cliff Kincaid ldquoPost Columnist Concerned About Media BiasrdquoAccuracy in Media viewed September 17 2001 httpwwwaimorgmedia_monitorA900_0_2_0_C

Jamieson Kathleen Hall Everything You Think You Know About Politics andWhy Yoursquore Wrong (New York Basic Books 2000)

Judge George G W E Griffiths R Carter Hill Helmut Lutkepohl and Tsoung-Chao Lee The Theory and Practice of Econometrics (New York John Wileyand Sons 1985)

Kurtz Howard ldquoFewer Republicans Trust the News Survey Findsrdquo WashingtonPost (June 9 2004) C01

Lichter S R S Rothman and L S Lichter The Media Elite (Bethesda MDAdler and Adler 1986)

Lott John R Jr ldquoPublic Schooling Indoctrination and Totalitarianismrdquo Jour-nal of Political Economy CVII (1999) S127ndashS157

mdashmdash The Bias Against Guns (Washington DC Regnery Publishing Inc 2003)Lott John R Jr and Kevin A Hassett ldquoIs Newspaper Coverage of Economic

Events Politically Biasedrdquo manuscript American Enterprise Institute 2004McChesney Robert and Ben Scott Our Unfree Press 100 Years of Radical Media

Criticism (New York The New Press 2004)


McFadden Daniel ldquoConditional Logit Analysis of Qualitative Choice Behaviorrdquo inFrontiers in Econometrics P Zarembka ed (New York Academic Press 1974)

Mullainathan Sendhil and Andrei Shleifer ldquoThe Market for Newsrdquo manuscriptHarvard University 2003

Nunberg Geoffrey ldquo lsquoLiberal Biasrsquo Noch Einmalrdquo Language Log viewed December20 2004 httpitrecisupennedumyllanguagelogarchives001169html

Parenti Michael Inventing Reality The Politics of the Mass Media (New York StMartinrsquos Press 1986)

Povich Elaine Partners and Adversaries The Contentious Connection BetweenCongress and the Media (Arlington VA Freedom Forum 1996)

Sperry Paul ldquoMyth of the Conservative Wall Street Journalrdquo WorldNetDailyviewed June 25 2002 wwwworldnetdailycomnewsarticleaspARTICLE_ID28078

Sutter Daniel ldquoCan the Media Be So Liberal The Economics of Media Biasrdquo TheCato Journal XX (2001) 431ndash451

mdashmdash ldquoAdvertising and Political Bias in the Mediardquo American Journal of Econom-ics and Sociology LXI (2002) 725ndash745

mdashmdash ldquoAn Indirect Test of the Liberal Media Thesis Using Newsmagazine Circu-lationrdquo manuscript University of Oklahoma 2004

Weaver D H and G C Wilhoit American Journalist in the 1990s (Mahwah NJLawrence Erlbaum 1996)

Woodward Bob The Agenda Inside the Clinton White House (New York Simon ampSchuster 1994)


when it calls Steve Forbesrsquo tax plan lsquowackyrsquordquo) and untestedtheories (ldquoif the news industry is a competitive market then howcan media outlets be systematically biasedrdquo)

Few studies provide an objective measure of the slant ofnews and none has provided a way to link such a measure toideological measures of other political actors That is none of theexisting measures can say for example whether the New YorkTimes is more liberal than Senator Edward Kennedy or whetherFox News is more conservative than Senator Bill Frist We pro-vide such a measure Namely we compute an adjusted Americansfor Democratic Action (ADA) score for various news outlets in-cluding the New York Times the Washington Post USA Todaythe Drudge Report Fox Newsrsquo Special Report and all three net-worksrsquo nightly news shows

Our results show a strong liberal bias All of the news outletsexcept Fox Newsrsquo Special Report and the Washington Times re-ceived a score to the left of the average member of Congress Anda few outlets including the New York Times and CBS EveningNews were closer to the average Democrat in Congress than thecenter These findings refer strictly to the news stories of theoutlets That is we omitted editorials book reviews and letters tothe editor from our sample

To compute our measure we count the times that a mediaoutlet cites various think tanks and other policy groups1 Wecompare this with the times that members of Congress cite thesame think tanks in their speeches on the floor of the House andSenate By comparing the citation patterns we can construct anADA score for each media outlet

As a simplified example imagine that there were only twothink tanks and suppose that the New York Times cited the firstthink tank twice as often as the second Our method asks what isthe estimated ADA score of a member of Congress who exhibitsthe same frequency (21) in his or her speeches This is the scorethat our method would assign the New York Times

A feature of our method is that it does not require us to make asubjective assessment of how liberal or conservative a think tank isThat is for instance we do not need to read policy reports of the

1 Our sample includes policy groups that are not usually called think tankssuch as the NAACP NRA and Sierra Club To avoid using the more unwieldyphrase ldquothink tanks and other policy groupsrdquo we often use a shorthand versionldquothink tanksrdquo When we use the latter phrase we mean to include the othergroups such as the NAACP etc



































II DATA






























group



media outlets













TABLE I(CONTINUED)



group



media outlets







































(1) aj bjxi eij





(2) expaj bjxi k1

J

expak bkxi



(3) aj bjcm emj


(4) expaj bjcm k1

J

expak bkcm














V RESULTS









Standarderror






















































TO THE CENTER



































































APPENDIX




Media outlet 1 2 3 4 5 6 7 8 9














Media outlet 1 2 3 4 5 6 7 8 9













k1

J









OF MISSOURI

REFERENCES







































































II DATA






























group



media outlets













TABLE I(CONTINUED)



group



media outlets







































(1) aj bjxi eij





(2) expaj bjxi k1

J

expak bkxi



(3) aj bjcm emj


(4) expaj bjcm k1

J

expak bkcm














V RESULTS









Standarderror






















































TO THE CENTER



































































APPENDIX




Media outlet 1 2 3 4 5 6 7 8 9














Media outlet 1 2 3 4 5 6 7 8 9













k1

J









OF MISSOURI

REFERENCES





























































II DATA






























group



media outlets













TABLE I(CONTINUED)



group



media outlets







































(1) aj bjxi eij





(2) expaj bjxi k1

J

expak bkxi



(3) aj bjcm emj


(4) expaj bjcm k1

J

expak bkcm














V RESULTS









Standarderror






















































TO THE CENTER



































































APPENDIX




Media outlet 1 2 3 4 5 6 7 8 9














Media outlet 1 2 3 4 5 6 7 8 9













k1

J









OF MISSOURI

REFERENCES























































II DATA






























group



media outlets













TABLE I(CONTINUED)



group



media outlets







































(1) aj bjxi eij





(2) expaj bjxi k1

J

expak bkxi



(3) aj bjcm emj


(4) expaj bjcm k1

J

expak bkcm














V RESULTS









Standarderror






















































TO THE CENTER



































































APPENDIX




Media outlet 1 2 3 4 5 6 7 8 9














Media outlet 1 2 3 4 5 6 7 8 9













k1

J









OF MISSOURI

REFERENCES

















































II DATA






























group



media outlets













TABLE I(CONTINUED)



group



media outlets







































(1) aj bjxi eij





(2) expaj bjxi k1

J

expak bkxi



(3) aj bjcm emj


(4) expaj bjcm k1

J

expak bkcm














V RESULTS









Standarderror






















































TO THE CENTER



































































APPENDIX




Media outlet 1 2 3 4 5 6 7 8 9














Media outlet 1 2 3 4 5 6 7 8 9













k1

J









OF MISSOURI

REFERENCES







































II DATA






























group



media outlets













TABLE I(CONTINUED)



group



media outlets







































(1) aj bjxi eij





(2) expaj bjxi k1

J

expak bkxi



(3) aj bjcm emj


(4) expaj bjcm k1

J

expak bkcm














V RESULTS









Standarderror






















































TO THE CENTER



































































APPENDIX




Media outlet 1 2 3 4 5 6 7 8 9














Media outlet 1 2 3 4 5 6 7 8 9













k1

J









OF MISSOURI

REFERENCES



























































group



media outlets













TABLE I(CONTINUED)



group



media outlets







































(1) aj bjxi eij





(2) expaj bjxi k1

J

expak bkxi



(3) aj bjcm emj


(4) expaj bjcm k1

J

expak bkcm














V RESULTS









Standarderror






















































TO THE CENTER



































































APPENDIX




Media outlet 1 2 3 4 5 6 7 8 9














Media outlet 1 2 3 4 5 6 7 8 9













k1

J









OF MISSOURI

REFERENCES




















































group



media outlets













TABLE I(CONTINUED)



group



media outlets







































(1) aj bjxi eij





(2) expaj bjxi k1

J

expak bkxi



(3) aj bjcm emj


(4) expaj bjcm k1

J

expak bkcm














V RESULTS









Standarderror






















































TO THE CENTER



































































APPENDIX




Media outlet 1 2 3 4 5 6 7 8 9














Media outlet 1 2 3 4 5 6 7 8 9













k1

J









OF MISSOURI

REFERENCES















































group



media outlets













TABLE I(CONTINUED)



group



media outlets







































(1) aj bjxi eij





(2) expaj bjxi k1

J

expak bkxi



(3) aj bjcm emj


(4) expaj bjcm k1

J

expak bkcm














V RESULTS









Standarderror






















































TO THE CENTER



































































APPENDIX




Media outlet 1 2 3 4 5 6 7 8 9














Media outlet 1 2 3 4 5 6 7 8 9













k1

J









OF MISSOURI

REFERENCES










































group



media outlets













TABLE I(CONTINUED)



group



media outlets







































(1) aj bjxi eij





(2) expaj bjxi k1

J

expak bkxi



(3) aj bjcm emj


(4) expaj bjcm k1

J

expak bkcm














V RESULTS









Standarderror






















































TO THE CENTER



































































APPENDIX




Media outlet 1 2 3 4 5 6 7 8 9














Media outlet 1 2 3 4 5 6 7 8 9













k1

J









OF MISSOURI

REFERENCES









































TABLE I(CONTINUED)



group



media outlets







































(1) aj bjxi eij





(2) expaj bjxi k1

J

expak bkxi



(3) aj bjcm emj


(4) expaj bjcm k1

J

expak bkcm














V RESULTS









Standarderror






















































TO THE CENTER



































































APPENDIX




Media outlet 1 2 3 4 5 6 7 8 9














Media outlet 1 2 3 4 5 6 7 8 9













k1

J









OF MISSOURI

REFERENCES









































































(1) aj bjxi eij





(2) expaj bjxi k1

J

expak bkxi



(3) aj bjcm emj


(4) expaj bjcm k1

J

expak bkcm














V RESULTS









Standarderror






















































TO THE CENTER



































































APPENDIX




Media outlet 1 2 3 4 5 6 7 8 9














Media outlet 1 2 3 4 5 6 7 8 9













k1

J









OF MISSOURI

REFERENCES

































































(1) aj bjxi eij





(2) expaj bjxi k1

J

expak bkxi



(3) aj bjcm emj


(4) expaj bjcm k1

J

expak bkcm














V RESULTS









Standarderror






















































TO THE CENTER



































































APPENDIX




Media outlet 1 2 3 4 5 6 7 8 9














Media outlet 1 2 3 4 5 6 7 8 9













k1

J









OF MISSOURI

REFERENCES

























































(1) aj bjxi eij





(2) expaj bjxi k1

J

expak bkxi



(3) aj bjcm emj


(4) expaj bjcm k1

J

expak bkcm














V RESULTS









Standarderror






















































TO THE CENTER



































































APPENDIX




Media outlet 1 2 3 4 5 6 7 8 9














Media outlet 1 2 3 4 5 6 7 8 9













k1

J









OF MISSOURI

REFERENCES


















































(1) aj bjxi eij





(2) expaj bjxi k1

J

expak bkxi



(3) aj bjcm emj


(4) expaj bjcm k1

J

expak bkcm














V RESULTS









Standarderror






















































TO THE CENTER



































































APPENDIX




Media outlet 1 2 3 4 5 6 7 8 9














Media outlet 1 2 3 4 5 6 7 8 9













k1

J









OF MISSOURI

REFERENCES













































(1) aj bjxi eij





(2) expaj bjxi k1

J

expak bkxi



(3) aj bjcm emj


(4) expaj bjcm k1

J

expak bkcm














V RESULTS









Standarderror






















































TO THE CENTER



































































APPENDIX




Media outlet 1 2 3 4 5 6 7 8 9














Media outlet 1 2 3 4 5 6 7 8 9













k1

J









OF MISSOURI

REFERENCES








































(1) aj bjxi eij





(2) expaj bjxi k1

J

expak bkxi



(3) aj bjcm emj


(4) expaj bjcm k1

J

expak bkcm














V RESULTS









Standarderror






















































TO THE CENTER



































































APPENDIX




Media outlet 1 2 3 4 5 6 7 8 9














Media outlet 1 2 3 4 5 6 7 8 9













k1

J









OF MISSOURI

REFERENCES






































(2) expaj bjxi k1

J

expak bkxi



(3) aj bjcm emj


(4) expaj bjcm k1

J

expak bkcm














V RESULTS









Standarderror






















































TO THE CENTER



































































APPENDIX




Media outlet 1 2 3 4 5 6 7 8 9














Media outlet 1 2 3 4 5 6 7 8 9













k1

J









OF MISSOURI

REFERENCES















































V RESULTS









Standarderror






















































TO THE CENTER



































































APPENDIX




Media outlet 1 2 3 4 5 6 7 8 9














Media outlet 1 2 3 4 5 6 7 8 9













k1

J









OF MISSOURI

REFERENCES









































V RESULTS









Standarderror






















































TO THE CENTER



































































APPENDIX




Media outlet 1 2 3 4 5 6 7 8 9














Media outlet 1 2 3 4 5 6 7 8 9













k1

J









OF MISSOURI

REFERENCES






































Documents

THE QUARTERLY JOURNAL OF ECONOMICS - UCLA