/Users/andrea/_magisterarbeit/korpus/clean/trainkorpus/1/file2.html NN ----------------------------------------- : A DT Measure NN of IN Media NP Bias NP Tim NP Groseclose NP Department NP of IN Political NP Science NP UCLA NP Jeff NP Milyo NP Department NP of IN Economics NP University NP of IN Missouri NP December NP 2004 CD We PP are VBP grateful JJ for IN the DT research NN assistance NN by IN Aviva NP Aminova NP , , Jose NP Bustos NP , , Anya NP Byers NP , , Evan NP Davidson NP , , Kristina NP Doan NP , , Wesley NP Hussey NP , , David NP Lee NP , , Pauline NP Mena NP , , Orges NP Obeqiri NP , , Byrne NP Offut NP , , Matt NP Patterson NP , , David NP Primo NP , , Darryl NP Reeves NP , , Susie NP Rieniets NP , , Tom NP Rosholt NP , , Michael NP Uy NP , , Diane NP Valos NP , , Michael NP Visconti NP , , Margaret NP Vo NP , , Rachel NP Ward NP , , and CC Andrew NP Wright NP . SENT Also RB , , we PP are VBP grateful JJ for IN comments NNS and CC suggestions NNS by IN Matt NP Baum NP , , Mark NP Crain NP , , Tim NP Groeling NP , , Phil NP Gussin NP , , Jay NP Hamilton NP , , Wesley NP Hussey NP , , Chap NN Lawson NP , , Steve NP Levitt NP , , Jeff NP Lewis NP , , Andrew NP Martin NP , , David NP Mayhew NP , , Jeff NP Minter NP , , Mike NP Munger NP , , David NP Primo NP , , Andy NP Waddell NP , , Barry NP Weingast NP , , John NP Zaller NP , , and CC Jeff NP Zwiebel NP . SENT We PP also RB owe VVP gratitude NN to TO UCLA NP , , University NP of IN Missouri NP , , Stanford NP University NP , , and CC the DT University NP of IN Chicago NP . SENT These DT universities NNS paid VVD our PP$ salaries NNS , , funded VVN our PP$ research NN assistants NNS , , and CC paid VVN for IN services NNS such JJ as IN Lexis NP Nexis NP , , which WDT were VBD necessary JJ for IN our PP$ data NN collection NN . SENT No DT other JJ organization NN or CC person NN helped VVD to TO fund VV this DT research NN project NN . SENT A DT Measure NN of IN Media NP Bias NP In IN this DT paper NN we PP estimate VVP ADA NP Americans NPS for IN Democratic NP Action NP scores NNS for IN major JJ media NNS outlets NNS such JJ as IN the DT New NP York NP Times NP , , USA NP Today NP , , Fox NP News NP Special JJ Report NP , , and CC all DT three CD network NN television NN news NN shows NNS . SENT Our PP$ estimates NNS allow VVP us PP to TO answer VV such JJ questions NNS as IN Is VBZ the DT average JJ article NN in IN the DT New NP York NP Times NP more RBR liberal JJ than IN the DT average JJ speech NN by IN Tom NP Daschle NP . SENT or CC Is VBZ the DT average JJ story NN on IN Fox NP News NP more RBR conservative JJ than IN the DT average JJ speech NN by IN Bill NP Frist NP . SENT To TO compute VV our PP$ measure NN , , we PP count VVP the DT times NNS that IN a DT media NNS outlet NN cites VVZ various JJ think NN tanks NNS and CC other JJ policy NN groups NNS . SENT We PP compare VVP this DT with IN the DT times NNS that IN members NNS of IN Congress NP cite VVP the DT same JJ think VVP tanks NNS in IN their PP$ speeches NNS on IN the DT floor NN of IN the DT House NP and CC Senate NP . SENT By IN comparing VVG the DT citation NN patterns NNS we PP construct VV an DT ADA NP score NN . SENT As IN a DT simplified VVN example NN , , imagine VV that IN there EX were VBD only RB two CD think NN tanks NNS , , one CD liberal NN and CC one CD conservative NN . SENT Suppose VVP that IN the DT New NP York NP Times NP cited VVD the DT liberal NN think VVP tank NN twice RB as RB often RB as IN the DT conservative JJ one NN . SENT Our PP$ method NN asks VVZ . SENT What WP is VBZ the DT typical JJ ADA NP score NN of IN members NNS of IN Congress NP who WP exhibit VV the DT same JJ frequency NN 2 CD . SENT 1 CD in IN their PP$ speeches NNS . SENT This DT is VBZ the DT score NN that IN we PP would MD assign VV to TO the DT New NP York NP Times NP . SENT Our PP$ results NNS show VVP a DT strong JJ liberal JJ bias NN . SENT All DT of IN the DT news NN outlets NNS except IN Fox NP News NP Special JJ Report NP and CC the DT Washington NP Times NP received VVD a DT score NN to TO the DT left NN of IN the DT average JJ member NN of IN Congress NP . SENT Consistent JJ with IN many JJ conservative JJ critics NNS , , CBS NP Evening NP News NP and CC the DT New NP York NP Times NP received VVD a DT score NN far RB left VVD of IN center NN . SENT Outlets NNS such JJ as IN the DT Washington NP Post NP , , USA NP Today NP , , NPR NP s PP Morning NP Edition NN , , NBC NP s PP Nightly JJ News NP and CC ABC NP s PP World NP News NP Tonight NP were VBD moderately RB left VVN . SENT The DT most RBS centrist JJ outlets NNS but CC still RB left VVD leaning VVG by IN our PP$ measure NN were VBD the DT Newshour NP with IN Jim NP Lehrer NP , , CNN NP s PP NewsNight NP with IN Aaron NP Brown NP , , and CC ABC NP s PP Good NP Morning NP America NP . SENT Fox NP News NP Special JJ Report NP , , while IN right NN of IN center NN , , was VBD closer JJR to TO the DT center NN than IN any DT of IN the DT three CD major JJ networks NNS evening NN news NN broadcasts NNS . SENT All DT of IN our PP$ findings NNS refer VVP strictly RB to TO the DT news NN stories NNS of IN the DT outlets NNS . SENT That DT is VBZ , , we PP omitted VVD editorials NNS , , book NN reviews NNS , , and CC letters NNS to TO the DT editor NN from IN our PP$ sample NN . SENT A DT Measure NN of IN Media NP Bias NP The DT editors NNS in IN Los NP Angeles NP killed VVD the DT story NN . SENT They PP told VVD Witcover NP that IN it PP didn NN t NN come VVN off RP and CC that IN it PP was VBD an DT opinion NN story NN . SENT The DT solution NN was VBD simple JJ , , they PP told VVD him PP . SENT All RB he PP had VHD to TO do VV was VBD get VV other JJ people NNS to TO make VV the DT same JJ points NNS and CC draw VV the DT same JJ conclusions NNS and CC then RB write VV the DT article NN in IN their PP$ words NNS . SENT emphasis NN in IN original JJ Timothy NP Crouse NP , , Boys NNS on IN the DT Bus NN , , 1973 CD , , p NN . SENT 116 CD . SENT Do VV the DT major JJ media NN outlets NNS in IN the DT U NP . SENT S NP . SENT have VHP a DT liberal JJ bias NN . SENT Few JJ questions NNS evoke VVP stronger JJR opinions NNS , , and CC we PP cannot NN think VVP of IN a DT more RBR important JJ question NN to TO which WDT objective JJ statistical JJ techniques NNS can MD lend VV their PP$ service NN . SENT So RB far RB , , the DT debate NN has VHZ largely RB been VBN one CD of IN anecdotes NNS How WRB can MD CBS NP News NP be VB balanced VVN when WRB it PP calls VVZ Steve NP Forbes NP tax NN plan NN wacky JJ . SENT and CC untested JJ theories NNS if IN the DT news NN industry NN is VBZ a DT competitive JJ market NN , , then RB how WRB can MD media NN outlets NNS be VB systematically RB biased VVN . SENT . SENT Few JJ studies NNS provide VVP an DT objective JJ measure NN of IN the DT slant NN of IN news NN , , and CC none NN has VHZ provided VVN a DT way NN to TO link VV such PDT a DT measure NN to TO ideological JJ measures NNS of IN other JJ political JJ actors NNS . SENT That DT is VBZ , , none NN of IN the DT existing JJ measures NNS can MD say VV , , for IN example NN , , whether IN the DT New NP York NP Times NP is VBZ more RBR liberal JJ than IN Tom NP Daschle NP or CC whether IN Fox NP News NP is VBZ more RBR conservative JJ than IN Bill NP Frist NP . SENT We PP provide VVP such PDT a DT measure NN . SENT Namely RB , , we PP compute VV an DT ADA NP score NN for IN various JJ news NN outlets NNS , , including VVG the DT New NP York NP Times NP , , the DT Washington NP Post NP , , USA NP Today NP , , the DT Drudge NN Report NP , , Fox NP News NP Special JJ Report NP , , and CC all DT three CD networks NNS nightly JJ news NN shows NNS . SENT Our PP$ results NNS show VVP a DT strong JJ liberal JJ bias NN . SENT All DT of IN the DT news NN outlets NNS except IN Fox NP News NP Special JJ Report NP and CC the DT Washington NP Times NP received VVD a DT score NN to TO the DT left NN of IN the DT average JJ member NN of IN Congress NP . SENT And CC a DT few JJ outlets NNS , , including VVG the DT New NP York NP Times NP and CC CBS NP Evening NP News NP , , were VBD closer JJR to TO the DT average JJ Democrat NP in IN Congress NP than IN the DT center NN . SENT These DT findings NNS refer VVP strictly RB to TO the DT news NN stories NNS of IN the DT outlets NNS . SENT That DT is VBZ , , we PP omitted VVD editorials NNS , , book NN reviews NNS , , and CC letters NNS to TO the DT editor NN from IN our PP$ sample NN . SENT To TO compute VV our PP$ measure NN , , we PP count VVP the DT times NNS that IN a DT media NNS outlet NN cites VVZ various JJ think NN tanks NNS and CC other JJ policy NN groups NNS . SENT 1 LS We PP compare VV this DT with IN the DT times NNS that IN members NNS of IN Congress NP cite VVP the DT same JJ think VVP tanks NNS in IN their PP$ speeches NNS on IN the DT floor NN of IN the DT House NP and CC Senate NP . SENT By IN comparing VVG the DT citation NN patterns NNS we PP can MD construct VV an DT ADA NP score NN for IN each DT media NNS outlet NN . SENT As IN a DT simplified VVN example NN , , imagine VV that IN there EX were VBD only RB two CD think NN tanks NNS , , one CD liberal NN and CC one CD conservative NN . SENT Suppose VVP that IN the DT New NP York NP Times NP cited VVD the DT liberal NN think VVP tank NN twice RB as RB often RB as IN the DT conservative JJ one NN . SENT Our PP$ method NN asks VVZ . SENT What WP is VBZ the DT estimated JJ ADA NP score NN of IN a DT member NN of IN Congress NP who WP exhibits VVZ the DT same JJ frequency NN 2 CD . SENT 1 CD in IN his PP$ or CC her PP$ speeches NNS . SENT This DT is VBZ the DT score NN that IN our PP$ method NN would MD assign VV the DT New NP York NP Times NP . SENT A DT feature NN of IN our PP$ method NN is VBZ that IN it PP does VVZ not RB require VV us PP to TO make VV a DT subjective JJ assessment NN of IN how WRB liberal JJ or CC conservative JJ a DT think VVP tank NN is VBZ . SENT That DT is VBZ , , for IN instance NN , , we PP do VVP we PP need VVP to TO read VV policy NN reports NNS of IN the DT think VVP tank NN or CC analyze VV its PP$ position NN on IN various JJ issues NNS to TO determine VV its PP$ ideology NN . SENT Instead RB , , we PP simply RB observe VVP the DT ADA NP scores NNS of IN the DT members NNS of IN Congress NP who WP cite VVP the DT think VVP tank NN . SENT This DT feature NN is VBZ important JJ , , since IN an DT active JJ controversy NN exists VVZ whether IN , , e NN . SENT g NN . SENT , , the DT Brookings NP Institution NP or CC the DT RAND NN Corporation NP is VBZ moderate JJ , , left JJ wing NN , , or CC right JJ wing NN . SENT Some DT Previous JJ Studies NNS of IN Media NP Bias NP Survey NP research NN has VHZ shown VVN that IN an DT almost RB overwhelming JJ fraction NN of IN journalists NNS are VBP liberal JJ . SENT For IN instance NN , , Elaine NP Povich NP 1996 CD reports NNS that IN only RB seven CD percent NN of IN all DT Washington NP correspondents NNS voted VVD for IN George NP H NP . SENT W NP . SENT Bush NP in IN 1992 CD , , compared VVN to TO 37 CD percent NN of IN the DT American JJ public NN . SENT 2 CD Lichter NP , , Rothman NP and CC Lichter NP , , 1986 CD and CC Weaver NP and CC Wilhoit NP 1996 CD report NN similar JJ findings NNS for IN earlier JJR elections NNS . SENT More RBR recently RB , , the DT New NP York NP Times NP reported VVD that IN only RB eight CD percent NN of IN Washington NP correspondents NNS thought VVD George NP W NP . SENT Bush NP would MD be VB a DT better JJR president NN than IN John NP Kerry NP . SENT 3 LS This DT compares VVZ to TO 51 CD of IN all DT American JJ voters NNS . SENT David NP Brooks NP notes VVZ that IN for IN every DT journalist NN who WP contributed VVD to TO George NP W NP . SENT Bush NP s PP campaign NN , , 93 CD contributed VVD to TO Kerry NP s PP . SENT 4 LS These DT statistics NNS suggest VVP that IN journalists NNS , , as IN a DT group NN , , are VBP more RBR liberal JJ than IN almost RB any DT congressional JJ district NN in IN the DT country NN . SENT For IN instance NN , , in IN the DT Ninth JJ California NP district NN , , which WDT includes VVZ Berkeley NP , , twelve NN percent NN voted VVD for IN Bush NP in IN 1992 CD , , nearly RB double VV the DT rate NN of IN journalists NNS . SENT In IN the DT Eighth JJ Massachusetts NP district NN , , which WDT includes VVZ Cambridge NP , , nineteen CD percent NN voted VVD for IN Bush NP , , approximately RB triple RB the DT rate NN of IN journalists NNS . SENT 5 CD Of IN course NN , , however RB , , just RB because IN a DT journalist NN has VHZ liberal JJ or CC conservative JJ views NNS , , this DT does VVZ not RB mean VV that IN his PP$ or CC her PP$ reporting NN will MD be VB slanted VVN . SENT For IN instance NN , , as IN Kathleen NP Hall NP Jamieson NP 2000 CD , , 188 CD notes NNS , , One PP might MD hypothesize VV instead RB that WDT reporters NNS respond VV to TO the DT cues NNS of IN those DT who WP pay VVP their PP$ salaries NNS and CC mask NN their PP$ own JJ ideological JJ dispositions NNS . SENT Another DT explanation NN would MD hold VV that IN norms NNS of IN journalism NN , , including VVG objectivity NN and CC balance VV blunt VV whatever WDT biases NNS exist VVP . SENT Or CC , , as IN Timothy NP Crouse NP explains VVZ . SENT It PP is VBZ an DT unwritten JJ law NN of IN current JJ political JJ journalism NN that IN conservative JJ Republican JJ Presidential JJ candidates NNS usually RB receive VVP gentler JJR treatment NN from IN the DT press NN than IN do VV liberal JJ Democrats NNS . SENT Since IN most JJS reporters NNS are VBP moderate JJ or CC liberal JJ Democrats NNS themselves PP , , they PP try VVP to TO offset VV their PP$ natural JJ biases NNS by IN going VVG out RP of IN their PP$ way NN to TO be VB fair JJ to TO conservatives NNS . SENT No DT candidate NN ever RB had VHD a DT more RBR considerate JJ press NN corps NN than IN Barry NP Goldwater NP in IN 1964 CD , , and CC four CD years NNS later RBR the DT campaign NN press NN gave VVD every DT possible JJ break NN to TO Richard NP Nixon NP . SENT Reporters NNS sense VVP a DT social JJ barrier NN between IN themselves PP and CC most RBS conservative JJ candidates NNS . SENT their PP$ relations NNS are VBP formal JJ and CC meticulously RB polite JJ . SENT But CC reporters NNS tend VVP to TO loosen VV up RP around IN liberal JJ candidates NNS and CC campaign NN staffs NNS . SENT since IN they PP share VV the DT same JJ ideology NN , , they PP can MD joke VV with IN the DT staffers NNS , , even RB needle NN them PP , , without IN being VBG branded VVN the DT enemy NN . SENT If IN a DT reporter NN has VHZ been VBN trained VVN in IN the DT traditional JJ , , objective JJ school NN of IN journalism NN , , this DT ideological JJ and CC social JJ closeness NN to TO the DT candidate NN and CC the DT staff NN makes VVZ him PP feel VVP guilty JJ . SENT he PP begins VVZ to TO compensate VV . SENT the DT more RBR he PP likes VVZ and CC agrees VVZ with IN the DT candidate NN personally RB , , the DT harder RBR he PP judges NNS him PP professionally RB . SENT Like IN a DT coach NN sizing VVG up RP his PP$ own JJ son NN in IN spring NN tryouts NNS , , the DT reporter NN becomes VVZ doubly RB severe JJ . SENT 1973 CD , , 355 CD 6 CD However RB , , a DT strong JJ form NN of IN the DT view NN that IN reporters NNS offset VV or CC blunt VV their PP$ own JJ ideological JJ biases NNS leads VVZ to TO a DT counterfactual JJ implication NN . SENT Suppose VVP it PP is VBZ true JJ that IN all DT reporters NNS report VVP objectively RB , , and CC their PP$ ideological JJ views NNS do VVP not RB color VV their PP$ reporting NN . SENT If IN so RB , , then RB all DT news NN would MD have VH the DT same JJ slant NN . SENT Moreover RB , , if IN one PP believes VVZ Crouse NP s PP claim VVP that WDT reporters NNS overcompensate VV in IN relation NN to TO their PP$ own JJ ideology NN , , then RB a DT news NN outlet NN filled VVN with IN conservatives NNS , , such JJ as IN Fox NP News NP , , should MD have VH a DT more RBR liberal JJ slant NN than IN a DT news NN outlet NN filled VVN with IN liberals NNS , , such JJ as IN the DT New NP York NP Times NP . SENT Spatial JJ models NNS of IN firm NN location NN , , such JJ as IN those DT by IN Hotelling NP 1929 CD or CC Mullainathan NP and CC Shleifer NP 2003 CD give VV theoretical JJ reasons NNS why WRB the DT media NNS should MD slant VV the DT news NN exactly RB as IN consumers NNS desire VVP . SENT 6 CD The DT idea NN is VBZ that IN if IN the DT media NNS did VVD not RB , , then RB an DT entrepreneur NN could MD form VV a DT new JJ outlet NN that WDT does VVZ , , and CC he PP or CC she PP could MD earn VV greater JJR than IN equilibrium NN profits NNS , , possibly RB even RB driving VVG the DT other JJ outlets NNS out IN of IN business NN . SENT This DT is VBZ a DT compelling JJ argument NN , , and CC even RB the DT libertarian JJ Cato NP Journal NP has VHZ published VVN an DT article NN agreeing VVG with IN the DT view NN . SENT In IN this DT article NN , , the DT author NN , , Daniel NP Sutter NP 2001 CD , , notes VVZ that IN Charges NNS of IN a DT liberal JJ bias NN essentially RB require VV the DT existence NN of IN a DT cartel NN 431 CD . SENT However RB , , contrary JJ to TO the DT prediction NN of IN the DT typical JJ firm NN location NN model NN , , we PP find VVP a DT a DT systematic JJ liberal JJ bias NN of IN the DT U NP . SENT S NP . SENT media NNS . SENT This DT is VBZ echoed VVN by IN three CD other JJ studies NNS Hamilton NP 2004 CD , , Lott NP and CC Hasset NP 2004 CD , , and CC Sutter NP 2004 CD , , the DT only JJ empirical JJ studies NNS of IN media NNS bias NN by IN economists NNS of IN which WDT we PP are VBP aware JJ . SENT Although IN his PP$ primary JJ focus NN is VBZ not RB on IN media NNS bias NN , , in IN one CD section NN of IN his PP$ book NN , , James NP Hamilton NP 2004 CD analyzes VVZ Pew NP Center NP surveys NNS of IN media NNS bias NN . SENT The DT surveys NNS show VVP unsurprsingly RB that IN conservatives NNS tend VVP to TO believe VV that IN there EX is VBZ a DT liberal JJ bias NN in IN the DT media NNS , , while IN liberals NNS tend VVP to TO believe VV there EX is VBZ a DT conservative JJ bias NN . SENT While IN many JJ would MD simply RB conclude VV that IN this DT is VBZ only RB evidence NN that IN bias NN is VBZ in IN the DT eyes NNS of IN the DT beholder NN , , Hamilton NP makes VVZ the DT astute JJ point NN that IN that DT individuals NNS are VBP more RBR likely JJ to TO perceive VV bias NN the DT further RBR the DT slant NN of IN the DT news NN is VBZ from IN their PP$ own JJ position NN . SENT Since IN the DT same JJ surveys NNS also RB show VVP that DT conservatives NNS tend VVP to TO see VV a DT bias NN more RBR than IN liberals NNS do VVP , , this DT is VBZ evidence NN that IN the DT news NN slants VVZ more RBR to TO the DT left NN . SENT John NP Lott NP and CC Kevin NP Hassett NP 2004 CD propose VV an DT innovative JJ test NN for IN media NNS bias NN . SENT They PP record VV whether IN the DT headlines NNS of IN various JJ economic JJ news NN stories NNS are VBP positive JJ or CC negative JJ . SENT For IN instance NN , , on IN the DT day NN that IN the DT Commerce NP Department NP reports VVZ that IN GDP NP grows VVZ by IN a DT large JJ degree NN , , a DT newspaper NN could MD instead RB report VV GDP NP Growth NP Less RBR than IN Expected VVN . SENT Lott NP and CC Hasset NP control NN for IN the DT actual JJ economic JJ figures NNS reported VVN by IN the DT Commerce NP Department NP , , and CC they PP include VVP an DT independent JJ variable NN that WDT indicates VVZ the DT political JJ party NN of IN the DT president NN . SENT Of IN the DT ten CD major JJ newspapers NNS that IN they PP examine VVP , , they PP find VVP that IN nine CD are VBP more RBR likely JJ to TO report VV a DT negative JJ headline NN if IN the DT president NN is VBZ Republican JJ . SENT 7 CD Daniel NP Sutter NP 2004 CD collects VVZ data NNS on IN the DT geographic JJ locations NNS of IN readers NNS of IN Time NP , , Newsweek NP , , and CC U NP . SENT S NP . SENT News NP and CC World NP Report NP . SENT He PP shows VVZ that IN as IN a DT region NN becomes VVZ more RBR liberal JJ as IN indicated VVN by IN its PP$ vote NN share NN for IN President NP Clinton NP , , its PP$ consumption NN of IN the DT three CD major JJ national JJ news NN magazines NNS increases NNS . SENT With IN a DT clever JJ and CC readers NNS that IN we PP are VBP ideologues NNS . SENT It PP is VBZ an DT exercise NN of IN , , in IN disinformation NN , , of IN alarming JJ proportions NNS . SENT This DT attempt NN to TO convince VV the DT audience NN of IN the DT world NN s VVZ most JJS ideology NN free JJ newspapers NNS that IN they PP re IN being VBG subjected VVN to TO agenda NN driven JJ news NN reflecting VVG a DT liberal JJ bias NN . SENT I PP don VV t NN believe VVP our PP$ viewers NNS and CC readers NNS will MD be VB , , in IN the DT long JJ run NN , , misled VVN by IN those DT who WP advocate VVP biased VVN journalism NN . SENT 8 CD when WRB it PP comes VVZ to TO free JJ publicity NN , , some DT of IN the DT major JJ broadcast NN media NNS are VBP simply RB biased VVN in IN favor NN of IN the DT Republicans NPS , , while IN the DT rest NN tend VVP to TO blur VV differences NNS between IN the DT parties NNS . SENT But CC that IN s PP the DT way NN it PP is VBZ . SENT Democrats NPS should MD complain VV as RB loudly RB about IN the DT real JJ conservative JJ bias NN of IN the DT media NNS as IN the DT Republicans NPS complain VVP about IN its PP$ entirely RB mythical JJ bias NN 9 CD The DT mainstream NN media NN does VVZ not RB have VH a DT liberal JJ bias NN . SENT . SENT . SENT . SENT ABC NP , , CBS NP , , NBC NP , , CNN NP , , the DT New NP York NP Times NP , , The DT Washington NP Post NP , , Time NP , , Newsweek NP and CC the DT rest NN at IN least JJS try VV to TO be VB fair JJ . SENT 10 CD I'm NN going VVG out IN telling VVG the DT story NN that IN I PP think VVP is VBZ the DT biggest JJS story NN of IN our PP$ time NN . SENT how WRB the DT right JJ wing NN media NNS has VHZ become VVN a DT partisan JJ propaganda NN arm NN of IN the DT Republican NP National NP Committee NP . SENT We PP have VHP an DT ideological JJ press NN that's NNS interested VVN in IN the DT election NN of IN Republicans NNS , , and CC a DT mainstream JJ press NN that's NNS interested VVN in IN the DT bottom JJ line NN . SENT Therefore RB , , we PP don't VVD have VH a DT vigilant JJ , , independent JJ press NN whose WP$ interest NN is VBZ the DT American JJ people NNS . SENT 11 CD Data NNS The DT web NN site NN , , www NP . SENT wheretodoresearch NN . SENT com NN lists VVZ 200 CD of IN the DT most RBS prominent JJ think NN tanks NNS and CC policy NN groups NNS in IN the DT U NP . SENT S NP . SENT Using VVG the DT official JJ web NN site NN of IN Congress NP , , http NN . SENT thomas NP . SENT loc NP . SENT gov NP , , we PP and CC our PP$ research NN assistants NNS searched VVD the DT Congressional NP Record NP for IN instances NNS where WRB a DT member NN of IN Congress NP cited VVD one CD of IN these DT think VVP tanks NNS . SENT We PP also RB recorded VVD the DT average NN adjusted VVN ADA NP score NN of IN the DT member NN who WP cited VVD the DT think NN tank NN . SENT We PP use VVP adjusted JJ scores NNS , , constructed VVN by IN Groseclose NP , , Levitt NP , , and CC Snyder NP 1999 CD , , because IN we PP need VVP the DT scores NNS to TO be VB comparable JJ across IN time NN and CC chambers NNS . SENT 12 CD Groseclose NP , , Levitt NP , , and CC Snyder NP 1999 CD use NN the DT 1980 CD House NP scale NN as IN their PP$ base NN year NN and CC chamber NN . SENT It PP is VBZ convenient JJ for IN us PP to TO choose VV a DT scale NN that WDT gives VVZ centrist JJ members NNS of IN Congress NP a DT score NN of IN about IN 50 CD . SENT For IN this DT reason NN we PP converted VVD scores NNS to TO the DT 1999 CD House NP scale NN . SENT 13 CD Along IN with IN direct JJ quotes NNS of IN think VVP tanks NNS , , we PP sometimes RB included VVD sentences NNS that WDT were VBD not RB direct JJ quotes NNS . SENT For IN instance NN , , many JJ of IN the DT citations NNS were VBD cases NNS where WRB a DT member NN of IN Congress NP noted VVD This DT bill NN is VBZ supported VVN by IN think NN tank NN X NN . SENT Also RB , , members NNS of IN Congress NP sometimes RB insert VV printed JJ material NN into IN the DT Congressional NP Record NP , , such JJ as IN a DT letter NN , , a DT newspaper NN article NN , , or CC a DT report NN . SENT If IN a DT think NN tank NN was VBD cited VVN in IN such JJ material NN or CC if IN a DT think VVP tank NN member NN wrote VVD the DT material NN , , we PP treated VVD it PP just RB as RB if IN the DT member NN of IN Congress NP had VHD read VVN the DT material NN in IN his PP$ or CC her PP$ speech NN . SENT We PP did VVD the DT same JJ exercise NN for IN stories NNS that IN media NNS outlets NNS report NN , , except IN with IN media NN outlets NNS we PP did VVD not RB record VV an DT ADA NP score NN . SENT Instead RB , , our PP$ method NN estimates VVZ such PDT a DT score NN . SENT Sometimes RB a DT legislator NN or CC journalist NN noted VVD an DT action NN that IN a DT think NN tank NN had VHD taken VVN e NN . SENT g NN . SENT that IN it PP raised VVD a DT certain JJ amount NN of IN money NN , , initiated VVD a DT boycott NN , , filed VVD a DT lawsuit NN , , elected VVN new JJ officers NNS , , or CC held VVD its PP$ annual JJ convention NN . SENT We PP did VVD not RB record VV such JJ cases NNS in IN our PP$ data NNS set VVD . SENT However RB , , sometimes RB in IN the DT process NN of IN describing VVG such JJ actions NNS , , the DT journalist NN or CC legislator NN would MD quote VV a DT member NN of IN the DT think NN tank NN , , and CC the DT quote NN revealed VVD the DT think NN tank NN s PP views VVZ on IN national JJ policy NN , , or CC the DT quote NN stated VVD a DT fact NN that WDT is VBZ relevant JJ to TO national JJ policy NN . SENT If IN so RB , , we PP would MD record VV that DT quote NN in IN our PP$ data NNS set VVD . SENT For IN instance NN , , suppose VVP a DT reporter NN noted VVD The DT NAACP NP has VHZ asked VVN its PP$ members NNS to TO boycott VV businesses NNS in IN the DT state NN of IN South NP Carolina NP . SENT We PP are VBP initiating VVG this DT boycott NN , , because IN we PP believe VVP that IN it PP is VBZ racist JJ to TO fly VV the DT Confederate JJ Flag NN on IN the DT state NN capitol NP , , a DT leader NN of IN the DT group NN noted VVD . SENT In IN this DT instance NN , , we PP would MD count VV the DT second JJ sentence NN that IN the DT reporter NN wrote VVD , , but CC not RB the DT first JJ . SENT Also RB , , we PP omitted VVD the DT instances NNS where WRB the DT member NN of IN Congress NP or CC journalist NN only RB cited VVD the DT think NN tank NN so IN he PP or CC she PP could MD criticize VV it PP or CC explain VV why WRB it PP was VBD wrong JJ . SENT About IN five CD percent NN of IN the DT congressional JJ citations NNS and CC about RB one CD percent NN of IN the DT media NNS citations NNS fell VVD into IN this DT category NN . SENT In IN the DT same JJ spirit NN , , we PP omitted VVD cases NNS where WRB a DT journalist NN or CC legislator NN gave VVD an DT ideological JJ label NN to TO a DT think NN tank NN e NN . SENT g NN . SENT Even RB the DT conservative JJ Heritage NP Foundation NP favors VVZ this DT bill NN . SENT . SENT The DT idea NN is VBZ that IN we PP only RB wanted VVD cases NNS were VBD the DT legislator NN or CC journalist NN cited VVD the DT think NN tank NN as IN if IN it PP were VBD a DT disinterested JJ expert NN on IN the DT topic NN at IN hand NN . SENT About IN two CD percent NN of IN the DT congressional JJ citations NNS and CC about RB five CD percent NN of IN the DT media NNS citations NNS involved VVD an DT ideological JJ label NN . SENT 14 CD For IN the DT congressional JJ data NNS , , we PP coded VVD all DT citations NNS that WDT occurred VVD during IN the DT period NN Jan NP . SENT 1 CD , , 1993 CD to TO December NP 31 CD , , 2002 CD . SENT This DT covered VVD the DT 103 CD rd NN thru IN 107 CD th NN Congresses NNS . SENT We PP used VVD the DT period NN 1993 CD to TO 1999 CD to TO calculate VV the DT average NN adjusted VVN ADA NP score NN for IN members NNS of IN Congress NP . SENT 15 CD As RB noted VVD earlier RBR , , our PP$ media NNS data NN does VVZ not RB include VV editorials NNS , , letters NNS to TO the DT editor NN , , or CC book NN reviews NNS . SENT That DT is VBZ , , all DT of IN our PP$ results NNS refer VVP only RB to TO the DT bias NN of IN the DT news NN of IN media NNS . SENT There EX are VBP several JJ reasons NNS why WRB we PP do VVP not RB include VV editorials NNS . SENT The DT primary JJ one NN is VBZ that IN there EX is VBZ little JJ controversy NN over IN the DT slant NN of IN editorial JJ pages NNS e NN . SENT g NN . SENT few JJ would MD disagree VV that IN Wall NP Street NP Journal NP editorials NNS are VBP conservative JJ , , while IN New NP York NP Times NP editorials NNS are VBP liberal JJ . SENT However RB , , there EX is VBZ a DT very RB large JJ controversy NN about IN the DT slant NN of IN the DT news NN of IN various JJ media NNS outlets NNS . SENT A DT second JJ reason NN involves VVZ the DT effect NN if IN any DT that IN the DT media NNS have VHP on IN individuals NNS political JJ views NNS . SENT It PP is VBZ reasonable JJ to TO believe VV that IN a DT biased VVN outlet NN that WDT pretends VVZ to TO be VB centrist JJ has VHZ more JJR of IN an DT effect NN on IN readers NNS or CC viewers NNS beliefs NNS than IN , , say VV , , an DT editorial NN page NN that WDT does VVZ not RB pretend VV to TO be VB centrist JJ . SENT Because IN of IN this DT , , we PP believe VVP it PP is VBZ more RBR important JJ to TO examine VV the DT news NN than IN editorials NNS . SENT A DT third JJ reason NN involves VVZ difficulties NNS in IN coding VVG the DT data NNS . SENT Editorial NN and CC opinion NN writers NNS , , much RB more JJR than IN news NN writers NNS , , are VBP sometimes RB sarcastic JJ when WRB they PP quote VVP members NNS of IN think VVP tanks NNS . SENT If IN our PP$ coders NNS do VVP not RB catch VV the DT sarcasm NN , , they PP record VV the DT citation NN as IN a DT favorable JJ one CD . SENT This DT biases NNS the DT results NNS toward IN making VVG the DT editorials NNS appear VVP more RBR centrist JJ than IN they PP really RB are VBP . SENT In IN Table NN 1 CD we PP list VVP the DT 50 CD groups NNS from IN our PP$ list NN that WDT were VBD most RBS commonly RB cited VVN by IN the DT media NNS . SENT The DT first JJ column NN lists VVZ the DT average JJ ADA NP score NN of IN the DT legislator NN citing VVG the DT think VVP tank NN . SENT These DT averages VVZ closely RB correspond VV to TO conventional JJ wisdom NN about IN the DT ideological JJ positions NNS of IN the DT groups NNS . SENT For IN instance NN , , the DT Heritage NP Foundation NP and CC Christian NP Coalition NP , , with IN average JJ scores NNS of IN 20 CD . SENT 0 CD and CC 22 CD . SENT 6 CD , , are VBP near IN the DT conservative JJ end NN . SENT the DT Economic NP Policy NP Institute NP and CC the DT Children NP s PP Defense NP Fund NP 80 CD . SENT 3 CD and CC 82 CD . SENT 0 CD are NN near IN the DT liberal JJ end NN . SENT and CC the DT Brookings NP Institution NP and CC the DT World NP Wildlife NP Fund NP 53 CD . SENT 3 CD and CC 50 CD . SENT 4 CD are NN in IN the DT middle NN of IN our PP$ mix NN of IN think NN tanks NNS . SENT While IN most JJS of IN these DT averages NNS closely RB agree VVP with IN the DT conventional JJ wisdom NN , , two CD cases NNS seem VVP somewhat RB anomalous JJ . SENT The DT first JJ is VBZ the DT ACLU NP . SENT The DT average JJ score NN of IN legislators NNS citing VVG it PP was VBD 49 CD . SENT 8 CD . SENT Later RBR , , we PP shall MD provide VV reasons NNS why WRB it PP makes VVZ sense NN to TO define VV the DT political JJ center NN at IN 50 CD . SENT 1 CD . SENT This DT suggests VVZ that IN the DT ACLU NP , , if IN anything NN is VBZ a DT right JJ leaning VVG organization NN . SENT The DT reason NN the DT ACLU NP has VHZ such PDT a DT low JJ score NN is VBZ that IN it PP opposed VVD the DT McCain NP Feingold NP Campaign NP Finance NP bill NN , , and CC conservatives NNS in IN Congress NP cited VVD this DT often RB . SENT In IN fact NN , , slightly RB more JJR than IN one CD eight CD of IN all DT ACLU NP citations NNS in IN Congress NP were VBD due JJ to TO one CD person NN alone RB , , Mitch NP McConnell NP R NP . SENT Kt NN . SENT , , perhaps RB the DT chief JJ critic NN of IN McCain NP Feingold NP . SENT If IN we PP omit VVP McConnell NP s PP citations NNS , , the DT ACLU NP s PP average JJ score NN increases NNS to TO 55 CD . SENT 9 CD . SENT Because IN of IN this DT anomaly NN , , in IN the DT Appendix NN we PP report VVP the DT results NNS when WRB we PP repeat VV all DT of IN our PP$ analyses NNS but CC omit VV the DT ACLU NP data NNS . SENT The DT second JJ apparent JJ anomaly NN is VBZ the DT RAND NN Corporation NP , , which WDT has VHZ a DT fairly RB liberal JJ average JJ score NN , , 60 CD . SENT 4 LS . SENT We PP mentioned VVD this DT finding NN to TO some DT employees NNS of IN RAND NNS , , who WP told VVD us PP they PP were VBD not RB surprised VVN . SENT While IN RAND NN strives VVZ to TO be VB middle JJ of IN the DT road NN ideologically RB , , the DT more JJR conservative JJ scholars NNS at IN RAND NNS tend VVP to TO work VV on IN military JJ studies NNS , , while IN the DT more JJR liberal JJ scholars NNS tend VVP to TO work VV on IN domestic JJ studies NNS . SENT Because IN the DT military JJ studies NNS are VBP sometimes RB classified VVN and CC often RB more RBR technocratic JJ than IN the DT domestic JJ studies NNS , , the DT media NNS and CC members NNS of IN Congress NP tend VVP to TO cite VV the DT domestic JJ studies NNS disproportionately RB . SENT As IN a DT consequence NN , , RAND NN appears VVZ liberal JJ when WRB judged VVN by IN these DT citations NNS . SENT It PP is VBZ important JJ to TO note VV that IN this DT fact NN that IN the DT research NN at IN RAND NNS is VBZ more RBR conservative JJ than IN the DT numbers NNS in IN Table NN 1 CD suggest VVP will MD not RB bias VV our PP$ results NNS . SENT To TO see VV this DT , , think VVP of IN RAND NNS as IN two CD think VVP tanks NNS . SENT RAND NNS I PP , , the DT left JJ leaning VVG think NN tank NN which WDT produces VVZ the DT research NN that IN the DT media NNS and CC members NNS of IN Congress NP tend VVP to TO cite VV , , and CC RAND NNS II NP , , the DT conservative JJ think NN tank NN which WDT produces VVZ the DT research NN that IN they PP tend VVP not RB to TO cite VV . SENT Our PP$ results NNS exclude VVP RAND NN II NP from IN the DT analysis NN . SENT This DT causes VVZ no DT more JJR bias NN than IN excluding VVG any DT other JJ think NN tank NN that WDT is VBZ rarely RB cited VVN in IN Congress NP or CC the DT media NNS . SENT The DT second JJ and CC third JJ columns NNS respectively RB report VVP the DT number NN of IN congressional JJ and CC media NNS citations NNS in IN our PP$ data NNS . SENT These DT columns NNS give VVP some DT preliminary JJ evidence NN that IN the DT media NNS is VBZ liberal JJ , , relative JJ to TO Congress NP . SENT To TO see VV this DT , , define VV as IN right JJ wing NN a DT think VVP tank NN that WDT has VHZ an DT average JJ score NN below IN 40 CD . SENT Next JJ , , consider VV the DT ten CD most JJS cited VVD think VV tanks NNS by IN the DT media NNS . SENT Only RB one CD right NN wing NN think VVP tank NN makes VVZ this DT list NN , , American NP Enterprise NP Institute NP . SENT In IN contrast NN , , consider VVP the DT ten CD most JJS cited VVD think VV tanks NNS by IN Congress NP . SENT These DT are VBP the DT National NP Taxpayers NPS Union NP , , AARP NP , , Amnesty NN International NP , , Sierra NP Club NP , , Heritage NP Foundation NP , , Citizens NP Against NP Government NP Waste NP , , RAND NN , , Brookings NP , , NFIB NP , , and CC ACLU NP . SENT Four CD of IN these DT are VBP right JJ wing NN . SENT For IN perspective NN , , in IN Table NN 2 CD we PP list VVP the DT average NN adjusted VVN ADA NP score NN of IN some DT prominent JJ members NNS of IN Congress NP , , including VVG some DT well RB known VVN moderates NNS . SENT These DT include VVP the DT most RBS conservative JJ Democrat NP in IN our PP$ sample NN , , Nathan NP Deal NP Ga NP . SENT , , and CC the DT most RBS liberal JJ Republican NP in IN our PP$ sample NN , , Constance NP Morella NP Md NP . SENT . SENT Although IN Nathan NP Deal NP became VVD a DT Republican JJ in IN 1995 CD , , the DT score NN that IN we PP list VVP in IN the DT table NN is VBZ calculated VVN only RB from IN his PP$ years NNS as IN a DT Democrat NP . SENT 16 CD The DT table NN also RB lists VVZ the DT average JJ scores NNS of IN the DT entire JJ House NP and CC Senate NP , , as RB well RB as IN the DT average JJ scores NNS of IN the DT Republican JJ and CC Democratic JJ parties NNS . SENT 17 CD To TO calculate VV average JJ scores NNS , , for IN each DT member NN we PP note VVP all DT of IN his PP$ or CC her PP$ scores NNS for IN the DT seven CD year NN period NN for IN which WDT we PP recorded VVD adjusted VVN scores NNS 1993 CD 1999 CD . SENT Then RB we PP calculated VVD the DT average NN over IN these DT years NNS . SENT Because RB , , at IN times NNS , , there EX is VBZ some DT subjectivity NN in IN coding VVG our PP$ data NNS , , when WRB we PP hired VVD our PP$ research NN assistants NNS we PP asked VVD i NP for IN whom WP they PP voted VVD or CC would MD have VH voted VVD if IN they PP were VBD limited VVN to TO choosing VVG only JJ Al NP Gore NP and CC George NP Bush NP . SENT We PP chose VVD research NN assistants NNS so RB that IN approximately RB half PDT our PP$ data NNS was VBD coded VVN by IN Gore NP supporters NNS and CC half NN by IN Bush NP supporters NNS . SENT For IN each DT media NNS outlet NN we PP selected VVD an DT observation NN period NN that IN we PP estimated VVD would MD yield VV at IN least JJS 300 CD observations NNS citations NNS . SENT Because IN magazines NNS , , television NN shows NNS , , and CC radio NN shows NNS produce VVP less JJR data NNS per IN show NN or CC issue NN e NN . SENT g NN . SENT a DT transcript NN for IN a DT 30 CD minute NN television NN show NN contains VVZ only RB a DT small JJ fraction NN of IN the DT sentences NNS that WDT are VBP contained VVN in IN a DT daily JJ newspaper NN , , with IN some DT outlets NNS we PP began VVD with IN the DT earliest JJS date NN available JJ in IN Lexis NP Nexis NP . SENT We PP did VVD this DT for IN . SENT i NP the DT three CD magazines NNS that IN we PP analyze VVP , , ii VVG the DT five CD evening NN television NN news NN broadcasts NNS that IN we PP analyze VVP . SENT and CC iii NNS the DT one CD radio NN program NN that IN we PP analyze VVP . SENT 18 CD Our NP Definition NN of IN Bias NP Before IN proceeding NN , , it PP is VBZ useful JJ to TO clarify VV our PP$ definition NN of IN bias NN . SENT Most RBS important JJ , , the DT definition NN has VHZ nothing NN to TO do VV with IN the DT honesty NN or CC accuracy NN of IN the DT news NN outlet NN . SENT Instead RB , , our PP$ notion NN is VBZ more JJR like IN a DT taste NN or CC preference NN . SENT For IN instance NN , , we PP estimate VVP that IN the DT centrist JJ U NP . SENT S NP . SENT voter NN during IN the DT late JJ 1990 CD s PP had VHD a DT left JJ right NN ideology NN approximately RB equal JJ to TO that DT of IN Arlen NP Specter NP R NP Pa NP . SENT or CC Sam NP Nunn NP D NP Ga NP . SENT . SENT Meanwhile RB , , we PP estimate VVP that IN the DT average JJ New NP York NP Times NP article NN is VBZ ideologically RB very RB similar JJ to TO the DT average JJ speech NN by IN Joe NP Lieberman NP D NP Ct NP . SENT . SENT Next RB , , since IN vote NN scores NNS show VVP Lieberman NP to TO be VB more RBR liberal JJ than IN Specter NP or CC Nunn NP , , our PP$ method NN concludes VVZ that IN the DT New NP York NP Times NP has VHZ a DT liberal JJ bias NN . SENT However RB , , in IN no DT way NN does VVZ this DT imply VVP that IN the DT New NP York NP Times NP is VBZ inaccurate JJ or CC dishonest JJ just RB as IN the DT vote NN scores NNS do VVP not RB imply VV that IN Joe NP Lieberman NP is VBZ any DT less RBR honest JJ than IN Sam NP Nunn NP or CC Arlen NP Specter NP . SENT In IN contrast NN , , other JJ writers NNS , , at IN least JJS at IN times NNS , , do VVP define VV bias NN as IN a DT matter NN of IN accuracy NN or CC honesty NN . SENT We PP emphasize VVP that IN our PP$ differences NNS with IN such JJ writers NNS are VBP ones NNS of IN semantics NNS , , not RB substance NN . SENT If IN , , say VV , , a DT reader NN insists VVZ that DT bias NN should MD refer VV to TO accuracy NN or CC honesty NN , , then RB we PP urge VVP him PP or CC her PP simply RB to TO substitute VV another DT word NN wherever WRB we PP write VVP bias NN . SENT Perhaps RB slant NN is VBZ a DT good JJ alternative NN . SENT However RB , , at IN the DT same JJ time NN , , we PP argue VVP that IN our PP$ notion NN of IN bias NN is VBZ meaningful JJ and CC relevant JJ , , and CC perhaps RB more RBR meaningful JJ and CC relevant JJ than IN the DT alternative JJ notion NN . SENT The DT main JJ reason NN , , we PP believe VVP , , is VBZ that IN only RB seldom RB do VVP journalists NNS make VV dishonest JJ statements NNS . SENT Cases NNS such JJ as IN Jayson NP Blair NP , , Stephen NP Glass NP , , or CC the DT falsified VVN memo NN at IN CBS NP are VBP rare JJ . SENT they PP make VVP headlines NNS when WRB they PP do VVP occur VV . SENT and CC much RB of IN the DT time NN they PP are VBP orthogonal JJ to TO any DT political JJ bias NN . SENT Instead RB , , for IN every DT sin NN of IN commission NN , , such JJ as IN those DT by IN Glass NP or CC Blair NP , , we PP believe VVP that IN there EX are VBP hundreds NNS , , and CC maybe RB thousands NNS , , of IN sins NNS of IN omission NN cases NNS where WRB a DT journalist NN chose VVD facts NNS or CC stories NNS that IN only RB one CD side NN of IN the DT political JJ spectrum NN is VBZ likely JJ to TO mention VV . SENT For IN instance NN , , in IN a DT story NN printed VVN on IN March NP 1 CD , , 2002 CD , , the DT New NP York NP Times NP reported VVD that IN i NP the DT IRS NP increased VVD its PP$ audit NN rate NN on IN the DT working VVG poor JJ a DT phrase NN that IN the DT article NN defines VVZ as IN any DT taxpayer NN who WP claimed VVD an DT earned VVN income NN tax NN credit NN . SENT while IN ii NNS the DT agency NN decreased VVD its PP$ audit NN rate NN on IN taxpayers NNS who WP earn VVP more JJR than IN 100 CD , , 000 CD . SENT and CC iii NNS more JJR than IN half NN of IN all DT IRS NP audits NN involve VVP the DT working VVG poor JJ . SENT The DT article NN also RB notes VVZ that IN iv JJS The DT roughly RB 5 CD percent NN of IN taxpayers NNS who WP make VVP more JJR than IN 100 CD , , 000 CD have VHP the DT greatest JJS opportunities NNS to TO shortchange VV the DT government NN because IN they PP receive VV most JJS of IN the DT nonwage NN income NN . SENT Most JJS would MD agree VV that IN the DT article NN contains VVZ only RB true JJ and CC accurate JJ statements NNS . SENT however RB , , most JJS would MD also RB agree VV that IN the DT statements NNS are VBP more RBR likely JJ to TO be VB made VVN by IN a DT liberal NN than IN a DT conservative JJ . SENT Indeed RB , , the DT centrist JJ and CC right JJ leaning VVG news NN outlets NNS by IN our PP$ measure NN the DT Washington NP Times NP , , Fox NP News NP Special JJ Report NP , , the DT Newshour NP with IN Jim NP Lehrer NP , , ABC NP s PP Good NP Morning NP America NP , , and CC CNN NP s PP Newsnight NP with IN Aaron NP Brown NP failed VVD to TO mention VV any DT of IN these DT facts NNS . SENT Meanwhile RB , , three CD of IN the DT outlets NNS on IN the DT left VVN side NN of IN our PP$ spectrum NN CBS NP Evening NP News NP , , USA NP Today NP , , and CC the DT news NN pages NNS of IN the DT Wall NP Street NP Journal NP did VVD mention VV at IN least JJS one CD of IN the DT facts NNS . SENT Likewise RB , , on IN the DT opposite JJ side NN of IN the DT political JJ spectrum NN there EX are VBP true JJ and CC accurate JJ facts NNS that IN conservatives NNS are VBP more RBR likely JJ to TO state NN than IN liberals NNS . SENT For IN instance NN , , on IN March NP 28 CD , , 2002 CD , , the DT Washington NP Times NP , , the DT most RBS conservative JJ outlet NN by IN our PP$ measure NN , , reported VVN that IN Congress NP earmarked VVD 304 CD , , 000 CD to TO restore VV opera NN houses NNS in IN Connecticut NP , , Michigan NP , , and CC Washington NP . SENT 19 CD Meanwhile RB , , none NN of IN the DT other JJ outlets NNS in IN our PP$ sample NN mentioned VVD this DT fact NN . SENT Moreover RB , , the DT Washington NP Times NP article NN failed VVD to TO mention VV facts NNS that IN a DT liberal NN would MD be VB more RBR likely JJ to TO note VV . SENT For IN instance NN , , it PP did VVD not RB mention VV that IN the DT 304 CD , , 000 CD comprises VVZ a DT very RB tiny JJ portion NN of IN the DT federal JJ budget NN . SENT We PP also RB believe VVP that IN our PP$ notion NN of IN bias NN is VBZ the DT one NN that WDT is VBZ more RBR commonly RB adopted VVN by IN other JJ authors NNS . SENT For IN instance NN , , Lott NP and CC Hasset NP 2004 CD do VVP not RB assert VV that IN one CD headline NN in IN their PP$ data NN set NN is VBZ false JJ e NN . SENT g NN . SENT GDP NP Rises VVZ 5 CD Percent NN while IN another DT headline NN is VBZ true JJ e NN . SENT g NN . SENT GDP NP Growth NP Less RBR Than IN Expected VVN . SENT Rather RB , , the DT choice NN of IN headlines NNS is VBZ more RBR a DT question NN of IN taste NN , , or CC perhaps RB fairness NN , , than IN a DT question NN of IN accuracy NN or CC honesty NN . SENT Also RB , , much RB of IN Goldberg NP s PP 2002 NP and CC Alterman NP s VVZ 2003 CD complaints NNS about IN media NNS bias NN are VBP that IN some DT stories NNS receive VVP scant JJ attention NN from IN the DT press NN , , not RB that IN the DT stories NNS receive VVP inaccurate JJ attention NN . SENT For IN instance NN , , Goldberg NP notes VVZ how WRB few JJ stories NNS the DT media NNS devote VVP to TO the DT problems NNS faced VVN by IN children NNS of IN dual JJ career NN parents NNS . SENT On IN the DT opposite JJ side NN , , Alterman NP notes VVZ how WRB few JJ stories NNS the DT media NNS devote VVP to TO corporate JJ fraud NN . SENT Our PP$ notion NN of IN bias NN also RB seems VVZ closely RB aligned VVN to TO the DT notion NN described VVN by IN Bozell NP and CC Baker NP 1990 CD , , 3 CD . SENT But CC though IN bias NN in IN the DT media NN exists VVZ , , it PP is VBZ rarely RB a DT conscious JJ attempt NN to TO distort VV the DT news NN . SENT It PP stems VVZ from IN the DT fact NN that IN most JJS members NNS of IN the DT media NNS elite NNS have VHP little JJ contact NN with IN conservatives NNS and CC make VV little JJ effort NN to TO understand VV the DT conservative JJ viewpoint NN . SENT Their PP$ friends NNS are VBP liberals NNS , , what WP they PP read VVP and CC hear VVP is VBZ written VVN by IN liberals NNS . SENT 20 CD Similar JJ to TO the DT facts NNS and CC stories NNS that IN journalists NNS report NN , , the DT citations NNS that IN they PP gather VVP from IN experts NNS are VBP also RB very RB rarely RB dishonest JJ or CC inaccurate JJ . SENT Many JJ , , and CC perhaps RB most JJS , , simply RB indicate VVP the DT side NN of IN an DT issue NN that IN the DT expert NN or CC his PP$ or CC her PP$ organization NN favors VVZ . SENT For IN instance NN , , on IN April NP 27 CD , , 2002 CD , , the DT New NP York NP Times NP reported VVD that IN Congress NP passed VVD a DT 100 CD billion CD farm NN subsidies NNS bill NN that WDT also RB gave VVD vouchers NNS to TO the DT elderly JJ to TO buy VV fresh JJ fruits NNS and CC vegetables NNS . SENT This DT is VBZ a DT terrific JJ outcome NN one NN of IN the DT most RBS important JJ pieces NNS of IN social JJ welfare NN legislation NN this DT year NN , , said VVD Stacy NP Dean NP of IN the DT Center NP on IN Budget NP and CC Policy NP Priorities NNS , , her PP only RB quote VVP in IN the DT article NN . SENT In IN another DT instance NN , , on IN May NP 19 CD , , 2001 CD Similarly RB , , another DT large JJ fraction NN of IN cases NNS involve VVP the DT organization NN s PP views VVZ of IN politicians NNS . SENT For IN instance NN , , on IN March NP 29 CD , , 2002 CD , , the DT Washington NP Times NP reported VVD that IN the DT National NP Taxpayers NPS Union NP gave VVD Hillary NP Clinton NP a DT score NN of IN 3 CD percent NN on IN its PP$ annual JJ rating NN of IN Congress NP . SENT The DT story NN noted VVD that IN the DT score NN , , according VVG to TO the DT NTU NP , , was VBD the DT worst JJS score NN for IN a DT Senate NP freshman NN in IN their PP$ first JJ year NN in IN 5 CD , , 1992 CD , , CBS NP Evening NP News NP reported VVD a DT fact NN that IN liberals NNS are VBP more RBR likely JJ to TO note VV than IN conservatives NNS . SENT The DT United NP States NP now RB has VHZ greater JJR disparities NNS of IN income NN than IN virtually RB any DT Western JJ European JJ country NN , , said VVD Robert NP Greenstein NP of IN the DT Center NP on IN Budget NP and CC Policy NP Priorities NNS . SENT Meanwhile RB , , on IN May NP 30 CD , , 2003 CD , , CNN NP s PP Newsnight NP with IN Aaron NP Brown NP noted VVD a DT fact NN that IN conservatives NNS are VBP more RBR likely JJ to TO state NN than IN liberals NNS . SENT In IN a DT story NN about IN the DT FCC NP s PP decision NN to TO weaken VV regulations NNS about IN media NNS ownership NN , , it PP quoted VVD Adam NP Thierer NP of IN the DT Cato NP Institute NP , , L NP et CC s PP start VVP by IN stepping VVG back RB and CC taking VVG a DT look NN at IN the DT landscape NN of IN today NN versus CC , , say VVP , , 10 CD , , 15 CD , , 25 CD , , 30 CD years NNS ago RB . SENT And CC by IN almost RB every DT measure NN that IN you PP can MD go VV by IN , , you PP can MD see VV that IN there EX is VBZ more JJR diversity NN , , more JJR competition NN , , more JJR choice NN for IN consumers NNS and CC citizens NNS in IN these DT marketplaces NNS . SENT 21 CD A NP Simple JJ Structural NP Model NP Define VV xi NN as IN the DT average NN adjusted VVN ADA NP score NN of IN the DT ith NP member NN of IN Congress NP . SENT Given VVN that IN the DT member NN cites VVZ a DT think VVP tank NN , , we PP assume VVP that IN the DT utility NN that IN he PP or CC she PP receives VVZ from IN citing VVG the DT jth NN think VVP tank NN is VBZ aj NP bj NP xi NN eij NN . SENT The DT parameter NN , , bj NP , , indicates VVZ the DT ideology NN of IN the DT think NN tank NN . SENT Note NN that IN if IN xi NN is VBZ large JJ i NP . SENT e SYM . SENT the DT legislator NN is VBZ liberal JJ , , then RB the DT legislator NN receives VVZ more JJR utility NN from IN citing VVG the DT think VVP tank NN if IN bj NP is VBZ large JJ . SENT The DT parameter NN , , aj NP , , represents VVZ a DT sort NN of IN valence NN factor NN as IN political JJ scientists NNS use VVP the DT term NN for IN the DT think NN tank NN . SENT It PP captures VVZ non JJ ideological JJ factors NNS that WDT lead VVP legislators NNS and CC journalists NNS to TO cite VV the DT think VVP tank NN . SENT Such JJ factors NNS may MD include VV such JJ things NNS as IN a DT reputation NN for IN high JJ quality NN and CC objective NN research NN , , which WDT may MD be VB orthogonal JJ to TO any DT ideological JJ leanings NNS of IN the DT think NN tank NN . SENT We PP assume VVP that DT eij NN is VBZ distributed VVN according VVG to TO a DT Weibull NP distribution NN . SENT As IN shown VVN by IN McFadden NP 1974 CD . SENT also RB see VV Judge NP , , et NP . SENT al NP , , 1985 CD , , pp NP . SENT 770 CD 2 CD , , this DT implies VVZ that IN the DT probability NN that IN member NN i NP selects VVZ the DT jth NN think VVP tank NN is VBZ exp NN aj NP bj NP xi NN k NN 1 CD J NP exp NN ak NN bk NN xi NN , , 1 CD where WRB J NP is VBZ the DT total JJ number NN of IN think NN tanks NNS in IN our PP$ sample NN . SENT Note NN that IN this DT probability NN term NN is VBZ no RB different JJ from IN the DT one NN we PP see VVP in IN a DT multinomial NN logit NN where WRB the DT only JJ independent JJ variable NN is VBZ xi NN . SENT Define VV cm NN as IN the DT estimated JJ adjusted JJ ADA NP score NN of IN the DT mth JJ media NN outlet NN . SENT Similar JJ to TO the DT members NNS of IN Congress NP , , we PP assume VVP that IN the DT utility NN that IN it PP receives VVZ from IN citing VVG the DT the DT jth NN think VVP tank NN is VBZ aj NP bj NP cm NN emj NN . SENT We PP assume VVP that DT emj NN is VBZ distributed VVN according VVG to TO a DT Weibull NP distribution NN . SENT This DT implies VVZ that IN the DT probability NN that IN media NNS outlet NN m NN selects VVZ the DT jth NN think VVP tank NN is VBZ exp NN aj NP bj NP cm NN k NN 1 CD J NP exp NN ak NN bk NN cm NN . SENT 2 CD Although IN this DT term NN is VBZ similar JJ to TO the DT term NN that WDT appears VVZ in IN a DT multinomial NN logit NN , , we PP cannot JJ use NN multinomial NN logit NN to TO estimate VV the DT parameters NNS . SENT The DT problem NN is VBZ that DT cm NN , , a DT parameter NN that IN we PP estimate VVP , , appears VVZ where WRB normally RB we PP would MD have VH an DT independent JJ variable NN . SENT Instead RB , , we PP construct VV a DT likelihood NN function NN from IN 1 CD and CC 2 CD , , and CC we PP use VVP the DT nlm NN non JJ linear JJ maximization NN command NN in IN R NN to TO obtain VV estimates NNS of IN each DT aj NN , , bj NP , , and CC cm NN . SENT Similar JJ to TO a DT multinomial NN logit NN , , it PP is VBZ impossible JJ to TO identify VV each DT aj NP and CC bj NP . SENT Consequently RB , , we PP arbitrarily RB choose VVP one PP think VVP tank NN and CC set VV its PP$ values NNS of IN aj NP and CC bj NP to TO zero CD . SENT It PP is VBZ convenient JJ to TO choose VV a DT think VVP tank NN that WDT is VBZ cited VVN frequently RB . SENT Also RB , , to TO make VV most JJS estimates NNS of IN the DT bj JJS s PP positive JJ , , it PP is VBZ convenient JJ to TO choose VV a DT think VVP tank NN that WDT is VBZ conservative JJ . SENT Consequently RB , , we PP chose VVD the DT Heritage NP Foundation NP . SENT It PP is VBZ easy JJ to TO prove VV that IN this DT choice NN does VVZ not RB affect VV our PP$ estimates NNS of IN cm NN . SENT That DT is VBZ , , if IN we PP had VHD chosen VVN a DT different JJ think NN tank NN , , then RB all DT estimates NNS of IN cm NN would MD be VB unchanged JJ . SENT This DT identification NN problem NN is VBZ not RB just RB a DT technical JJ point NN . SENT it PP also RB has VHZ an DT important JJ substantive JJ implication NN . SENT Our PP$ method NN does VVZ not RB need VV to TO determine VV any DT sort NN of IN assessment NN of IN the DT absolute JJ ideological JJ position NN of IN a DT think NN tank NN . SENT It PP only RB needs VVZ to TO assess VV the DT relative JJ position NN . SENT In IN fact NN , , our PP$ method NN cannot NN assess VV absolute JJ positions NNS . SENT As IN a DT concrete JJ example NN , , consider VVP the DT estimated JJ bj NN s PP for IN AEI NP and CC the DT Brookings NP Institution NP . SENT These DT values NNS are VBP . SENT 026 CD and NP . SENT 038 CD . SENT The DT fact NN that IN the DT Brookings NP estimate NN is VBZ larger JJR than IN the DT AEI NP estimate NN means VVZ that IN Brookings NP is VBZ more RBR liberal JJ than IN AEI NP . SENT More RBR precisely RB , , it PP means VVZ that IN as IN a DT legislator NN or CC journalist NN becomes VVZ more RBR liberal JJ , , he PP or CC she PP prefers VVZ more JJR and CC more JJR to TO cite VV Brookings NP than IN AEI NP . SENT These DT estimates NNS are VBP consistent JJ with IN the DT claim NN that IN AEI NP is VBZ conservative JJ in IN an DT absolute JJ sense NN , , while IN Brookings NP is VBZ liberal JJ . SENT But CC they PP are VBP also RB consistent JJ with IN a DT claim NN , , e NN . SENT g NN . SENT , , that IN AEI NP is VBZ moderate JJ left NN while IN Brookings NP is VBZ far RB left VVN or CC also RB the DT possibility NN that IN AEI NP is VBZ far RB right JJ while IN Brookings NP is VBZ moderate JJ right NN . SENT This DT is VBZ related VVN to TO the DT fact NN that IN our PP$ model NN cannot NN fully RB identify VV the DT bj NN s PP that WDT is VBZ , , we PP could MD add VV the DT same JJ constant JJ to TO each DT and CC the DT value NN of IN the DT likelihood NN function NN and CC therefore RB the DT estimates NNS of IN the DT cm NN s PP would MD remain VV unchanged JJ . SENT One CD difficulty NN that WDT arose VVD in IN the DT estimation NN process NN is VBZ that IN it PP takes VVZ an DT unwieldy JJ amount NN of IN time NN to TO estimate VV all DT of IN the DT parameters NNS . SENT If IN we PP had VHD computed VVN a DT separate JJ aj NN and CC bj NN for IN each DT think VVP tank NN in IN our PP$ sample NN , , then RB we PP estimate VVP that IN our PP$ model NN would MD take VV over IN two CD weeks NNS to TO converge VV and CC produce VV estimates NNS . SENT 22 CD Complicating VVG this DT , , we PP compute VV estimates NNS for IN approximately RB two CD dozen NN different JJ specifications NNS of IN our PP$ basic JJ model NN . SENT Most JJS of IN these DT are VBP to TO test VV restrictions NNS of IN parameters NNS . SENT E SYM . SENT g NN . SENT we PP run VVP one CD specification NN where WRB the DT New NP York NP Times NP and CC NPR NP s PP Morning NP Edition NN are VBP constrained VVN to TO have VH the DT same JJ estimate NN of IN cm NN . SENT Thus RB , , if IN we PP estimated VVD the DT full JJ version NN of IN the DT model NN for IN each DT specification NN , , our PP$ computer NN would MD take VV approximately RB one CD year NN to TO produce VV all PDT the DT estimates NNS . SENT Instead RB , , we PP collapsed VVD data NNS from IN many JJ of IN the DT rarely RB cited VVN think VVP tanks NNS into IN six CD mega JJ think NN tanks NNS . SENT Specifically RB , , we PP estimated VVD a DT separate JJ aj NN and CC bj NN for IN the DT 44 CD think VVP tanks NNS that WDT were VBD most RBS cited VVN by IN the DT media NNS . SENT These DT comprised VVN 85 CD . SENT 6 CD of IN the DT total JJ number NN of IN media NNS citations NNS . SENT With IN the DT remaining JJ think NN tanks NNS , , we PP ordered VVD them PP left VVD to TO right RB according VVG to TO the DT average JJ ADA NP score NN of IN the DT legislators NNS who WP cited VVD them PP . SENT Let VV pmin NN and CC pmax NN be VB the DT minimum JJ and CC maximum JJ average JJ scores NNS for IN these DT think VVP tanks NNS . SENT To TO create VV the DT mega JJ think NN tanks NNS , , we PP defined VVD five CD cut NN points NNS to TO separate VV them PP . SENT Specifically RB , , we PP define VV cut NN point NN i NP as IN pi NN pmin NN i NP 6 CD pmax NN pmin NN . SENT In IN practice NN , , these DT five CD cut NN points NNS were VBD 22 CD . SENT 04 CD , , 36 CD . SENT 10 CD , , 50 CD . SENT 15 CD , , 64 CD . SENT 21 CD , , and CC 78 CD . SENT 27 CD . SENT The DT number NN of IN actual JJ and CC mega JJ think VVP tanks NNS to TO include VV respectively RB , , 44 CD and CC 6 CD is VBZ a DT somewhat RB arbitrary JJ choice NN . SENT We PP chose VVD 50 CD as IN the DT total JJ number NN because IN we PP often RB used VVD the DT mlogit NN procedure NN in IN Stata NP to TO compute VV seed NN values NNS . SENT This DT procedure NN is VBZ limited VVN to TO at IN most JJS 50 CD choices NNS , , which WDT meant VVD that IN we PP could MD estimate VV aj JJR and CC b LS s PP for IN at IN most JJS 50 CD think NN tanks NNS . SENT This DT still RB leaves VVZ an DT arbitrary JJ choice NN about IN how WRB many JJ of IN the DT 50 CD think VVP tanks NNS should MD be VB actual JJ think VVP tanks NNS and CC how WRB many JJ should MD be VB mega JJ think NN tanks NNS . SENT We PP experimented VVD with IN several JJ different JJ choices NNS . SENT Some DT choices NNS made VVD the DT media NNS appear VVP slightly RB more RBR liberal JJ than IN others NNS . SENT We PP chose VVD six CD as IN the DT number NN of IN mega JJ think VVP tanks NNS , , because IN it PP produced VVD approximately RB the DT average NN of IN the DT estimates NNS . SENT In IN the DT Appendix NN we PP also RB report VV results NNS when WRB instead RB we PP choose VVP 2 CD , , 3 CD , , 4 CD , , 5 CD , , 7 CD , , or CC 8 CD as IN the DT number NN of IN mega JJ think VVP tanks NNS . SENT Our PP$ choice NN to TO use VV 50 CD as IN the DT total JJ number NN of IN actual JJ and CC mega JJ think VVP tanks NNS , , if IN anything NN , , appears VVZ to TO makes NNS the DT media NNS appear VVP more RBR conservative JJ than IN they PP really RB are VBP . SENT In IN the DT Appendix NN we PP report VVP results NNS when WRB instead RB we PP chose VVD 60 CD , , 70 CD , , 80 CD , , and CC 90 CD as IN the DT total JJ number NN of IN actual JJ and CC mega JJ think VVP tanks NNS . SENT In IN general NN , , these DT choices NNS cause VVP average JJ estimate NN of IN cm NN to TO increase VV by IN approximately RB one CD or CC two CD points NNS . SENT Results NNS In IN Table NN 3 CD we PP list VVP the DT estimates NNS of IN cm NN , , the DT adjusted JJ ADA NP scores NNS for IN media NNS outlets NNS . SENT The DT ordering VVG of IN the DT scores NNS is VBZ largely RB consistent JJ with IN conventional JJ wisdom NN . SENT For IN instance NN , , the DT two CD most RBS conservative JJ outlets NNS are VBP the DT Washington NP Times NP and CC Fox NP News NP Special JJ Report NP , , two CD outlets NNS that WDT are VBP often RB called VVN conservative JJ e NN . SENT g NN . SENT see VV Alterman NP , , 2003 CD . SENT Near IN the DT liberal JJ end NN are VBP CBS NP Evening NP News NP and CC the DT New NP York NP Times NP . SENT Again RB , , these DT are VBP largely RB consistent JJ with IN the DT conventional JJ wisdom NN . SENT For IN instance NN , , CBS NP Evening NP News NP was VBD the DT target NN of IN best JJS selling NN book NN by IN Bernard NP Godberg NP 2002 CD , , a DT former JJ reporter NN who WP documents VVZ several JJ instances NNS of IN liberal JJ bias NN at IN the DT news NN show NN . SENT Further RBR , , some DT previous JJ scholarly JJ work NN shows VVZ CBS NP Evening NP News NP to TO be VB the DT most RBS liberal JJ of IN the DT three CD network NN evening NN news NN shows NNS . SENT James NP Hamilton NP 2004 CD recorded VVD the DT congressional JJ roll NN call NN votes NNS that IN the DT Americans NPS for IN Democratic NP Action NP chose VVD for IN its PP$ annual JJ scorecard NN , , and CC he PP examined VVD how WRB often RB each DT network NN covered VVD the DT roll NN calls VVZ . SENT Between IN 1969 CD and CC 1998 CD , , CBS NP Evening NP News NP consistently RB covered VVD these DT roll NN calls VVZ more RBR frequently RB than IN did VVD the DT other JJ two CD networks NNS . SENT 23 CD One CD surprise NN is VBZ the DT Wall NP Street NP Journal NP , , which WDT we PP find VVP as IN the DT most RBS liberal JJ of IN all DT 20 CD news NN outlets NNS . SENT We PP should MD first RB remind VV readers NNS that IN this DT estimate NN as RB well RB as IN all DT other JJ newspaper NN estimates NNS refers VVZ only RB to TO the DT news NN of IN the DT Wall NP Street NP Journal NP . SENT we PP omitted VVD all DT data NNS that WDT came VVD from IN its PP$ editorial JJ page NN . SENT If IN we PP included VVD data NNS from IN the DT editorial JJ page NN , , surely RB it PP would MD appear VV more RBR conservative JJ . SENT Second RB , , some DT anecdotal JJ evidence NN agrees VVZ with IN our PP$ result NN . SENT For IN instance NN , , Reed NP Irvine NP and CC Cliff NP Kincaid NP 2001 CD note NN that IN The DT Journal NP has VHZ had VHN a DT long JJ standing NN separation NN between IN its PP$ conservative JJ editorial NN pages NNS and CC its PP$ liberal JJ news NN pages NNS . SENT Paul NP Sperry NP , , in IN an DT article NN titled VVD the DT Myth NN of IN the DT Conservative NP Wall NP Street NP Journal NP , , notes VVZ that IN the DT news NN division NN of IN the DT Journal NP sometimes RB calls VVZ the DT editorial JJ division NN Nazis NNS . SENT Fact NN is VBZ , , Sperry NP writes VVZ , , the DT Journal NP s PP news NN and CC editorial NN departments NNS are VBP as RB politically RB polarized VVN as IN North NP and CC South NP Korea NP . SENT 24 CD Third NP , , a DT recent JJ poll NN from IN the DT Pew NP Research NP Center NP indicates VVZ that IN a DT greater JJR percentage NN of IN Democrats NPS , , 29 CD , , say VVP they PP trust VV the DT Journal NN than IN do VV Republicans NNS , , 23 CD . SENT Importantly RB , , the DT question NN did VVD not RB say VV the DT news NN division NN at IN the DT Wall NP Street NP Journal NP . SENT If IN it PP had VHD , , Democrats NNS surely RB would MD have VH said VVD they PP trusted VVD the DT Journal NP even RB more JJR , , and CC Republicans NPS even RB less JJR . SENT 25 CD Finally RB , , and CC perhaps RB most JJS important JJ , , a DT scholarly JJ study NN by IN Lott NP and CC Hasset NP 2004 CD gives VVZ evidence NN that WDT is VBZ consistent JJ with IN our PP$ result NN . SENT As RB far RB as IN we PP are VBP aware JJ this DT is VBZ the DT only RB other JJ study NN that WDT examines VVZ the DT political JJ bias NN of IN the DT news NN pages NNS of IN the DT Wall NP Street NP Journal NP . SENT Of IN the DT ten CD major JJ newspapers NNS that IN it PP examines VVZ , , the DT study NN estimates VVZ the DT Wall NP Street NP Journal NP as IN the DT second JJ most RBS liberal JJ . SENT 26 CD Only RB Newsday NP is VBZ more RBR liberal JJ , , and CC the DT Journal NP is VBZ substantially RB more RBR liberal JJ than IN the DT New NP York NP Times NP , , Washington NP Post NP , , L NP . SENT A DT . SENT Times NNS , , and CC USA NP Today NP . SENT Another DT somewhat RB surprising JJ result NN is VBZ our PP$ estimate NN of IN NPR NP s PP Morning NP Edition NN . SENT Conservatives NNS frequently RB list VVP NPR NP as IN an DT egregious JJ example NN of IN a DT liberal JJ news NN outlet NN . SENT 27 CD However RB , , by IN our PP$ estimate NN the DT outlet NN hardly RB differs VVZ from IN the DT average JJ mainstream NN news NN outlet NN . SENT For IN instance NN , , its PP$ score NN is VBZ approximately RB equal JJ to TO those DT of IN Time NP , , Newsweek NP , , and CC U NP . SENT S NP . SENT News NP and CC World NP Report NP , , and CC its PP$ score NN is VBZ slightly RB less JJR than IN the DT Washington NP Post NP s PP . SENT Further RBR , , our PP$ estimate NN places VVZ it PP well RB to TO the DT right NN of IN the DT New NP York NP Times NP , , and CC also RB to TO the DT right NN of IN the DT average JJ speech NN by IN Joe NP Lieberman NP . SENT These DT differences NNS are VBP statistically RB significant JJ . SENT 28 CD We PP mentioned VVD this DT finding NN to TO Terry NP Anderson NP , , an DT academic JJ economist NN and CC Executive NP Director NP of IN the DT Political NP Economy NP Research NP Center NP , , which WDT is VBZ among IN the DT list NN of IN think NN tanks NNS in IN our PP$ sample NN . SENT The DT average JJ score NN of IN legislators NNS citing VVG PERC NP was VBD 39 CD . SENT 9 CD , , which WDT places VVZ it PP as IN a DT moderate JJ right NN think VVP tank NN , , approximately RB as RB conservative JJ as IN RAND NNS is VBZ liberal JJ . SENT Anderson NP told VVD us PP , , When WRB NPR NP interviewed VVD us PP , , they PP were VBD nothing NN but CC fair JJ . SENT I PP think VVP the DT conventional JJ wisdom NN has VHZ overstated VVN any DT liberal JJ bias NN at IN NPR NP . SENT Our PP$ NPR NP estimate NN is VBZ also RB consistent JJ with IN James NP Hamilton NP s VVZ 2004 CD , , 108 CD research NN on IN audience NN ideology NN of IN news NN outlets NNS . SENT Hamilton NP finds VVZ that IN the DT average JJ NPR NP listener NN holds VVZ approximately RB the DT same JJ ideology NN as IN the DT average JJ network NN news NN viewer NN or CC the DT average JJ viewer NN of IN morning NN news NN shows NNS , , such JJ as IN Today NP or CC Good NP Morning NP America NP . SENT Indeed RB , , of IN the DT outlets NNS that IN he PP examines VVZ in IN this DT section NN of IN his PP$ book NN , , by IN this DT measure NN NPR NP is VBZ the DT ninth JJ most RBS liberal JJ out IN of IN eighteen CD . SENT Another DT result NN , , which WDT appears VVZ anomalous JJ , , is VBZ not RB so RB anomalous JJ upon IN further JJR examination NN . SENT This DT is VBZ the DT estimate NN for IN the DT Drudge NN Report NP , , which WDT at IN 60 CD . SENT 4 CD , , places VVZ it PP approximately RB in IN the DT middle NN of IN our PP$ mix NN of IN media NNS outlets NNS and CC approximately RB as RB liberal JJ as IN a DT typical JJ Southern NP Democrat NP , , such JJ as IN John NP Breaux NP D NP La NP . SENT . SENT We PP should MD emphasize VV that IN this DT estimate NN reflects VVZ both CC the DT news NN flashes VVZ that DT Matt NP Drudge NN reports NNS and CC the DT news NN stories NNS to TO which WDT his PP$ site NN links NNS on IN other JJ web NN sites NNS . SENT In IN fact NN , , of IN the DT entire JJ 311 CD think VVP tank NN citations NNS we PP found VVD in IN the DT Drudge NN Report NN , , only RB five CD came VVD from IN reports NNS written VVN by IN Matt NP Drudge NN . SENT Thus RB , , for IN all DT intents NNS and CC purposes NNS , , our PP$ estimate NN for IN the DT Drudge NN Report NN refers VVZ only RB to TO the DT articles NNS to TO which WDT the DT Report NP links NNS on IN other JJ web NN sites NNS . SENT Although IN the DT conventional JJ wisdom NN often RB asserts VVZ that IN the DT Drudge NN Report NN is VBZ relatively RB conservative JJ , , we PP believe VVP that IN the DT conventional JJ wisdom NN would MD also RB assert VV that IN if IN confined VVN only RB to TO the DT news NN stories NNS to TO which WDT the DT Report NP links NNS on IN other JJ web NN sites NNS this DT set NN would MD have VH a DT slant NN approximately RB equal JJ to TO the DT average JJ slant NN of IN all DT media NNS outlets NNS , , since RB , , after IN all DT , , it PP is VBZ comprised VVN of IN stories NNS from IN a DT broad JJ mix NN of IN other JJ outlets NNS . SENT 29 CD Digression NN . SENT Defining VVG the DT Center NP While IN the DT main JJ goal NN of IN our PP$ research NN is VBZ to TO provide VV a DT measure NN that WDT allows VVZ us PP to TO compare VV the DT ideological JJ positions NNS of IN media NNS outlets NNS to TO political JJ actors NNS , , a DT separate JJ goal NN is VBZ to TO express VV whether IN a DT news NN outlet NN is VBZ left VVN or CC right RB of IN center NN . SENT To TO do VV the DT latter NN , , we PP must MD define VV center NN . SENT This DT is VBZ a DT little RB more RBR arbitrary JJ than IN the DT first JJ exercise NN . SENT For IN instance NN , , the DT results NNS of IN the DT previous JJ section NN show NN that IN the DT average JJ NY NP Times NP article NN is VBZ approximately RB as RB liberal JJ as IN the DT average JJ Joe NP Lieberman NP D NP Ct NP . SENT speech NN . SENT While IN Lieberman NP is VBZ left VVN of IN center NN in IN the DT U NP . SENT S NP . SENT Senate NP , , many DT would MD claim VV that IN , , compared VVN to TO all DT persons NNS in IN the DT entire JJ world NN , , he PP is VBZ centrist JJ or CC even RB right RB leaning VVG . SENT And CC if IN the DT latter NN is VBZ one CD s NN criterion NN , , then RB nearly RB all RB of IN the DT media NNS outlets NNS that IN we PP examine VVP are VBP right JJ of IN center NN . SENT However RB , , we PP are VBP more RBR interested JJ in IN defining VVG centrist JJ by IN U NP . SENT S NP . SENT views NNS , , rather RB than IN world NN views NNS or CC , , say VVP , , European JJ views NNS . SENT One CD reason NN is VBZ that IN the DT primary JJ consumers NNS for IN the DT 20 CD news NN outlets NNS that IN we PP examine VVP are VBP in IN the DT U NP . SENT S NP . SENT If IN , , for IN example NN , , we PP wish VVP to TO test VV economic JJ theories NNS about IN whether IN U NP . SENT S NP . SENT news NN producers NNS are VBP adequately RB catering VVG to TO the DT demands NNS of IN their PP$ consumers NNS , , then RB U NP . SENT S NP . SENT consumers NNS are VBP the DT ones NNS on IN which WDT we PP should MD focus VV . SENT A DT second JJ reason NN is VBZ that IN the DT popular JJ debate NN on IN media NNS bias NN has VHZ focused VVN on IN U NP . SENT S NP . SENT views NNS , , not RB world NN views NNS . SENT For IN instance NN , , in IN Bernard NP Goldberg NP s VVZ 2002 CD insider NN account NN of IN CBS NP News NP , , he PP only RB claims VVZ that IN CBS NP is VBZ more RBR liberal JJ than IN the DT average JJ American NP , , not RB the DT average JJ European JJ or CC world NN citizen NN . SENT Given VVN this DT , , one CD of IN the DT simplest JJS definitions NNS of IN centrist JJ is VBZ simply RB to TO use VV the DT mean JJ or CC median JJ ideological JJ score NN of IN the DT U NP . SENT S NP . SENT House NP or CC Senate NP . SENT We PP focus VVP on RB mean VV scores NNS since IN the DT median NN tends VVZ to TO be VB unstable JJ . SENT 30 CD This NP is VBZ due JJ to TO the DT bi NP modal JJ nature NN that IN ADA NP scores NNS have VHP followed VVN in IN recent JJ years NNS . SENT For IN instance NN , , in IN 1999 CD only RB three CD senators NNS , , out RB of IN a DT total NN of IN 100 CD , , received VVD a DT score NN between IN 33 CD and CC 67 CD . SENT In IN contrast NN , , 33 CD senators NNS would MD have VH received VVN scores NNS in IN this DT range NN if IN the DT scores NNS had VHD been VBN distributed VVN uniformly RB , , and CC the DT number NN would MD be VB even RB larger JJR if IN scores NNS had VHD been VBN distributed VVN uni NN modally RB . SENT 31 CD We PP are VBP most RBS interested VVN in IN comparing VVG news NN outlets NNS to TO the DT centrist JJ voter NN , , who WP , , for IN a DT number NN of IN reasons NNS , , might MD not RB have VH the DT same JJ ideology NN as IN the DT centrist JJ member NN of IN Congress NP . SENT For IN instance NN , , because IN Washington NP , , D NP . SENT C SYM . SENT is VBZ not RB represented VVN in IN Congress NP and CC because IN D NP . SENT C SYM . SENT residents NNS tend VVP to TO be VB more RBR liberal JJ than IN the DT rest NN of IN the DT country NN , , the DT centrist JJ member NN of IN Congress NP should MD tend VV to TO be VB more RBR conservative JJ than IN the DT centrist JJ voter NN . SENT Another DT problem NN , , which WDT applies VVZ only RB to TO the DT Senate NP , , involves VVZ the DT fact NN that IN voters NNS from IN small JJ states NNS are VBP overrepresented VVN . SENT Since IN in IN recent JJ years NNS small JJ states NNS have VHP tended VVN to TO vote VV more RBR conservatively RB than IN large JJ states NNS , , this DT would MD cause VV the DT centrist JJ member NN of IN the DT Senate NP to TO be VB more RBR conservative JJ than IN the DT centrist JJ voter NN . SENT A DT third JJ reason NN , , which WDT applies VVZ only RB to TO the DT House NP , , is VBZ that IN gerrymandered VVN districts NNS can MD skew VV the DT relationship NN between IN a DT centrist JJ voter NN and CC a DT centrist JJ member NN of IN the DT House NP . SENT For IN instance NN , , although IN the DT total JJ votes NNS for IN Al NP Gore NP and CC George NP W NP . SENT Bush NP favored VVD Gore NP slightly RB , , the DT median JJ House NP district NN slightly RB favored VVN Bush NP . SENT Specifically RB , , if IN we PP exclude VVP the DT District NP of IN Columbia NP since IN it PP does VVZ not RB have VH a DT House NP member NN , , Al NP Gore NP received VVD 50 CD . SENT 19 CD of IN the DT two CD party NN vote NN . SENT Yet RB in IN the DT median JJ House NP district NN judging VVG by IN Gore NP Bush NP vote NN percentages NNS , , Al NP Gore NP received VVD only RB 48 CD . SENT 96 CD of IN the DT two CD party NN vote NN . SENT Twelve CD districts NNS had VHD percentages NNS between IN the DT median JJ and CC mean JJ percentages NNS . SENT The DT fact NN that IN the DT latter JJ number NN is VBZ smaller JJR than IN the DT former JJ number NN means VVZ that IN House NP districts NNS are VBP drawn VVN to TO favor VV Republicans NNS slightly RB . SENT Similar JJ results NNS occurry NN region NN First NP , , to TO account VV for IN the DT D NP . SENT C SYM . SENT bias NN , , we PP can MD add VV phantom JJ D NP . SENT C SYM . SENT legislators NNS to TO the DT House NP and CC Senate NP . SENT Of IN course NN , , we PP necessarily RB do VVP not RB know VV the DT ADA NP scores NNS of IN such JJ legislators NNS . SENT However RB , , it PP is VBZ reasonable JJ to TO believe VV that IN they PP would MD be VB fairly RB liberal JJ , , since IN D NP . SENT C SYM . SENT residents NNS tend VVP to TO vote VV overwhelmingly RB Democratic JJ in IN presidential JJ elections NNS . SENT They PP voted VVD 90 CD . SENT 5 CD for IN Gore NP in IN 2000 CD , , and CC they PP voted VVD 90 CD . SENT 6 CD for IN Kerry NP in IN 2004 CD . SENT For IN each DT year NN , , we PP gave VVD the DT phantom JJ D NP . SENT C SYM . SENT House NP member NN and CC senators NNS the DT highest JJS respective JJ House NP and CC Senate NP scores NNS that WDT occurred VVD that DT year NN . SENT Of IN course NN , , actual JJ D NP . SENT C SYM . SENT legislators NNS might MD not RB be VB quite RB so RB liberal JJ . SENT However RB , , one CD of IN our PP$ main JJ conclusions NNS is VBZ that IN the DT media NNS are VBP liberal JJ compared VVN to TO U NP . SENT S NP . SENT voters NNS . SENT Consequently RB , , it PP is VBZ better RBR err VV on IN the DT side NN of IN making VVG voters NNS appear VVP more RBR liberal JJ than IN they PP really RB are VBP than IN the DT opposite NN . SENT 32 RB The DT second JJ problem NN , , the DT small JJ state NN bias NN in IN the DT Senate NP , , can MD be VB overcome VVN simply RB by IN weighting NN each DT senator NN s PP score VVP by IN the DT population NN of IN his PP$ or CC her PP$ state NN . SENT The DT third JJ problem NN , , gerrymandered VVD districts NNS in IN the DT House NP , , is VBZ overcome VVN simply RB by IN the DT fact NN that IN we PP use VVP mean JJ scores NNS instead RB of IN the DT median NN . SENT 33 CD In IN Figure NP 1 CD , , we PP list VVP the DT mean JJ House NP and CC Senate NP scores NNS over IN the DT period NN 1947 CD 99 CD when WRB we PP use VVP this DT methodology NN i NP . SENT e SYM . SENT including VVG phantom JJ D NP . SENT C SYM . SENT legislators NNS and CC weighting NN senators NNS scores NNS by IN the DT population NN of IN their PP$ state NN . SENT The DT focus NN of IN our PP$ results NNS is VBZ for IN the DT period NN 1995 CD 99 CD . SENT We PP chose VVD 1999 CD as IN the DT end NN year NN simply RB because IN this DT is VBZ the DT last JJ year NN for IN which WDT Groseclose NP , , Levitt NP , , and CC Snyder NP 1999 CD computed VVD adjusted JJ ADA NP scores NNS . SENT However RB , , any DT conclusions NNS that IN we PP make VVP for IN this DT period NN should MD also RB hold VV for IN the DT 2000 CD 04 CD period NN , , since IN in IN the DT latter JJ period NN the DT House NP and CC Senate NP had VHD almost RB identical JJ party NN ratios NNS . SENT We PP chose VVD 1995 CD as IN the DT beginning NN year NN , , because IN it PP is VBZ the DT first JJ year NN after IN the DT historic JJ 1994 CD elections NNS , , where WRB Republicans NNS gained VVD 52 CD House NP seats NNS and CC eight CD Senate NP seats NNS . SENT This DT year NN , , it PP is VBZ reasonable JJ to TO believe VV , , marks VVZ the DT beginning NN of IN separate JJ era NN of IN American JJ politics NNS . SENT As IN a DT consequence NN , , if IN one PP wanted VVD to TO test VV hypotheses NNS about IN the DT typical JJ U NP . SENT S NP . SENT voter NN of IN , , say VV , , 1999 CD , , then RB the DT years NNS 1998 CD , , 1997 CD , , 1996 CD , , and CC 1995 CD would MD also RB provide VV helpful JJ data NNS . SENT However RB , , prior JJ years NNS would MD not RB . SENT Over IN this DT period NN the DT mean JJ score NN of IN the DT Senate NP after IN including VVG phantom JJ D NP . SENT C SYM . SENT senators NNS and CC weighting NN by IN state NN population NN varied VVN between IN 49 CD . SENT 28 CD and CC 50 CD . SENT 87 CD . SENT The DT mean NN of IN these DT means NN was VBD 49 CD . SENT 94 CD . SENT The DT similar JJ figure NN for IN the DT House NP was VBD 50 CD . SENT 18 CD . SENT After IN rounding VVG , , we PP use VVP the DT midpoint NN of IN these DT numbers NNS , , 50 CD . SENT 1 CD , , as IN our PP$ estimate NN of IN the DT adjusted JJ ADA NP score NN of IN the DT centrist JJ U NP . SENT S NP . SENT voter NN . SENT 34 CD A NP counter NN view NN is VBZ that IN the DT 1994 CD elections NNS did VVD not RB mark VV a DT new JJ era NN . SENT Instead RB , , as IN some DT might MD argue VV , , these DT elections NNS were VBD an DT anomaly NN , , and CC the DT congresses NNS of IN the DT decade NN or CC so RB before IN the DT 1994 CD elections NNS are VBP a DT more RBR appropriate JJ representation NN of IN voter NN sentiment NN of IN the DT late JJ 1990 CD s PP and CC early RB 2000 CD s PP . SENT Although IN we PP do VVP not RB agree VV , , we PP think VVP it PP is VBZ a DT useful JJ straw NN man NN . SENT Consequently RB , , we PP construct VV an DT alternative JJ measure NN based VVN on IN the DT congresses NNS that WDT served VVD between IN 1975 CD and CC 1994 CD . SENT We PP chose VVD 1975 CD , , because IN this DT was VBD the DT first JJ year NN of IN the DT Watergate NP babies NNS in IN Congress NP . SENT As IN Figure NN 1 CD shows NNS , , this DT year NN produced VVD a DT large JJ liberal JJ shift NN in IN Congress NP . SENT This DT period NN , , 1975 CD 94 CD , , also RB happens VVZ to TO be VB the DT most RBS liberal JJ 20 CD year NN period NN for IN the DT entire JJ era NN that IN the DT ADA NP has VHZ been VBN recording VVG vote NN scores NNS . SENT As IN we PP show VVP later RBR , , even RB if IN we PP use VVP this DT period NN to TO define VV centrist JJ , , all RB but CC two CD of IN the DT media NNS outlets NNS in IN our PP$ sample NN are VBP still RB left VVN of IN center NN . SENT The DT average JJ ADA NP score NN of IN senators NNS during IN the DT 1975 CD 94 CD period NN after IN including VVG phantom JJ D NP . SENT C SYM . SENT senators NNS and CC weighting NN according VVG to TO state NN population NN was VBD 53 CD . SENT 51 CD . SENT The DT similar JJ figure NN for IN the DT House NP was VBD 54 CD . SENT 58 CD . SENT After IN rounding VVG , , we PP use VVP the DT midpoint NN of IN these DT two CD scores NNS to TO define VV 54 CD . SENT 0 CD as IN the DT centrist JJ U NP . SENT S NP . SENT voter NN during IN 1975 CD 94 CD . SENT 35 CD Further RBR Results NNS . SENT How WRB Close RB are VBP Media NN Outlets NNS to TO the DT Center NP . SENT Next RB , , we PP compute VV the DT difference NN of IN a DT media NNS outlet NN s PP score VVP from IN 50 CD . SENT 1 CD to TO judge VV how WRB centrist JJ it PP is VBZ . SENT We PP list VVP these DT results NNS in IN Table NN 4 CD . SENT Most RBS striking JJ is VBZ that IN all DT but CC two CD of IN the DT outlets NNS we PP examine VVP are VBP left VVN of IN center NN . SENT Even RB more RBR striking JJ is VBZ that IN if IN we PP use VVP the DT more RBR liberal JJ definition NN of IN center NN 54 CD . SENT 0 CD the DT one NN constructed VVN from IN congressional JJ scores NNS from IN 1975 CD 94 CD it PP is VBZ still RB the DT case NN that IN eighteen CD of IN twenty CD outlets NNS are VBP left VVN of IN center NN . SENT The DT first JJ , , second JJ , , and CC third NN most RBS centrist JJ outlets NNS are VBP respectively RB Newshour NP with IN Jim NP Lehrer NP , , CNN NP s PP Newsnight NP with IN Aaron NP Brown NP , , and CC ABC NP s PP Good NP Morning NP America NP . SENT The DT scores NNS of IN Newsnight NP and CC Good NP Morning NP America NP were VBD not RB statistically RB different JJ from IN the DT center NN , , 50 CD . SENT 1 CD . SENT Although IN the DT point NN estimate NN of IN Newshour NP was VBD more RBR centrist JJ than IN the DT other JJ two CD outlets NNS , , its PP$ difference NN from IN the DT center NN is VBZ statistically RB significant JJ . SENT The DT reason NN is VBZ that IN its PP$ margin NN of IN error NN is VBZ smaller JJR than IN the DT other JJ two CD , , which WDT is VBZ due JJ primarily RB to TO the DT fact NN that IN we PP collected VVD more JJR observations NNS for IN this DT outlet NN . SENT Interestingly RB , , in IN the DT four CD presidential JJ and CC vice NN presidential JJ debates NNS of IN the DT 2004 CD election NN , , three CD of IN the DT four CD moderators NNS were VBD selected VVN from IN these DT three CD outlets NNS . SENT The DT fourth JJ moderator NN , , Bob NP Schieffer NP , , works VVZ at IN an DT outlet NN that IN we PP did VVD not RB examine VV , , CBS NP s PP Face NP the DT Nation NN . SENT The DT fourth JJ and CC fifth RB most JJS centrist JJ outlets NNS are VBP the DT Drudge NN Report NP and CC Fox NP News NP Special JJ Report NP with IN Brit NP Hume NP . SENT Their PP$ scores NNS are VBP significantly RB different JJ from IN the DT center NN at IN a DT 95 CD significance NN level NN . SENT Nevertheless RB , , top JJ five CD outlets NNS in IN Table NN 4 CD are VBP in IN a DT statistical JJ dead JJ heat NN for IN most RBS centrist JJ . SENT Even RB at IN an DT 80 CD level NN of IN significance NN , , none NN of IN these DT outlets NNS can MD be VB called VVN more RBR centrist JJ than IN any DT of IN the DT others NNS . SENT The DT sixth JJ and CC seventh RB most JJS centrist JJ outlets NNS are VBP ABC NP World NP News NP Tonight NP and CC NBC NP Nightly JJ News NN . SENT These DT outlets NNS are VBP almost RB in IN a DT statistical JJ tie NN with IN the DT five CD most RBS centrist JJ outlets NNS . SENT For IN instance NN , , each DT has VHZ a DT score NN that WDT is VBZ significantly RB different JJ from IN Newshour NP s PP at IN the DT 90 CD confidence NN level NN , , but CC not RB at IN the DT 95 CD confidence NN level NN . SENT The DT eighth JJ most RBS centrist JJ outlet NN , , USA NP Today NP , , received VVD a DT score NN that WDT is VBZ significantly RB different JJ from IN Newshour NP s PP at IN the DT 95 CD confidence NN level NN . SENT Fox NP News NP Special JJ Report NN is VBZ approximately RB one CD point NN more RBR centrist JJ than IN ABC NP s PP World NP News NP Tonight NP with IN Peter NP Jennings NP or CC NBC NP s PP Nightly JJ News NN with IN Tom NP Brokaw NP . SENT In IN neither DT case NN is VBZ the DT difference NN statistically RB significant JJ . SENT Given VVN that IN Special JJ Report NP is VBZ one CD hour NN long JJ and CC the DT other JJ two CD shows NNS are VBP a DT half NN hour NN long RB , , our PP$ measure NN implies VVZ that IN if IN a DT viewer NN watched VVD all DT three CD shows NNS each DT night NN , , he PP or CC she PP would MD receive VV a DT nearly RB perfectly RB balanced JJ version NN of IN the DT news NN . SENT In IN fact NN , , it PP would MD be VB slanted VVN slightly RB left VVN by IN 0 CD . SENT 4 CD ADA NP points NNS . SENT Special JJ Report NP is VBZ approximately RB thirteen CD points NNS more RBR centrist JJ than IN CBS NP Evening NP News NP with IN Dan NP Rather NP . SENT This DT difference NN is VBZ significant JJ at IN the DT 99 CD confidence NN level NN . SENT Also RB at IN 99 CD confidence NN levels NNS , , we PP can MD conclude VV that IN NBC NP Nightly JJ News NP and CC ABC NP World NP News NP Tonight NP are VBP more RBR centrist JJ than IN CBS NP Evening NP News NP . SENT The DT most RBS centrist JJ newspaper NN in IN our PP$ sample NN is VBZ USA NP Today NP . SENT However RB , , its PP$ distance NN from IN the DT center NN is VBZ not RB significantly RB different JJ from IN the DT distances NNS of IN the DT Washington NP Times NP or CC the DT Washington NP Post NP . SENT Interestingly RB , , our PP$ measure NN implies VVZ that IN if IN one CD spent VVD an DT equal JJ amount NN of IN time NN reading VVG the DT Washington NP Times NP and CC Washington NP Post NP , , he PP or CC she PP would MD receive VV a DT nearly RB perfectly RB balanced JJ version NN of IN the DT news NN . SENT It PP would MD be VB slanted VVN left VVN by IN only RB 0 CD . SENT 9 CD ADA NP points NNS . SENT If IN instead RB we PP use VVP the DT 54 CD . SENT 1 CD as IN our PP$ measure NN of IN centrist JJ which WDT is VBZ based VVN on IN congressional JJ scores NNS of IN the DT 1975 CD 94 CD period NN , , the DT rankings NNS change NN , , but CC not RB greatly RB . SENT The DT most RBS substantial JJ is VBZ the DT Fox NP News NP Special JJ Report NP , , which WDT drops VVZ from IN fifth JJ to TO fifteenth NN most RBS centrist JJ . SENT The DT Washington NP Times NP also RB changes VVZ significantly RB . SENT It PP drops VVZ from IN tenth NN to TO seventeenth NN most RBS centrist JJ . SENT Another DT implication NN of IN the DT scores NNS concerns VVZ the DT New NP York NP Times NP . SENT Although IN some DT claim NN that IN the DT liberal JJ bias NN of IN the DT New NP York NP Times NP is VBZ balanced VVN by IN the DT conservative JJ bias NN of IN other JJ outlets NNS , , such JJ as IN the DT Washington NP Times NP or CC Fox NP News NP Special JJ Report NP , , this DT is VBZ not RB quite RB true JJ . SENT The DT New NP York NP Times NP is VBZ slightly RB more RBR than IN twice RB as RB far RB from IN the DT center NN as IN Special JJ Report NP . SENT Consequently RB , , to TO gain VV a DT balanced JJ perspective NN , , a DT news NN consumer NN would MD need VV to TO spend VV twice RB as IN much JJ time NN watching VVG Special JJ Report NN as IN he PP or CC she PP spends VVZ reading VVG the DT New NP York NP Times NP . SENT Alternatively RB , , to TO gain VV a DT balanced JJ perspective NN , , a DT reader NN would MD need VV to TO spend VV 50 CD more JJR time NN reading VVG the DT Washington NP Times NP than IN the DT New NP York NP Times NP . SENT Potential JJ Biases NNS A DT frequent JJ concern NN of IN our PP$ method NN is VBZ a DT form NN of IN the DT following VVG claim NN . SENT The DT sample NN of IN think NN tanks NNS has VHZ a DT rightward JJ leftward JJ tilt NN rather RB than IN an DT ideological JJ balance NN . SENT E SYM . SENT g NN . SENT it PP does VVZ not RB include VV Public NP Citizen NP and CC many JJ other JJ Nader NP groups NNS . SENT E SYM . SENT g NN . SENT , , it PP does VVZ not RB include VV the DT National NP Association NP of IN Manufacturers NPS or CC the DT Conference NN of IN Catholic JJ Bishops NNS . SENT Consequently RB this DT will NN bias NN estimates VVZ to TO the DT right NN left VVD . SENT However RB , , the DT claim NN is VBZ not RB true JJ , , and CC here RB is VBZ the DT intuition NN . SENT If IN the DT sample NN of IN think NN tanks NNS were VBD say VV disproportionately RB conservative JJ , , this RB , , of IN course NN , , would MD cause VV media NNS outlets NNS to TO cite VV conservative JJ think VVP tanks NNS more RBR frequently RB as IN a DT proportion NN of IN citations NNS that IN we PP record NN in IN our PP$ sample NN . SENT This DT might MD seem VV to TO cause VV the DT media NNS to TO appear VV more RBR conservative JJ . SENT However RB , , at IN the DT same JJ time NN it PP causes VVZ members NNS of IN Congress NP to TO appear VV more RBR conservative JJ . SENT Our PP$ method NN only RB measures VVZ the DT degree NN to TO which WDT media NN is VBZ liberal JJ or CC conservative JJ , , relative JJ to TO Congress NP . SENT Since IN it PP is VBZ unclear JJ how WRB such PDT a DT disproportionate JJ sample NN would MD affect VV the DT relative JJ degree NN to TO which WDT the DT media NNS cite VVP conservative JJ or CC liberal JJ think VVP tanks NNS , , there EX is VBZ no RB a DT priori NN reason NN for IN this DT to TO cause VV a DT bias NN to TO our PP$ method NN . SENT In IN fact NN , , a DT similar JJ concern NN could MD be VB leveled VVN against IN regression NN analysis NN . SENT As IN a DT simple JJ example NN , , consider VVP a DT researcher NN who WP regresses VVZ the DT arm NN lengths NNS of IN subjects NNS on IN their PP$ heights NNS . SENT Suppose VV instead RB of IN choosing VVG a DT balance NN of IN short JJ and CC tall JJ subjects NNS , , he PP or CC she PP chooses VVZ a DT disproportionate JJ number NN of IN tall JJ subjects NNS . SENT This DT will MD not RB affect VV his PP$ or CC her PP$ findings NNS about IN the DT relationship NN between IN height NN and CC arm NN length NN . SENT That DT is VBZ , , he PP or CC she PP will MD find VV that DT arm NN length NN is VBZ approximately RB half PDT the DT subject JJ s NN height NN , , and CC this DT estimate NN , , half NN , , would MD be VB the DT same JJ in IN expectation NN whether IN he PP or CC she PP chooses VVZ many JJ or CC few JJ tall JJ subjects NNS . SENT For IN similar JJ reasons NNS , , to TO achieve VV unbiased JJ estimates NNS in IN a DT regression NN , , econometrics NN textbooks NNS place VVP no DT restrictions NNS on IN the DT distribution NN of IN independent JJ variables NNS . SENT They PP only RB place VVP restrictions NNS upon IN , , e NN . SENT g NN . SENT , , the DT correlation NN of IN the DT independent JJ variables NNS and CC the DT error NN term NN . SENT Another DT frequent JJ concern NN of IN our PP$ method NN takes VVZ a DT form NN of IN the DT following VVG claim NN . SENT Most JJS of IN the DT congressional JJ data NNS came VVD from IN years NNS in IN which WDT the DT Republicans NPS were VBD the DT majority NN party NN . SENT Since IN the DT majority NN can MD control VV the DT rules NNS , , and CC hence RB debate NN time NN given VVN to TO each DT side NN , , this DT will MD cause VV the DT sample NN to TO have VH a DT disproportionate JJ number NN of IN citations NNS by IN Republicans NPS . SENT In IN turn NN , , this DT will MD cause VV media NNS outlets NNS to TO appear VV to TO be VB more RBR liberal JJ than IN they PP really RB are VBP . SENT First RB , , it PP is VBZ not RB true JJ that IN the DT majority NN party NN gives VVZ itself PP a DT disproportionate JJ amount NN of IN debate NN time NN . SENT Instead RB , , the DT usual JJ convention NN is VBZ for IN debate NN time NN to TO be VB divided VVN equally RB between IN proponents NNS and CC opponents NNS of IN any DT issue NN . SENT This DT means VVZ that IN the DT majority NN party NN actually RB gives VVZ itself PP less JJR than IN the DT proportionate JJ share NN . SENT However RB , , this DT convention NN is VBZ countered VVN by IN two CD other JJ factors NNS , , which WDT tend VVP to TO give VV the DT majority NN and CC minority NN party NN their PP$ proportionate JJ share NN of IN speech NN time NN . SENT 1 LS Many JJ of IN the DT speeches NNS in IN the DT Congressional NP Record NP are VBP not RB part NN of IN the DT debate NN on IN a DT particular JJ bill NN or CC amendment NN but CC are VBP from IN special JJ orders NNS generally RB in IN the DT evening NN after IN the DT chamber NN has VHZ adjourned VVN from IN official JJ business NN or CC one CD minutes NNS generally RB in IN the DT morning NN before IN the DT chamber NN has VHZ convened VVN for IN official JJ business NN . SENT For IN these DT types NNS of IN speeches NNS there EX are VBP no DT restrictions NNS of IN party NN balance NN , , and CC for IN the DT most JJS part NN , , any DT legislator NN who WP shows VVZ up RP at IN the DT chamber NN is VBZ allowed VVN to TO make VV such PDT a DT speech NN . SENT 2 CD Members NNS often RB place VVP printed JJ material NN into IN the DT Record NP . SENT We PP included VVD such PDT printed VVN material NN as IN a DT part NN of IN any DT member NN s PP speech NN . SENT In IN general JJ , , there EX are VBP no DT restrictions NNS on IN the DT amount NN of IN material NN that IN a DT legislator NN can MD place VV into IN the DT Record NP or CC whether IN he PP or CC she PP can MD do VV this DT . SENT Thus RB , , e NN . SENT g NN . SENT if IN a DT legislator NN has VHZ run VVN out RP of IN time NN to TO make VV his PP$ or CC her PP$ speech NN , , he PP or CC she PP can MD request VV that IN the DT remainder NN be VB placed VVN in IN written JJ form NN into IN the DT Record NP . SENT But CC even RB if IN the DT majority NN party NN were VBD given VVN more RBR or CC less RBR than IN its PP$ proportionate JJ share NN of IN speech NN time NN , , this DT would MD not RB bias VV our PP$ estimates NNS . SENT With IN each DT media NNS outlet NN , , our PP$ method NN seeks VVZ the DT legislator NN who WP has VHZ a DT citation NN pattern NN that WDT is VBZ most RBS similar JJ to TO that DT outlet NN . SENT For IN instance NN , , suppose VVP that IN the DT New NP York NP Times NP cites VVZ liberal JJ think NN tanks NNS about RB twice RB as IN often RB as RB conservative JJ think VVP tanks NNS . SENT Suppose VVP as IN we PP actually RB find VVP that IN Joe NP Lieberman NP is VBZ the DT legislator NN who WP has VHZ the DT mix NN of IN citations NNS most RBS similar JJ to TO the DT New NP York NP Times NP that WDT is VBZ , , suppose VVP he PP also RB tends VVZ to TO cite VV liberal JJ think VVP tanks NNS twice RB as RB often RB as IN conservative JJ think VVP tanks NNS . SENT Now RB consider VV a DT congressional JJ rules NNS change NN that WDT cuts VVZ the DT speech NN time NN of IN Democrats NNS in IN half NN . SENT Although IN this DT will MD affect VV the DT number NN of IN total JJ citations NNS that IN Lieberman NP makes VVZ , , it PP will MD not RB affect VV the DT proportion NN of IN citations NNS that IN he PP makes VVZ to TO liberal JJ and CC conservative JJ think VVP tanks NNS . SENT Hence RB , , our PP$ method NN would MD still RB give VV the DT New NP York NP Times NP an DT ADA NP score NN equal JJ to TO Joe NP Lieberman NP s PP . SENT 36 CD More JJR problematic JJ is VBZ a DT concern NN that IN congressional JJ citations NNS and CC media NNS citations NNS do VVP not RB follow VV the DT same JJ data NN generating NN process NN . SENT For IN instance NN , , suppose VVP that IN a DT factor NN besides IN ideology NN affects VVZ the DT probability NN that IN a DT legislator NN or CC reporter NN will MD cite VV a DT think VVP tank NN , , and CC suppose VVP that IN this DT factor NN affects VVZ reporters NNS and CC legislators NNS differently RB . SENT Indeed RB , , John NP Lott NP and CC Kevin NP Hasset NP have VHP invoked VVN a DT form NN of IN this DT claim NN to TO argue VV that IN our PP$ results NNS are VBP biased VVN toward IN making VVG the DT media NNS appear VVP more RBR conservative JJ than IN they PP really RB are VBP . SENT They PP note VVP . SENT For IN example NN , , Lott NP 2003 CD , , Chapter NN 2 CD shows NNS that IN the DT New NP York NP Times NP stories NNS on IN gun NN regulations NNS consistently RB interview NN academics NNS who WP favor VVP gun NN control NN , , but CC uses VVZ gun NN dealers NNS or CC the DT National NP Rifle NP Association NP to TO provide VV the DT other JJ side NN In IN this DT case NN , , this DT bias NN makes VVZ Groseclose NP and CC Milyo NP s PP measure VV of IN the DT New NP York NP Times NP look VVP more RBR conservative JJ than IN is VBZ likely RB accurate JJ . SENT 2004 CD , , 8 CD However RB , , it PP is VBZ possible JJ , , and CC perhaps RB likely JJ , , that IN members NNS of IN Congress NP practice NN the DT same JJ tendency NN that IN Lott NP and CC Hassett NP have VHP identified VVN with IN reporters NNS that WDT is VBZ , , to TO cite VV academics NNS when WRB they PP make VVP an DT anti JJ gun NN argument NN and CC to TO cite VV , , say VV , , the DT NRA NN when WRB they PP make VVP a DT pro JJ gun NN argument NN . SENT If IN so RB , , then RB our PP$ method NN will MD have VH no DT bias NN . SENT On IN the DT other JJ hand NN , , if IN members NNS of IN Congress NP do VVP not RB practice VV the DT same JJ tendency NN as IN journalists NNS , , then RB this DT can MD cause VV a DT bias NN to TO our PP$ method NN . SENT But CC even RB here RB , , it PP is VBZ not RB clear JJ which WDT direction NN the DT bias NN will MD occur VV . SENT For IN instance NN , , it PP is VBZ possible JJ that IN members NNS of IN Congress NP have VHP a DT greater JJR lesser JJR tendency NN than IN journalists NNS to TO cite VV such JJ academics NNS . SENT If IN so RB , , then RB this DT will MD cause VV our PP$ method NN to TO make VV media NNS outlets NNS appear VVP more RBR liberal JJ conservative NN than IN they PP really RB are VBP . SENT In IN fact NN , , the DT criticism NN we PP have VHP heard VVN most RBS frequently RB is VBZ a DT form NN of IN this DT concern NN , , but CC it PP is VBZ usually RB stated VVN in IN a DT way NN that WDT suggests VVZ the DT bias NN is VBZ in IN the DT opposite JJ direction NN . SENT Here RB is VBZ a DT typical JJ variant NN . SENT It PP is VBZ possible JJ that IN 1 CD Journalists NNS care VVP more JJR about IN the DT quality NN of IN a DT think VVP tank NN than IN do VV legislators NNS e NN . SENT g NN . SENT suppose VVP they PP prefer VVP to TO cite VV a DT think VVP tank NN with IN a DT reputation NN for IN serious JJ scholarship NN than IN another DT group NN that WDT is VBZ known VVN more RBR for IN its PP$ activism NN . SENT and CC 2 CD the DT liberal NN think VVP tanks NNS in IN the DT sample NN tend VVP to TO be VB of IN higher JJR quality NN than IN the DT conservative JJ think NN tanks NNS . SENT If IN statements NNS 1 CD and CC 2 CD are VBP true JJ , , then RB our PP$ method NN will MD indeed RB make VV media NNS outlets NNS appear VVP more RBR liberal JJ than IN they PP really RB are VBP . SENT That DT is VBZ , , the DT media NNS will MD cite VV liberal JJ think VVP tanks NNS more RBR , , not RB because IN they PP prefer VVP to TO cite VV liberal JJ think VVP tanks NNS , , but CC because IN they PP prefer VVP to TO cite VV high JJ quality NN think VVP tanks NNS . SENT On IN the DT other JJ hand NN , , if IN one CD statement NN is VBZ true JJ and CC the DT other JJ is VBZ false JJ , , then RB our PP$ method NN will MD make VV media NNS outlets NNS appear VVP more RBR conservative JJ than IN they PP really RB are VBP . SENT E SYM . SENT g NN . SENT suppose VVP journalists NNS care VVP about IN quality NN more RBR than IN legislators NNS , , but CC suppose VVP that IN the DT conservative JJ groups NNS in IN our PP$ sample NN tend VVP to TO be VB of IN higher JJR quality NN than IN the DT liberal JJ groups NNS . SENT Then RB the DT media NNS will MD tend VV to TO cite VV the DT conservative JJ groups NNS disproportionately RB , , but CC not RB because IN the DT media NNS are VBP conservative JJ , , rather RB because IN they PP have VHP a DT taste NN for IN quality NN . SENT This DT will MD cause VV our PP$ method NN to TO judge VV the DT media NNS as RB more RBR conservative JJ than IN they PP really RB are VBP . SENT Finally RB , , if IN neither DT statement NN is VBZ true JJ , , then RB our PP$ method NN will MD make VV media NNS outlets NNS appear VVP more RBR liberal JJ than IN they PP really RB are VBP . SENT Note NN that IN there EX are VBP four CD possibilities NNS by IN which WDT statements NNS 1 CD and CC 2 CD can MD be VB true JJ or CC false JJ . SENT Two CD lead NN to TO a DT liberal JJ bias NN and CC two CD lead NN to TO a DT conservative JJ bias NN . SENT To TO test VV this DT concern NN , , we PP collected VVD two CD variables NNS that WDT indicate VVP whether IN a DT think VVP tank NN or CC policy NN group NN is VBZ more RBR likely JJ to TO produce VV quality NN scholarship NN . SENT The DT first JJ variable NN , , staff NN called VVD fellows NNS , , is VBZ coded VVN as IN 1 CD if IN any DT staff NN members NNS on IN the DT group NN s PP website NN are VBP given VVN one CD of IN the DT following VVG titles NNS . SENT fellow JJ including VVG research NN fellow NN or CC senior JJ fellow NN , , researcher NN , , economist NN , , or CC analyst NN . SENT The DT second JJ variable NN , , closed VVD membership NN , , is VBZ coded VVN as IN a DT 0 CD if IN the DT web NN site NN of IN the DT group NN asks VVZ visitors NNS to TO join VV the DT group NN and CC 1 CD otherwise RB . SENT The DT idea NN behind IN this DT is VBZ that DT more JJR activist NN groups NNS are VBP more RBR likely JJ to TO recruit VV laypersons NNS for IN things NNS such JJ as IN protests NNS and CC letter NN writing VVG campaigns NNS to TO politicians NNS . SENT More RBR scholarly JJ groups NNS are VBP less RBR likely JJ to TO engage VV in IN these DT activities NNS . SENT Both DT variables NNS seem VVP to TO capture VV the DT conventional JJ wisdom NN about IN which WDT think VVP tanks NNS are VBP known VVN for IN quality NN scholarship NN . SENT For IN instance NN , , of IN the DT top JJ 25 CD most JJS cited VVN groups NNS in IN Table NN 1 CD , , the DT following NN had VHD both DT closed JJ membership NN and CC staff NN called VVD fellows NNS . SENT Brookings NP , , Center NP for IN Strategic NP and CC International NP Studies NPS , , Council NP on IN Foreign NP Relations NPS , , AEI NP , , RAND NN , , Carnegie NP Endowment NP for IN Intl NP . SENT Peace NN , , Cato NP , , Institute NP for IN International NP Economics NP , , Urban NP Institute NP , , Family NP Research NP Council NP , , and CC Center NP on IN Budget NP and CC Policy NP Priorities NNS . SENT Meanwhile RB , , the DT following VVG groups NNS , , which WDT most JJS would MD agree VV are VBP more RBR commonly RB known VVN for IN activism NN than IN high JJ quality NN scholarship NN , , had VHD neither DT closed JJ membership NN nor CC staff NN called VVD fellows NNS . SENT ACLU NP , , NAACP NP , , Sierra NP Club NP , , NRA NP , , AARP NP , , Common NP Cause NP , , Christian NP Coalition NP , , NOW RB , , and CC Federation NP of IN American JJ Scientists NNS . SENT 37 RB These DT two CD variables NNS provide VVP some DT weak JJ evidence NN that IN statement NN 1 CD is VBZ true JJ that IN journalists NNS indeed RB prefer VVP to TO cite VV high JJ quality NN groups NNS more RBR than IN legislators NNS do VVP . SENT When WRB we PP restrict VVP the DT sample NN only RB to TO citations NNS from IN the DT top JJ 44 CD most JJS cited VVN think NN tanks NNS recall VVP it PP is VBZ only RB these DT 44 CD that WDT receive VVP their PP$ own JJ estimate NN of IN aj NP and CC bj NP , , journalists NNS cite VVP these DT think VVP tanks NNS approximately RB 46 CD more JJR frequently RB in IN our PP$ data NNS set VVN than IN legislators NNS cite VVP them PP . SENT This DT is VBZ due JJ simply RB to TO the DT fact NN that IN the DT data NNS set VVP we PP collect VVP for IN media NNS outlets NNS is VBZ approximately RB 46 CD larger JJR than IN the DT data NNS set VVP we PP collect VVP for IN Congress NP . SENT However RB , , if IN we PP restrict VVP the DT sample NN only RB to TO the DT top JJ 44 CD think VVP tanks NNS that WDT also RB have VHP closed VVN membership NN , , then RB the DT media NNS cite VVP this DT set NN of IN groups NNS 82 CD more RBR frequently RB than IN legislators NNS do VVP . SENT Thus RB , , to TO the DT extent NN closed VVD membership NN indicates VVZ quality NN , , this DT result NN suggests VVZ that IN the DT mso NN spacerun NN . SENT yes UH Recall VV that DT high JJ ADA NP scores NNS indicate VVP that IN the DT group NN is VBZ liberal JJ . SENT The DT correlation NN between IN staff NN called VVD fellows NNS and CC the DT average JJ ADA NP score NN is VBZ also RB negative JJ , , specifically RB . SENT 071 CD . SENT This DT evidence NN suggests VVZ that IN , , if IN anything NN , , our PP$ estimates NNS are VBP biased VVN in IN the DT direction NN of IN making VVG the DT media NNS look VVP more RBR conservative JJ than IN they PP really RB are VBP . SENT However RB , , because IN the DT correlations NNS are VBP so RB close JJ to TO zero VV , , we PP believe VVP that IN any DT bias NN is VBZ small JJ . SENT A DT final JJ anecdote NN gives VVZ some DT compelling JJ evidence NN that IN our PP$ method NN is VBZ not RB biased VVN . SENT Note NN that IN none NN of IN the DT above JJ arguments NNS suggest VVP a DT problem NN with IN the DT way NN our PP$ method NN ranks VVZ media NN outlets NNS . SENT Now RB , , suppose VVP that IN there EX is VBZ no DT problem NN with IN the DT rankings NNS , , yet RB our PP$ method NN is VBZ plagued VVN with IN a DT significant JJ bias NN that WDT systematically RB causes VVZ media NN outlets NNS to TO appear VV more RBR liberal JJ conservative NN than IN they PP really RB are VBP . SENT If IN so RB , , then RB this DT means NN that IN the DT three CD outlets NNS we PP find VVP to TO be VB most RBS centrist JJ Newshour NP with IN Jim NP Lehrer NP , , Good NP Morning NP America NP , , and CC Newsnight NP with IN Aaron NP Brown NP are VBP actually RB consast JJ four CD broad JJ empirical JJ regularities NNS emerge VVP from IN our PP$ results NNS . SENT In IN this DT section NN we PP document VV the DT regularities NNS and CC analyze VV their PP$ significance NN for IN some DT theories NNS about IN the DT industrial JJ organization NN of IN the DT news NN industry NN . SENT First RB , , we PP find VVP a DT systematic JJ tendency NN for IN the DT U NP . SENT S NP . SENT media NNS outlets NNS to TO slant VV the DT news NN to TO the DT left NN . SENT As IN mentioned VVN earlier RBR , , this DT is VBZ inconsistent JJ with IN basic JJ spatial JJ models NNS of IN firm NN location NN such JJ as IN Harold NP Hotelling NP s VVZ 1929 CD and CC others NNS . SENT In IN such JJ models NNS the DT median JJ firm NN locates VVZ at IN the DT ideal JJ location NN of IN the DT median JJ consumer NN , , which WDT our PP$ results NNS clearly RB do VVP not RB support VV . SENT Some DT scholars NNS have VHP extended VVN the DT basic JJ spatial JJ model NN to TO provide VV a DT theory NN why WRB the DT media NNS could MD be VB systematically RB biased VVN . SENT For IN instance NN , , James NP Hamilton NP 2004 CD notes NNS that IN news NN producers NNS may MD prefer VV to TO cater VV to TO some DT consumers NNS more RBR than IN others NNS . SENT In IN particular JJ , , Hamilton NP notes NNS that IN young JJ females NNS tend VVP to TO be VB one CD of IN the DT most RBS marginal JJ news NN consumers NNS i NP . SENT e SYM . SENT they PP are VBP the DT most RBS willing JJ to TO switch VV to TO activities NNS besides IN reading VVG or CC watching VVG the DT news NN . SENT Further RBR , , this DT group NN often RB makes VVZ the DT consumption NN decisions NNS for IN the DT household NN . SENT For IN these DT two CD reasons NNS , , advertisers NNS are VBP willing JJ to TO pay VV more RBR to TO outlets NNS that WDT reach VVP this DT group NN . SENT Since IN this DT young JJ females NNS tend VVP to TO be VB more RBR liberal JJ on IN average NN , , a DT news NN outlet NN may MD want VV to TO slant VV its PP$ coverage NN to TO the DT left NN . SENT Thus RB , , according VVG to TO Hamilton NP s PP theory NN , , U NP . SENT S NP . SENT news NN outlets NNS slant VV their PP$ news NN coverage NN leftward RB , , not RB in IN spite NN of IN consumer NN demand NN , , but CC because IN of IN it PP . SENT 38 CD A NP more JJR compelling JJ explanation NN for IN the DT liberal JJ slant NN of IN news NN outlets NNS , , in IN our PP$ view NN , , involves VVZ production NN factors NNS , , not RB demand NN factors NNS . SENT As IN Daniel NP Sutter NP 2001 CD has VHZ noted VVN , , journalists NNS might MD systematically RB have VH a DT taste NN to TO slant VV their PP$ stories NNS to TO the DT left NN . SENT Indeed RB , , this DT is VBZ consistent JJ with IN the DT survey NN evidence NN that IN we PP noted VVD earlier RBR . SENT As IN a DT consequence NN , , If IN the DT majority NN of IN journalists NNS have VHP left VVN of IN center NN views NNS , , liberal JJ news NN might MD cost VV less RBR to TO supply VV than IN unbiased JJ news NN 444 CD . SENT David NP Baron NP 2004 CD constructs NNS a DT rigorous JJ mathematical JJ model NN along IN these DT lines NNS . SENT In IN his PP$ model NN journalists NNS are VBP driven VVN , , not RB just RB by IN money NN , , but CC also RB a DT desire NN to TO influence VV their PP$ readers NNS or CC viewers NNS . SENT Baron NP shows VVZ that DT profit NN maximizing VVG firms NNS may MD choose VV to TO allow VV reporters NNS to TO slant VV their PP$ stories NNS , , and CC consequently RB in IN equilibrium NN the DT media NNS will MD have VH a DT systematic JJ bias NN . SENT 39 CD A NP second JJ empirical JJ regularity NN is VBZ that IN the DT media NNS outlets NNS that IN we PP examine VVP are VBP fairly RB centrist JJ relative NN to TO members NNS of IN Congress NP . SENT For IN instance NN , , as IN Figure NN 2 CD shows NNS , , all DT outlets NNS but CC one CD have VHP ADA NP scores NNS between IN the DT average JJ Democrat NP and CC average JJ Republican NP in IN Congress NP . SENT In IN contrast NN , , it PP is VBZ reasonable JJ to TO believe VV that IN at IN least JJS half PDT the DT voters NNS consider VVP themselves PP more RBR extreme JJ than IN the DT party NN averages NNS . SENT 40 CD If IN so RB , , then RB a DT basic JJ spatial JJ model NN , , where WRB are NN firms NNS are VBP constrained VVN to TO charge VV the DT same JJ exogenous JJ price NN , , implies VVZ that IN approximately RB half PDT the DT media NNS outlets NNS should MD choose VV a DT slant NN outside IN the DT party NN averages NNS . SENT 41 CD Clearly RB , , our PP$ results NNS do VVP not RB support VV this DT prediction NN . SENT Moreover RB , , when WRB we PP add VVP price NN competition NN to TO the DT basic JJ spatial JJ model NN , , then RB , , as IN Mullainathan NP and CC Schleifer NP 2003 CD show NN , , even RB fewer JJR media NNS outlets NNS should MD be VB centrist JJ . SENT Specifically RB , , their PP$ two CD firm NN model NN predicts VVZ that IN both DT media NNS firms NNS should MD choose VV slants NNS that WDT are VBP outside IN the DT preferred JJ slants NNS of IN all DT consumers NNS . SENT The DT intuition NN is VBZ that IN in IN the DT first JJ round NN , , when WRB firms NNS choose VVP locations NNS , , they PP want VVP to TO differentiate VV their PP$ products NNS significantly RB , , so RB in IN the DT next JJ round NN they PP will MD have VH less JJR incentive NN to TO compete VV on IN price NN . SENT Given VVN this DT theoretical JJ result NN , , it PP is VBZ puzzling JJ that IN media NNS outlets NNS in IN the DT U NP . SENT S NP . SENT are VBP not RB more RBR heterogeneous JJ . SENT We PP suspect VVP that IN , , once RB again RB , , the DT reason NN may MD lie VV with IN production NN factors NNS . SENT For IN instance NN , , one CD possibility NN may MD involve VV the DT sources NNS for IN news NN stories NNS what WP one PP could MD consider VV as IN the DT raw JJ materials NNS of IN the DT news NN industry NN . SENT If IN a DT news NN outlet NN is VBZ too RB extreme JJ , , many JJ of IN the DT newsmakers NNS may MD refuse VV to TO grant VV interviews NNS to TO the DT reporters NNS . SENT A DT third JJ empirical JJ regularity NN involves VVZ the DT question NN whether IN reporters NNS will MD be VB faithful JJ agents NNS of IN the DT owners NNS of IN the DT firms NNS for IN which WDT they PP work VVP . SENT That DT is VBZ , , will MD the DT slant NN of IN their PP$ news NN stories NNS reflect VVP their PP$ own JJ ideological JJ preferences NNS or CC the DT firm NN s PP owners NNS . SENT The DT conventional JJ wisdom NN , , at IN least JJS among IN left JJ wing NN commentators NNS , , is VBZ that IN the DT latter NN is VBZ true JJ . SENT For IN instance NN , , Eric NP Alterman NP 2003 CD entitles VVZ a DT chapter NN of IN his PP$ book NN You PP re IN Only RB as RB Liberal JJ as IN the DT Man NN Who WP Owns VVZ You PP . SENT A DT weaker JJR assertion NN is VBZ that IN the DT particular JJ news NN outlet NN will MD be VB a DT faithful JJ agent NN of IN the DT firm NN that WDT owns VVZ it PP . SENT However RB , , our PP$ results NNS provide VVP some DT weak JJ evidence NN that IN this DT is VBZ not RB true JJ . SENT For IN instance NN , , although IN Time NP magazine NN and CC CNN NP s PP Newsnight NP are VBP owned VVN by IN the DT same JJ firm NN Time NP Warner NP , , their PP$ ADA NP scores NNS differ VVP substantially RB , , by IN 9 CD . SENT 4 CD points NNS . SENT 42 CD Further RBR , , almost RB half NN of IN the DT other JJ outlets NNS have VHP scores NNS between IN the DT scores NNS of IN Newsnight NP and CC Time NP Magazine NP . SENT A DT fourth JJ regularity NN concerns NNS the DT question NN whether IN one CD should MD expect VV a DT government NN funded JJ news NN outlet NN to TO be VB more RBR liberal JJ than IN a DT privately RB funded VVN outlet NN . SENT Radical JJ democratic JJ media NNS scholars NNS Robert NP McChesney NP and CC Ben NP Scott NP claim VVP that IN it PP will MD . SENT For IN instance NN , , they PP note VVP Commercial JJ journalism NN has VHZ more RBR often RB served VVN the DT minority NN interests NNS of IN dominant JJ political JJ , , military JJ , , and CC business NN concerns NNS than IN it PP has VHZ the DT majority NN interests NNS of IN disadvantaged JJ social JJ classes NNS 2004 CD , , 4 CD . SENT And CC conservatives NNS , , who WP frequently RB complain VVP that IN NPR NP is VBZ far RB left VVN , , seem VVP also RB to TO agree VV . SENT However RB , , our PP$ results NNS do VVP not RB support VV such JJ claims NNS . SENT If IN anything NN , , the DT government NN funded VVN outlets NNS in IN our PP$ sample NN NPR NP s PP Morning NP Edition NN and CC Newshour NP with IN Jim NP Lehrer NP have VHP a DT slightly RB lower JJR average JJ ADA NP score NN 61 CD . SENT 0 CD , , than IN the DT private JJ outlets NNS in IN our PP$ sample NN 62 CD . SENT 8 CD . SENT 43 CD Related NP , , some DT claim NN that IN a DT free JJ market NN system NN of IN news NN will MD produce VV less JJR diversity NN of IN news NN than IN a DT government NN run NN system NN . SENT However RB , , again RB , , our PP$ results NNS do VVP not RB support VV such PDT a DT claim NN . SENT The DT variance NN of IN the DT ADA NP scores NNS of IN the DT privately RB run VVN outlets NNS is VBZ substantially RB higher JJR 131 CD . SENT 3 CD than IN the DT variance NN of IN the DT two CD government NN funded JJ outlets NNS that IN we PP examine VVP 55 CD . SENT 1 CD . SENT In IN interpreting VVG some DT of IN the DT above JJ regularities NNS , , especially RB perhaps RB the DT latter JJ two CD , , we PP advise VVP caution NN . SENT For IN instance NN , , with IN regard NN to TO our PP$ comparisons NNS of IN government NN funded VVN vs NP . SENT privately RB funded VVN news NN outlets NNS , , we PP should MD emphasize VV that IN our PP$ sample NN of IN government NN funded VVN outlets NNS is VBZ small JJ only RB two CD , , and CC our PP$ total JJ sample NN of IN news NN outlets NNS might MD not RB be VB representative JJ of IN all DT news NN outlets NNS . SENT Related VVN , , in IN our PP$ attempts NNS to TO explain VV these DT patterns NNS , , we PP in IN no DT way NN claim VVP to TO have VH provided VVN the DT last JJ word NN on IN a DT satisfactory JJ theory NN . SENT Nor CC do VVP we PP claim VVP to TO have VH performed VVN an DT exhaustive JJ review NN of IN potential JJ theories NNS in IN the DT literature NN . SENT Rather RB , , the DT main JJ goal NN of IN our PP$ research NN is VBZ simply RB to TO demonstrate VV that IN it PP is VBZ possible JJ to TO create VV an DT objective JJ measure NN of IN the DT slant NN of IN the DT news NN . SENT Once RB this DT is VBZ done VVN , , as IN we PP hope VVP we PP have VHP demonstrated VVN in IN this DT section NN , , it PP is VBZ easy JJ to TO raise VV a DT host NN of IN theoretical JJ issues NNS to TO which WDT such JJ a DT measure NN can MD be VB applied VVN . SENT Appendix NN We PP believe VVP that IN the DT most RBS appropriate JJ model NN specification NN is VBZ the DT one NN that IN we PP used VVD to TO generate VV estimates NNS in IN Table NN 3 CD . SENT However RB , , in IN this DT Appendix NN we PP show VVP how WRB the DT estimates NNS change VVP when WRB we PP adopt VVP alternative JJ specifications NNS . SENT Recall NN , , that WDT we PP excluded VVD observations NNS in IN which WDT the DT journalist NN or CC legislator NN gave VVD an DT ideological JJ label NN to TO the DT think NN tank NN or CC policy NN group NN . SENT The DT first JJ column NN of IN Table NN A DT 1 CD lists NNS ADA NP estimates VVZ when WRB instead RB we PP include VVP these DT observations NNS , , while IN maintaining VVG all PDT the DT other JJ assumptions NNS that IN we PP used VVD to TO create VV Table NN 3 CD e NN . SENT g NN . SENT that IN we PP use VVP 44 CD actual JJ think NN tanks NNS and CC 6 CD mega JJ think NN tanks NNS , , etc FW . SENT As IN mentioned VVN earlier RBR , , when WRB we PP include VVP labeled VVN observations NNS , , the DT main JJ effect NN is VBZ to TO make VV the DT media NNS outlets NNS appear VVP more RBR centrist JJ . SENT For IN example NN , , note NN that IN this DT causes VVZ the DT New NP York NP Times NP score NN to TO become VV more RBR conservative JJ by IN about IN 3 CD . SENT 8 CD points NNS . SENT while IN it PP makes VVZ the DT score NN of IN the DT Fox NP News NP Special JJ Report NN become VVN more RBR liberal JJ by IN 1 CD . SENT 8 CD points NNS . SENT In IN column NN 2 CD we PP report VVP the DT results NNS when WRB we PP exclude VVP citations NNS of IN the DT ACLU NP while IN we PP maintain VVP all PDT the DT other JJ model NN specifications NNS we PP used VVD to TO construct VV Table NN 3 CD , , including VVG the DT decision NN to TO omit VV labeled VVN observations NNS . SENT In IN columns NNS 3 CD to TO 8 CD we PP report VVP the DT results NNS when WRB , , instead RB of IN using VVG 44 CD actual JJ think NN tanks NNS and CC 6 CD mega JJ think NN tanks NNS , , we PP use VVP 48 CD respectively RB , , 47 CD , , 46 CD , , 45 CD , , 43 CD , , and CC 42 CD actual JJ and CC 2 CD respectively RB 3 CD , , 4 CD , , 5 CD , , 7 CD , , and CC 8 CD mega JJ think NN tanks NNS . SENT In IN columns NNS 1 CD to TO 4 CD of IN Table NN A DT 2 CD we PP report VVP the DT results NNS when WRB , , instead RB of IN using VVG 44 CD actual JJ think NN tanks NNS and CC 6 CD mega JJ think NN tanks NNS , , we PP use VVP 54 CD respectively RB 64 CD , , 74 CD , , and CC 84 CD actual JJ think NN tanks NNS and CC 6 CD mega JJ think NN tanks NNS . SENT That DT is VBZ , , we PP let VVD the DT total JJ number NN of IN think NN tanks NNS that IN we PP use VVP change NN to TO 60 CD , , 70 CD , , 80 CD , , and CC 90 CD . SENT In IN column NN 5 CD of IN Table NN A DT 2 CD we PP use VVP sentences NNS as IN the DT level NN of IN observation NN instead RB of IN citations NNS . SENT That DT is VBZ , , for IN instance NN , , suppose VVP that IN a DT news NN outlet NN lists VVZ a DT four CD sentence NN quotation NN from IN a DT member NN of IN a DT think NN tank NN . SENT In IN the DT earlier JJR analysis NN we PP would MD count VV this DT as IN one CD observation NN . SENT However RB , , the DT estimates NNS in IN column NN 5 CD would MD treat VV this DT as IN four CD observations NNS . SENT One CD problem NN with IN this DT specification NN is VBZ that IN the DT data NNS are VBP very RB lumpy JJ that WDT is VBZ , , some DT quotes NNS contain VVP an DT inordinate JJ number NN of IN sentences NNS , , which WDT cause VVP some DT anomalies NNS . SENT One PP is VBZ that IN some DT relatively RB obscure JJ think VVP tanks NNS become VV some DT of IN the DT most RBS cited VVN under IN this DT specification NN . SENT For IN instance NN , , the DT Alexis NP de NP Tocqueville NP Institute NP , , which WDT most JJS readers NNS would MD agree VV is VBZ not RB one CD of IN the DT most RBS well RB known VVN and CC prominent JJ think VVP tanks NNS , , is VBZ the DT 13 JJ th NN most RBS cited VVN think NN tank NN by IN members NNS of IN Congress NP , , when WRB we PP use VVP sentences NNS as IN the DT level NN of IN observation NN . SENT It PP is VBZ the DT 58 JJ th NN most RBS cited VVN , , when WRB we PP use VVP citations NNS as IN the DT level NN of IN observation NN . SENT Members NNS of IN Congress NP cited VVD it PP only RB 35 CD times NNS , , yet CC they PP cited VVD an DT average NN of IN 39 CD sentences NNS for IN each DT citation NN . SENT This DT compares VVZ to TO approximately RB five CD sentences NNS per IN citation NN for IN the DT other JJ think NN tanks NNS . SENT Meanwhile RB , , it PP is VBZ one CD of IN the DT 30 CD least JJS cited VVD think VV tank NN by IN the DT media NNS . SENT 44 CD A NP related JJ problem NN is VBZ that IN these DT data NNS are VBP serially RB correlated VVN . SENT That DT is VBZ , , for IN instance NN , , if IN a DT given VVN observation NN for IN the DT New NP York NP Times NP is VBZ a DT citation NN to TO the DT Brookings NP Institution NP , , then RB the DT probability NN is VBZ high JJ that IN the DT next JJ observation NN will MD also RB be VB a DT citation NN to TO the DT same JJ think NN tank NN since IN the DT average JJ citation NN contains VVZ more JJR than IN one CD sentence NN . SENT However RB , , the DT likelihood NN function NN that IN we PP use VVP assumes VVZ that IN the DT observations NNS are VBP not RB serially RB correlated VVN . SENT Finally RB , , related VVN to TO these DT problems NNS , , the DT estimates NNS from IN this DT specification NN sometimes RB are VBP in IN stark JJ disagreement NN with IN common JJ wisdom NN . SENT For IN instance NN , , the DT estimates NNS imply VVP that IN the DT Washington NP Times NP is VBZ more RBR liberal JJ than IN Good NP Morning NP America NP . SENT For IN these DT reasons NNS , , we PP base VVP our PP$ conclusions NNS on IN the DT estimates NNS that WDT use VVP citations NNS as IN the DT level NN of IN observation NN , , rather RB than IN sentences NNS . SENT References NNS Alterman NP , , Eric NP . SENT 2003 CD . SENT What WP Liberal NP Media NP . SENT The DT Truth NN about IN Bias NP and CC the DT News NP . SENT New NP York NP . SENT Basic JJ Books NNS . SENT Baron NP , , David NP . SENT 2004 CD . SENT Persistent JJ Media NP Bias NP . SENT Manuscript NN . SENT Stanford NP University NP . SENT Black NP , , Duncan NP . SENT 1958 CD . SENT The DT Theory NN of IN Committees NNS and CC Elections NNS . SENT London NP . SENT Cambridge NP University NP Press NP . SENT Bozell NP , , L NP . SENT B SYM . SENT , , and CC B NP . SENT H NP . SENT Baker NP . SENT 1990 CD . SENT editors NNS That IN s PP the DT Way NP It PP Isn NP t NN . SENT A DT Reference NN Guide NP to TO Media NP Bias NP . SENT Alexandria NP , , Va NP . SENT . SENT Media NP Research NP Center NP . SENT Crouse NP , , Timothy NP . SENT 1973 CD . SENT Boys NNS on IN the DT Bus NN . SENT New NP York NP . SENT Ballantine NP Books NPS . SENT Djankov NP , , Simeon NP , , Caralee NP McLiesh NP , , Tatiana NP Nenova NP , , and CC Andrei NP Shleifer NP . SENT 2003 CD . SENT Who WP Owns VVZ the DT Media NP . SENT Journal NP of IN Law NP and CC Economics NP . SENT 46 CD October NP . SENT 341 CD 81 CD . SENT Franken NP , , Al NP . SENT 2003 CD . SENT Lies NNS and CC the DT Lying VVG Liars NNS Who WP Tell VV Them NP . SENT A DT Fair NP and CC Balanced JJ Look NN at IN the DT Right NP . SENT New NP York NP . SENT Dutton NP . SENT Goff NP , , Brian NP , , and CC Robert NP Tollison NP . SENT 1990 CD . SENT Why WRB is VBZ the DT Media NN so IN Liberal NP . SENT Journal NP of IN Public NP Finance NP and CC Public NP Choice NP . SENT 1 LS . SENT 13 CD 21 CD . SENT Goldberg NP , , Bernard NP . SENT 2002 CD . SENT Bias NN . SENT A DT CBS NP Insider NP Exposes VVZ How WRB the DT Media NP Distort VVP the DT News NP . SENT Washington NP , , D NP . SENT C SYM . SENT Regnery NP . SENT Groeling NP , , Tim NP , , and CC Samuel NP Kernell NP . SENT 1998 CD . SENT Is VBZ Network NP News NP Coverage NP of IN the DT President NP Biased VVN . SENT Journal NP of IN Politics NP . SENT 60 CD November NP . SENT 1063 CD 87 CD . SENT Groseclose NP , , Tim NP , , Steven NP D NP . SENT Levitt NP , , and CC James NP M NP . SENT Snyder NP , , Jr NP . SENT 1999 CD . SENT Comparing VVG Interest NP Group NP Scores NPS across IN Time NP and CC Chambers NPS . SENT Adjusted JJ ADA NP Scores NNS for IN the DT U NP . SENT S NP . SENT Congress NP , , American NP Political NP Science NP Review NP . SENT 93 CD March NP . SENT 33 CD 50 CD . SENT Groseclose NP , , Tim NP , , and CC Jeff NP Milyo NP . SENT 2004 CD . SENT Response NN to TO Liberal NP Bias NP , , Noch NP Einmal NP . SENT http NN . SENT itre NN . SENT cis NNS . SENT upenn NP . SENT edu NN myl NN languagelog NN archives NN 001301 CD . SENT html NN . SENT Hamilton NP , , James NP . SENT 2004 CD . SENT All PDT the DT News NN That IN s PP Fit VVP to TO Sell VV . SENT How WRB the DT Market NN Transforms VVZ Information NP into IN News NP . SENT Princeton NP , , N NP . SENT J NP . SENT . SENT Princeton NP University NP Press NP . SENT Herman NP , , Edward NP S NP . SENT , , and CC Noam NP Chomsky NP . SENT 1988 CD . SENT Manufacturing NP Consent NP . SENT The DT Political JJ Economy NN of IN the DT Mass NP Media NP . SENT New NP York NP . SENT Pantheon NN Books NNS . SENT Hotelling NP , , Harold NP . SENT 1929 CD . SENT Stability NN In IN Competition NN . SENT Economic NP Journal NP . SENT 39 CD . SENT 41 CD 57 CD . SENT Irvine NP , , Reed NP , , and CC Cliff NP Kincaid NP . SENT 2001 CD . SENT Post NN Columnist NN Concerned VVN About IN Media NP Bias NP . SENT September NP 17 CD , , 2001 CD . SENT http NN . SENT www JJ . SENT aim NN . SENT org NP media NNS monitor VVP A DT 900 CD 0 CD 2 CD 0 CD C NP website NN for IN Accuracy NN in IN Media NP . SENT Jamieson NP , , Kathleen NP Hall NP . SENT 2000 CD . SENT Everything NN You PP Think VVP You PP Know VVP About IN Politics NNS and CC Why WRB You PP re IN Wrong JJ . SENT New NP York NP . SENT Basic JJ Books NNS . SENT Judge NP , , George NP G NP . SENT , , W NP . SENT E SYM . SENT Griffiths NP , , R NP . SENT Carter NP Hill NP , , Helmut NP Lutkepohl NP , , and CC Tsoung NP Chao NP Lee NP . SENT 1985 CD . SENT The DT Theory NN and CC Practice NN of IN Econometrics NNS . SENT New NP York NP . SENT John NP Wiley NP and CC Sons NP . SENT Kurtz NP , , Howard NP . SENT 2004 CD . SENT Fewer JJR Republicans NPS Trust NP the DT News NP , , Survey NP Finds VVZ . SENT Washington NP Post NP . SENT P NN . SENT C NP 01 CD . SENT June NP 9 CD , , 2004 CD . SENT Lichter NP , , S NP . SENT R SYM . SENT , , S NP . SENT Rothman NP , , and CC L NP . SENT S NP . SENT Lichter NP . SENT 1986 CD . SENT The DT Media NP Elite NP . SENT Bethesda NP , , MD NP . SENT Adler NP and CC Adler NP . SENT Lott NP , , John NP R NP , , Jr NP . SENT 2003 CD . SENT The DT Bias NP Against NP Guns NNS . SENT Washington NP , , D NP . SENT C SYM . SENT . SENT Regnery NP Publishing NP , , Inc NP . SENT Lott NP , , John NP R NN . SENT , , Jr NP . SENT . SENT 1999 CD . SENT Public NP Schooling NP , , Indoctrination NN , , and CC Totalitarianism NN . SENT Journal NP of IN Political NP Economy NP . SENT 6 CD , , part NN 2 CD . SENT S NP 127 CD S NP 157 CD . SENT Lott NP , , John NP R NN . SENT , , Jr NP . SENT , , and CC Kevin NP A NP . SENT Hasset NP . SENT 2004 CD . SENT Is VBZ Newspaper NP Coverage NP of IN Economic NP Events NNS Politically RB Biased VVN . SENT Manuscript NN . SENT July NP 20 CD , , 2004 CD . SENT American NP Enterprise NP Institute NP . SENT McChesney NP , , Robert NP , , and CC Ben NP Scott NP . SENT 2004 CD . SENT Our PP$ Unfree JJ Press NP . SENT 100 CD Years NNS of IN Radical JJ New NP York NP . SENT Academic NP Press NP . SENT Mullainathan NP , , Sendhil NP , , and CC Andrei NP Shleifer NP . SENT 2003 CD . SENT The DT Market NN for IN News NP . SENT Manuscript NN . SENT Harvard NP University NP . SENT Nunberg NP , , Geoffrey NP . SENT 2004 CD . SENT Liberal NP Bias NP , , Noch NP Einmal NP . SENT http NN . SENT itre NN . SENT cis NNS . SENT upenn NP . SENT edu NN myl NN languagelog NN archives NN 001169 CD . SENT html NP Parenti NP , , Michael NP . SENT 1986 CD . SENT Inventing VVG Reality NN . SENT The DT Politics NNS of IN the DT Mass NP Media NP . SENT New NP York NP . SENT St NP . SENT Martin NP s PP Press NP . SENT Povich NP , , Elaine NP . SENT 1996 CD . SENT Partners NNS and CC Adversaries NNS . SENT The DT Contentious JJ Connection NN Between IN Congress NP and CC the DT Media NP . SENT Arlington NP , , VI NP . SENT Freedom NP Forum NP . SENT Sperry NP , , Paul NP . SENT 2002 CD . SENT Myth NN of IN the DT Conservative NP Wall NP Street NP Journal NP . SENT Sutter NP , , Daniel NP . SENT 2001 CD . SENT Can MD the DT Media NN Be VB So RB Liberal JJ . SENT The DT Economics NP of IN Media NP Bias NP . SENT The DT Cato NP Journal NP . SENT 20 CD Winter NPS . SENT 431 CD 51 CD . SENT Sutter NP , , Daniel NP . SENT 2002 CD . SENT Advertising NP and CC Political NP Bias NP in IN the DT Media NP , , American NP Journal NP of IN Economics NP and CC Sociology NN , , 61 CD 3 CD . SENT 725 CD 745 CD . SENT Sutter NP , , Daniel NP . SENT 2004 CD . SENT An DT Indirect JJ Test NN of IN the DT Liberal NP Media NP Thesis NN Using VVG Newsmagazine NN Circulation NN . SENT Manuscript NN . SENT University NP of IN Oklahoma NP . SENT Weaver NP , , D NP . SENT H NP . SENT and CC G NP . SENT C SYM . SENT Wilhoit NP . SENT 1996 CD . SENT American JJ Journalist NN in IN the DT 1990 CD s PP . SENT Mahwah NP , , NJ NP . SENT Lawrence NP Erlbaum NP . SENT Woodward NP , , Bob NP . SENT 1994 CD . SENT The DT Agenda NN . SENT Inside IN the DT Clinton NP White NP House NP . SENT New NP York NP . SENT Simon NP Schuster NP . SENT Table NN 1 CD . SENT The DT 50 CD Most JJS Cited VVD Think VV Tanks NP and CC Policy NP Groups NNS by IN the DT Media NN in IN our PP$ Sample NP Ave NP . SENT Score NN of IN Number NP of IN Number NP of IN legislator NN citing VVG Citations NNS by IN Citations NNS by IN Think VVP Tank NP Policy NP Group NP think VVP tank NN Legislators NNS Media NP Outlets NNS 1 CD Brookings NP Institution NP 53 CD . SENT 3 CD 320 JJ 1392 CD 2 CD American NP Civil NP Liberties NPS Union NP 49 CD . SENT 8 CD 273 JJ 1073 CD 3 CD NAACP NP 75 CD . SENT 4 CD 134 JJ 559 CD 4 CD Center NP for IN Strategic NP and CC International NP Studies NP 46 CD . SENT 3 CD 79 JJ 432 CD 5 CD Amnesty NN International NP 57 CD . SENT 4 CD 394 JJ 419 CD 6 CD Council NP on IN Foreign NP Relations NP 60 CD . SENT 2 CD 45 JJ 403 CD 7 CD Sierra NP Club NP 68 CD . SENT 7 CD 376 JJ 393 CD 8 CD American NP Enterprise NP Institute NP 36 CD . SENT 6 CD 154 JJ 382 CD 9 CD RAND NN Corporation NP 60 CD . SENT 4 CD 352 JJ 350 CD 10 CD National NP Rifle NP Association NP 45 CD . SENT 9 CD 143 JJ 336 CD 11 CD American NP Association NP of IN Retired NP Persons NP 66 CD . SENT 0 CD 411 JJ 333 CD 12 CD Carnegie NP Endowment NP for IN International NP Peace NP 51 CD . SENT 9 CD 26 JJ 328 CD 13 CD Heritage NP Foundation NP 20 CD . SENT 0 CD 369 JJ 288 CD 14 CD Common JJ Cause NP 69 CD . SENT 0 CD 222 JJ 287 CD 15 CD Center NN for IN Responsive JJ Politics NP 66 CD . SENT 9 CD 75 JJ 264 CD 16 CD Consumer NP Federation NP of IN America NP 81 CD . SENT 7 CD 224 JJ 256 CD 17 CD Christian JJ Coalition NP 22 CD . SENT 6 CD 141 JJ 220 CD 18 CD Cato NP Institute NP 36 CD . SENT 3 CD 224 JJ 196 CD 78 CD . SENT 9 CD 62 JJ 195 CD 20 CD Institute NP for IN International NP Economics NP 48 CD . SENT 8 CD 61 JJ 194 CD 21 CD Urban NP Institute NP 73 CD . SENT 8 CD 186 JJ 187 CD 22 CD Family NP Research NP Council NP 20 CD . SENT 3 CD 133 JJ 160 CD 23 CD Federation NP of IN American NP Scientists NNS 67 CD . SENT 5 CD 36 JJ 139 CD 24 CD Economic NP Policy NP Institute NP 80 CD . SENT 3 CD 130 JJ 138 CD 25 CD Center NP on IN Budget NP and CC Policy NP Priorities NNS 88 CD . SENT 3 CD 224 JJ 115 CD 26 CD National NP Right NP to TO Life NP Committee NP 21 CD . SENT 6 CD 81 JJ 109 CD 27 CD Electronic NP Privacy NN Information NP Center NP 57 CD . SENT 4 CD 19 JJ 107 CD 28 CD International NP Institute NP for IN Strategic NP Studies NP 41 CD . SENT 2 CD 16 JJ 104 CD 29 CD World NP Wildlife NP Fund NP 50 CD . SENT 4 CD 130 JJ 101 CD 30 CD Cent NN . SENT for IN Strategic NP and CC Budgetary JJ Assessments NNS 33 CD . SENT 9 CD 7 CD 89 CD 31 CD Nat NP . SENT Abort VV . SENT and CC Reproductive NP Rights NPS Action NP Lg NP . SENT 71 CD . SENT 9 CD 30 JJ 88 CD 32 CD Children's NP Defense NP Fund NP 82 CD . SENT 0 CD 231 JJ 78 CD 33 CD Employee NP Benefit NP Research NP Institute NP 49 CD . SENT 1 CD 41 JJ 78 CD 34 CD Citizens NPS Against IN Government NP Waste NP 36 CD . SENT 3 CD 367 JJ 76 CD 35 CD People NNS for IN the DT American JJ Way NP 76 CD . SENT 1 CD 63 JJ 76 CD 36 CD Environmental NP Defense NP Fund NP 66 CD . SENT 9 CD 137 JJ 74 CD 37 CD Economic NP Strategy NP Institute NP 71 CD . SENT 9 CD 26 JJ 71 CD 38 CD People NNS for IN the DT Ethical JJ Treatment NP of IN Animals NPS 73 NP . SENT 4 LS 5 CD 70 CD 39 CD Americans NPS for IN Tax NP Reform NP 18 CD . SENT 7 CD 211 JJ 67 CD 40 CD Citizens NPS for IN Tax NP Justice NP 87 CD . SENT 8 CD 92 JJ 67 CD 41 CD National NP Federation NP of IN Independent NP Businesses NNS 73 CD 64 CD 43 CD National NP Taxpayers NPS Union NP 34 CD . SENT 3 CD 566 JJ 63 CD 44 CD Stimson NP Center NP 63 CD . SENT 6 CD 26 CD 63 CD p NN class NN MsoNormal NN align VV center NN style NN text NN align VV . SENT center NN 79 CD . SENT 0 CD 28 JJ 61 CD 46 CD Handgun NN Control NP , , Inc NP . SENT 77 CD . SENT 2 CD 58 JJ 61 CD 47 CD Hoover NP Institution NP 36 CD . SENT 5 CD 35 JJ 61 CD 48 CD Nixon NP Center NP 21 CD . SENT 7 CD 6 CD 61 CD 49 CD American JJ Conservative JJ Union NP 16 CD . SENT 1 CD 43 JJ 56 CD 50 CD Manhattan NP Institute NP 32 CD . SENT 0 CD 18 CD 54 CD Table NN 2 CD . SENT Average JJ Adjusted JJ Scores NNS of IN Legislators NNS Legislator NN Ave NP . SENT Score NN Maxine NP Waters NP D NP . SENT Calif NP . SENT 99 CD . SENT 6 CD Ted NP Kennedy NP D NP . SENT Mass NP . SENT 88 CD . SENT 8 CD John NP Kerry NP D NP . SENT Mass NP . SENT 87 CD . SENT 6 CD average JJ Democrat NP 84 CD . SENT 3 CD Tom NP Daschle NP D NP . SENT S NP . SENT D SYM . SENT 80 CD . SENT 9 CD Joe NP Lieberman NP D NP Ct NP . SENT 74 CD . SENT 2 CD Constance NP Morella NP R NP Md NP . SENT 68 CD . SENT 2 CD Ernest NP Hollings NP D NP S NP . SENT C SYM . SENT 63 CD . SENT 7 CD John NP Breaux NP D NP La NP . SENT 59 CD . SENT 5 CD Christopher NP Shays NP R NP Ct NP . SENT 54 CD . SENT 6 CD Arlen NP Specter NP R NP Pa NP . SENT 51 CD . SENT 3 CD James NP Leach NP R NP Iowa NP 50 CD . SENT 3 CD Howell NP Heflin NP D NP Ala NP . SENT 49 CD . SENT 7 CD Sam NP Nunn NP D NP Ga NP . SENT 48 CD . SENT 0 CD Dave NP McCurdy NP D NP Ok NP . SENT 46 CD . SENT 9 CD Olympia NP Snowe NP R NP Me PP . SENT 43 CD . SENT 0 CD Susan NP Collins NP R NP Me PP . SENT 39 CD . SENT 3 CD Charlie NP Stenholm NP D NP Tex NP . SENT 36 CD . SENT 1 CD Rick NP Lazio NP R NP N NP . SENT Y NP . SENT 35 CD . SENT 8 CD Nathan NP Deal NP D NP Ga NP . SENT 21 CD . SENT 5 CD Joe NP Scarborough NP R NP . SENT Fla NP . SENT 17 CD . SENT 7 CD average JJ Republican NP 16 CD . SENT 1 CD John NP McCain NP R NP . SENT Ariz NP . SENT 12 CD . SENT 7 CD Bill NP Frist NP R NP . SENT Tenn NP . SENT 10 CD . SENT 3 CD Tom NP Delay NP R NP . SENT Tex NP . SENT 4 LS . SENT 7 CD Note NN . SENT The DT table NN lists VVZ average JJ adjusted JJ ADA NP scores NNS . SENT The DT method NN for IN adjusting VVG scores NNS is VBZ described VVN in IN Groseclose NP , , Levitt NP , , and CC Snyder NP 1999 CD . SENT Scores NNS are VBP converted VVN to TO the DT 1999 CD House NP scale NN . SENT The DT scores NNS listed VVN are VBP an DT average NN of IN the DT legislator's JJ scores NNS during IN the DT 1993 CD 1999 CD period NN . SENT The DT one CD exception NN is VBZ Nathan NP Deal NP . SENT His PP$ score NN is VBZ the DT average NN for IN the DT two CD years NNS of IN the DT sample NN in IN which WDT he PP was VBD a DT Democrat NP , , 1993 CD and CC 1994 CD . SENT He PP switched VVD parties NNS in IN 1995 CD . SENT Deal NN is VBZ the DT most RBS conservative JJ Democrat NP in IN the DT sample NN . SENT Constance NP Morella NP is VBZ the DT most RBS liberal JJ Republican NP in IN the DT sample NN . SENT Table NN 3 CD . SENT Results NNS from IN Maximum JJ Likelihood NN Estimation NN Period NN of IN ADA NP Standard NP Observation NN Score NN Error NN ABC NP Good NP Morning NP America NP 6 CD 27 CD 97 JJ 6 CD 26 CD 03 CD 56 CD . SENT 1 CD 3 CD . SENT 2 CD ABC NP World NP News NP Tonight RB 1 CD 1 CD 94 JJ 6 CD 26 CD 03 CD 61 CD . SENT 0 CD 1 CD . SENT 7 CD CBS NP Early NP Show NP 11 CD 1 CD 99 JJ 6 CD 26 CD 03 CD 66 CD . SENT 6 CD 4 CD . SENT 0 CD CBS NP Evening NP News NP 1 CD 1 CD 90 JJ 6 CD 26 CD 03 CD 73 CD . SENT 7 CD 1 CD . SENT 6 CD CNN NP NewsNight NP with IN Aaron NP Brown NP 11 CD 9 CD 01 JJ 2 CD 5 CD 04 CD 56 CD . SENT 0 CD 4 CD . SENT 1 CD Fox NP News NP Special JJ Report NP with IN Brit NP Hume NP 6 CD 1 CD 98 JJ 6 CD 26 CD 03 CD 39 CD . SENT 7 CD 1 CD . SENT 9 CD LA NP Times NP 6 CD 28 CD 02 JJ 12 CD 29 CD 02 CD 70 CD . SENT 0 CD 2 CD . SENT 2 CD NBC NP Nightly JJ News NP 1 CD 1 CD 97 JJ 6 CD 26 CD 03 CD 61 CD . SENT 6 CD 1 CD . SENT 8 CD NBC NP Today NN Show NP 6 CD 27 CD 97 JJ 6 CD 26 CD 03 CD 64 CD . SENT 0 CD 2 CD . SENT 5 CD New NP York NP Times NP 7 CD 1 CD 01 JJ 5 CD 1 CD 02 CD 73 CD . SENT 7 CD 1 CD . SENT 6 CD Newshour NP with IN Jim NP Lehrer NP 11 CD 29 CD 99 JJ 6 CD 26 CD 03 CD 55 CD . SENT 8 CD 2 CD . SENT 3 CD Newsweek NP 6 CD 27 CD 95 JJ 6 CD 26 CD 03 CD 66 CD . SENT 3 CD 1 CD . SENT 8 CD NPR NP Morning NP Edition NN 1 CD 1 CD 92 JJ 6 CD 26 CD 03 CD 66 CD . SENT 3 CD 1 CD . SENT 0 CD Time NP Magazine NP 8 CD 6 CD 01 JJ 6 CD 26 CD 03 CD 65 CD . SENT 4 CD 4 CD . SENT 8 CD U NP . SENT S NP . SENT News NP and CC World NP Report NP 6 CD 27 CD 95 JJ 6 CD 26 CD 03 CD 65 CD . SENT 8 CD 1 CD . SENT 8 CD USA NP Today NN 1 CD 1 CD 02 JJ 9 CD 1 CD 02 CD 63 CD . SENT 4 CD 2 CD . SENT 7 CD Wall NP Street NP Journal NP 1 CD 1 CD 02 JJ 5 CD 1 CD 02 CD 85 CD . SENT 1 CD 3 CD . SENT 9 CD Washington NP Post NP 1 CD 1 CD 02 JJ 5 CD 1 CD 02 CD 66 CD . SENT 6 CD 2 CD . SENT 5 CD Washington NP Times NP 1 CD 1 CD 02 JJ 5 CD 1 CD 02 CD 35 CD . SENT 4 CD 2 CD . SENT 7 CD average NN 62 CD . SENT 6 CD Note NN . SENT The DT table NN gives VVZ our PP$ estimates NNS of IN adjusted JJ ADA NP scores NNS for IN media NNS outlets NNS , , converted VVN to TO the DT 1999 CD House NP scale NN . SENT As IN a DT comparison NN , , 50 CD . SENT 06 CD is VBZ our PP$ estimate NN of IN the DT average JJ American JJ voter NN . SENT This DT is VBZ based VVN upon IN taking VVG average JJ scores NNS of IN House NP members NNS and CC senators NNS after IN adding VVG phantom JJ extreme JJ liberal JJ D NP . SENT C SYM . SENT House NP members NNS and CC senators NNS and CC weighting NN Senate NP scores NNS by IN the DT population NN of IN the DT state NN . SENT The DT average JJ score NN of IN Republicans NNS serving VVG in IN Congress NP between IN 1995 CD and CC 1999 CD was VBD 16 CD . SENT 1 CD . SENT The DT average JJ score NN of IN Democrats NPS was VBD 84 JJ . SENT 3 LS . SENT All DT data NNS for IN the DT news NN outlets NNS came VVD from IN news NN stories NNS only RB . SENT That DT is VBZ , , we PP omitted VVD op NN eds NPS , , letters NNS to TO the DT editor NN , , and CC book NN reviews NNS . SENT Table NN 4 CD . SENT Rankings NNS Based VVN on IN Distance NP from IN Center NP Rank NP News NP Outlet NN ADA NP Score NN 1 CD Newshour NP with IN Jim NP Lehrer NP 55 CD . SENT 8 CD 2 CD CNN NP NewsNight NP with IN Aaron NP Brown NP 56 CD . SENT 0 CD 3 CD ABC NP Good NP Morning NP America NP 56 CD . SENT 1 CD 4 CD Drudge NN Report NP 60 CD . SENT 4 CD 5 CD Fox NP News NP Special JJ Report NP with IN Brit NP Hume NP 39 CD . SENT 7 CD 6 CD ABC NP World NP News NP Tonight NP 61 CD . SENT 0 CD 7 CD NBC NP Nightly JJ News NP 61 CD . SENT 6 CD 8 CD USA NP Today NP 63 CD . SENT 4 CD 9 CD NBC NP Today NN Show NP 64 CD . SENT 0 CD 10 CD Washington NP Times NP 35 CD . SENT 4 CD 11 CD Time NP Magazine NP 65 CD . SENT 4 CD 12 CD U NP . SENT S NP . SENT News NP and CC World NP Report NP 65 CD . SENT 8 CD 13 CD NPR NP Morning NP Edition NN 66 CD . SENT 3 CD 14 CD Newsweek NP 66 CD . SENT 3 CD 15 CD CBS NP Early NP Show NP 66 CD . SENT 6 CD 16 CD Washington NP Post NP 66 CD . SENT 6 CD 17 CD LA NP Times NP 70 CD . SENT 0 CD 18 CD CBS NP Evening NP News NP 73 CD . SENT 7 CD 19 CD New NP York NP Times NP 73 CD . SENT 7 CD 20 CD Wall NP Street NP Journal NP 85 CD . SENT 1 CD Note NN . SENT The DT table NN gives VVZ our PP$ method's JJ rankings NNS of IN the DT most JJS to TO least RBS centrist JJ news NN outlet NN . SENT The DT rankings NNS are VBP based VVN on IN the DT distance NN of IN the DT outlet's NNS estimated VVN ADA NP score NN from IN Table NN 3 CD to TO 50 CD . SENT 06 CD , , our PP$ estimate NN of IN the DT average JJ U NP . SENT S NP . SENT voter's NP ADA NP score NN . SENT Table NN A NP 1 CD . SENT ADA NP Scores NPS for IN Alternative NP Specifications NNS News NP Outlet NN 1 CD 2 CD 3 CD 4 CD 5 CD 6 CD 7 CD 8 CD ABC NP Good NP Morning NP America NP 56 CD . SENT 7 CD 56 CD . SENT 0 CD 55 CD . SENT 0 CD 56 CD . SENT 0 CD 59 CD . SENT 3 CD 59 CD . SENT 5 CD 56 CD . SENT 3 CD 55 CD . SENT 5 CD ABC NP World NP News NP Tonight NP 61 CD . SENT 4 CD 61 CD . SENT 3 CD 61 CD . SENT 0 CD 62 CD . SENT 0 CD 61 CD . SENT 6 CD 62 CD . SENT 4 CD 60 CD . SENT 9 CD 59 CD . SENT 8 CD CBS NP Early NP Show NP 67 CD . SENT 5 CD 67 CD . SENT 1 CD 64 CD . SENT 2 CD 67 CD . SENT 5 CD 67 CD . SENT 8 CD 68 CD . SENT 3 CD 66 CD . SENT 0 CD 64 CD . SENT 9 CD CBS NP Evening NP News NP 72 CD . SENT 1 CD 74 CD . SENT 0 CD 74 CD . SENT 0 CD 74 CD . SENT 7 CD 73 CD . SENT 2 CD 74 CD . SENT 1 CD 72 CD . SENT 8 CD 71 CD . SENT 7 CD CNN NP NewsNight NP with IN Aaron NP Brown NP 55 CD . SENT 8 CD 55 CD . SENT 8 CD 54 CD . SENT 8 CD 58 CD . SENT 0 CD 56 CD . SENT 0 CD 56 CD . SENT 4 CD 55 CD . SENT 5 CD 53 CD . SENT 3 CD Drudge NN Report NP 55 CD . SENT 3 CD 60 CD . SENT 6 CD 59 CD . SENT 1 CD 62 CD . SENT 6 CD 60 CD . SENT 8 CD 62 CD . SENT 1 CD 60 CD . SENT 267 CD . SENT 8 CD 70 CD . SENT 4 CD 69 CD . SENT 5 CD 71 CD . SENT 7 CD 70 CD . SENT 5 CD 70 CD . SENT 9 CD 69 CD . SENT 3 CD 68 CD . SENT 5 CD NBC NP Nightly JJ News NP 62 CD . SENT 1 CD 61 CD . SENT 7 CD 63 CD . SENT 1 CD 63 CD . SENT 0 CD 61 CD . SENT 3 CD 62 CD . SENT 3 CD 61 CD . SENT 2 CD 60 CD . SENT 2 CD NBC NP Today NN Show NP 64 CD . SENT 0 CD 64 CD . SENT 8 CD 64 CD . SENT 7 CD 65 CD . SENT 2 CD 65 CD . SENT 1 CD 66 CD . SENT 1 CD 63 CD . SENT 8 CD 62 CD . SENT 9 CD New NP York NP Times NP 69 CD . SENT 9 CD 74 CD . SENT 9 CD 72 CD . SENT 6 CD 74 CD . SENT 3 CD 73 CD . SENT 9 CD 74 CD . SENT 7 CD 73 CD . SENT 3 CD 71 CD . SENT 6 CD Newshour NP with IN Jim NP Lehrer NP 55 CD . SENT 1 CD 56 CD . SENT 0 CD 54 CD . SENT 4 CD 57 CD . SENT 0 CD 55 CD . SENT 8 CD 55 CD . SENT 9 CD 56 CD . SENT 0 CD 53 CD . SENT 6 CD Newsweek NP 65 CD . SENT 7 CD 66 CD . SENT 7 CD 64 CD . SENT 5 CD 67 CD . SENT 0 CD 66 CD . SENT 9 CD 67 CD . SENT 5 CD 65 CD . SENT 7 CD 64 CD . SENT 4 CD NPR NP Morning NP Edition NN 65 CD . SENT 6 CD 66 CD . SENT 9 CD 66 CD . SENT 2 CD 67 CD . SENT 5 CD 66 CD . SENT 1 CD 67 CD . SENT 1 CD 66 CD . SENT 1 CD 64 CD . SENT 6 CD Fox NP News NP Special JJ Report NP 41 CD . SENT 5 CD 39 CD . SENT 0 CD 38 CD . SENT 8 CD 41 CD . SENT 2 CD 40 CD . SENT 5 CD 40 CD . SENT 6 CD 39 CD . SENT 8 CD 38 CD . SENT 8 CD Time NP Magazine NP 68 CD . SENT 2 CD 65 CD . SENT 5 CD 62 CD . SENT 4 CD 66 CD . SENT 2 CD 64 CD . SENT 3 CD 65 CD . SENT 4 CD 64 CD . SENT 2 CD 63 CD . SENT 3 CD U NP . SENT S NP . SENT News NP and CC World NP Report NP 65 CD . SENT 2 CD 65 CD . SENT 8 CD 65 CD . SENT 4 CD 67 CD . SENT 0 CD 65 CD . SENT 8 CD 66 CD . SENT 4 CD 64 CD . SENT 8 CD 63 CD . SENT 6 CD USA NP Today NP 61 CD . SENT 7 CD 63 CD . SENT 2 CD 62 CD . SENT 5 CD 63 CD . SENT 7 CD 62 CD . SENT 8 CD 63 CD . SENT 9 CD 62 CD . SENT 4 CD 60 CD . SENT 4 CD Wall NP Street NP Journal NP 86 CD . SENT 1 CD 85 CD . SENT 1 CD 85 CD . SENT 9 CD 86 CD . SENT 2 CD 85 CD . SENT 5 CD 86 CD . SENT 4 CD 84 CD . SENT 8 CD 82 CD . SENT 5 CD Washington NP Post NP 64 CD . SENT 7 CD 67 CD . SENT 0 CD 65 CD . SENT 5 CD 67 CD . SENT 4 CD 66 CD . SENT 8 CD 67 CD . SENT 2 CD 66 CD . SENT 7 CD 64 CD . SENT 3 CD Washington NP Times NP 35 CD . SENT 7 CD 33 CD . SENT 8 CD 34 CD . SENT 4 CD 36 CD . SENT 2 CD 35 CD . SENT 3 CD 36 CD . SENT 2 CD 34 CD . SENT 8 CD 32 CD . SENT 9 CD Average NN of IN 20 CD Outlets NNS 62 CD . SENT 1 CD 62 CD . SENT 8 CD 61 CD . SENT 9 CD 63 CD . SENT 7 CD 63 CD . SENT 0 CD 63 CD . SENT 7 CD 62 CD . SENT 2 CD 60 CD . SENT 7 CD Table NN A NP 2 CD . SENT ADA NP Scores NPS for IN Alternative NP Specifications NNS News NP Outlet NN 1 CD 2 CD 3 CD 4 CD 5 CD ABC NP Good NP Morning NP America NP 56 CD . SENT 9 CD 59 CD . SENT 9 CD 60 CD . SENT 2 CD 60 CD . SENT 3 CD 45 CD . SENT 4 CD ABC NP World NP News NP Tonight NP 61 CD . SENT 6 CD 62 CD . SENT 4 CD 62 CD . SENT 9 CD 62 CD . SENT 9 CD 58 CD . SENT 7 CD CBS NP Early NP Show NP 67 CD . SENT 1 CD 68 CD . SENT 9 CD 68 CD . SENT 9 CD 69 CD . SENT 0 CD 56 CD . SENT 8 CD CBS NP Evening NP News NP 74 CD . SENT 0 CD 74 CD . SENT 3 CD 75 CD . SENT 0 CD 75 CD . SENT 0 CD 69 CD . SENT 6 CD CNN NP NewsNight NP with IN Aaron NP Brown NP 56 CD . SENT 2 CD 56 CD . SENT 6 CD 57 CD . SENT 3 CD 57 CD . SENT 3 CD 51 CD . SENT 7 CD Drudge NN Report NP 60 CD . SENT 2 CD 61 CD . SENT 1 CD 61 CD . SENT 0 CD 60 CD . SENT 7 CD 56 CD . SENT 0 CD LA NP Times NP 69 CD . SENT 5 CD 70 CD . SENT 0 CD 70 CD . SENT 1 CD 69 CD . SENT 8 CD 65 CD . SENT 8 CD NBC NP Nightly JJ News NP 63 CD . SENT 3 CD 63 CD . SENT 4 CD 63 CD . SENT 6 CD 63 CD . SENT 5 CD 60 CD . SENT 9 CD NBC NP Today NN Show NP 65 CD . SENT 2 CD 66 CD . SENT 6 CD 66 CD . SENT 4 CD 66 CD . SENT 6 CD 55 CD . SENT 9 CD New NP York NP Times NP 74 CD . SENT 1 CD 75 CD . SENT 0 CD 75 CD . SENT 3 CD 74 CD . SENT 9 CD 70 CD . SENT 8 CD Newshour NP with IN Jim NP Lehrer NP 58 CD . SENT 1 CD 58 CD . SENT 5 CD 59 CD . SENT 0 CD 59 CD . SENT 3 CD 50 CD . SENT 9 CD Newsweek NP 66 CD . SENT 9 CD 67 CD . SENT 6 CD 68 CD . SENT 4 CD 68 CD . SENT 0 CD 68 CD . SENT 9 CD NPR NP Morning NP Edition NN 67 CD . SENT 2 CD 67 CD . SENT 9 CD 68 CD . SENT 3 CD 68 CD . SENT 2 CD 59 CD . SENT 2 CD Fox NP News NP Special JJ Report NP 41 CD . SENT 7 CD 42 CD . SENT 2 CD 42 CD . SENT 5 CD 42 CD . SENT 3 CD 33 CD . SENT 4 CD Time NP Magazine NP 65 CD . SENT 6 CD 65 CD . SENT 7 CD 65 CD . SENT 9 CD 65 CD . SENT 7 CD 64 CD . SENT 7 CD U NP . SENT S NP . SENT News NP and CC World NP Report NP 66 CD . SENT 1 CD 67 CD . SENT 2 CD 68 CD . SENT 4 CD 68 CD . SENT 5 CD 65 CD . SENT 7 CD USA NP Today NP 63 CD . SENT 3 CD 64 CD . SENT 5 CD 64 CD . SENT 9 CD 65 CD . SENT 0 CD 66 CD . SENT 9 CD Wall NP Street NP Journal NP 84 CD . SENT 9 CD 86 CD . SENT 6 CD 86 CD . SENT 9 CD 86 CD . SENT 8 CD 82 CD . SENT 1 CD Washington NP Post NP 66 CD . SENT 2 CD 66 CD . SENT 8 CD 66 CD . SENT 9 CD 66 CD . SENT 8 CD 56 CD . SENT 7 CD Washington NP Times NP 35 CD . SENT 4 CD 35 CD . SENT 8 CD 36 CD . SENT 1 CD 35 CD . SENT 1 CD 48 CD . SENT 0 CD Average NN of IN 20 CD Outlets NNS 63 CD . SENT 2 CD 64 CD . SENT 1 CD 64 CD . SENT 4 CD 64 CD . SENT 3 CD 59 CD . SENT 4 CD 1 CD Our PP$ sample NN includes VVZ policy NN groups NNS that WDT are VBP not RB usually RB called VVN think VVP tanks NNS , , such JJ as IN the DT NAACP NP , , NRA NP , , and CC Sierra NP Club NP . SENT To TO avoid VV using VVG the DT more RBR unwieldy JJ phrase NN think VVP tanks NNS and CC other JJ policy NN groups NNS we PP often RB use VVP a DT shorthand NN version NN , , think VVP tanks NNS . SENT When WRB we PP use VVP the DT latter JJ phrase NN we PP mean VVP to TO include VV the DT other JJ groups NNS , , such JJ as IN NAACP NP , , etc FW . SENT 2 LS Eighty CD nine CD percent NN of IN the DT Washington NP correspondents NNS voted VVD for IN Bill NP Clinton NP , , and CC two CD percent NN voted VVD for IN Ross NP Perot NP . SENT 3 CD Finding VVG Biases NNS on IN the DT Bus NN , , John NP Tierney NP , , New NP York NP Times NP , , August NP , , 1 CD , , 2004 CD . SENT The DT article NN noted VVD that IN journalists NNS outside IN Washington NP were VBD not RB as RB liberal JJ . SENT Twenty CD five CD percent NN of IN these DT journalists NNS favored VVD Bush NP over IN Kerry NP . SENT 4 CD Ruling NN Class NP War NP , , New NP York NP Times NP , , September NP 11 CD , , 2004 CD . SENT 5 CD Cambridge NP and CC Berkeley NP s PP preferences NNS for IN Republican JJ presidential JJ candidates NNS have VHP remained VVN fairly RB constant JJ . SENT In IN the DT House NP district NN that WDT contains VVZ Cambridge NP , , Bob NP Dole NP received VVD 17 CD percent NN of IN the DT two CD party NN vote NN in IN 1996 CD , , and CC George NP W NP . SENT Bush NP received VVD 19 CD percent NN in IN 2000 CD . SENT In IN the DT House NP district NN that WDT contains VVZ Berkeley NP , , Bob NP Dole NP received VVD 14 CD percent NN of IN the DT two CD party NN vote NN , , and CC George NP W NP . SENT Bush NP received VVD 13 CD percent NN . SENT 6 CD Some RB scholars NNS claim VVP that IN news NN outlets NNS cater VVP not RB to TO the DT desires NNS of IN consumers NNS , , but CC to TO the DT desires NNS of IN advertisers NNS . SENT Consequently RB , , since IN advertisers NNS have VHP preferences NNS that WDT are VBP more RBR pro JJ business NN or CC pro JJ free JJ market NN than IN the DT average JJ consumer NN , , these DT scholars NNS predict VVP that IN news NN outlets NNS will MD slant VV their PP$ coverage NN to TO the DT right NN of IN consumers NNS preferences NNS . SENT E SYM . SENT g NN . SENT , , see VV Parenti NP , , 1986 CD , , or CC Herman NP and CC Chomsky NP , , 1988 CD . SENT While IN our PP$ work NN finds VVZ empirical JJ problems NNS with IN such JJ predictions NNS , , Sutter NP 2002 CD notes NNS several JJ theoretical JJ problems NNS . SENT Most RBS important JJ , , although IN an DT advertiser NN has VHZ great JJ incentive NN to TO pressure VV a DT news NN outlet NN to TO give VV favorable JJ treatment NN to TO his PP$ own JJ product NN or CC his PP$ own JJ business NN , , he PP has VHZ little JJ incentive NN to TO pressure VV for IN favorable JJ treatment NN of IN business NN in IN general NN . SENT Although IN the DT total JJ benefits NNS of IN the DT latter JJ type NN of IN pressure NN may MD be VB large JJ , , they PP are VBP dispersed VVN across IN a DT large JJ number NN of IN businesses NNS , , and CC the DT advertiser NN himself PP would MD receive VV only RB a DT tiny JJ fraction NN of IN the DT benefits NNS . SENT 7 CD One CD of IN the DT most RBS novel JJ features NNS of IN the DT Lott NP Hasset NP paper NN is VBZ that IN to TO define VV unbiased JJ , , it PP constructs VVZ a DT baseline NN that WDT can MD vary VV with IN exogenous JJ factors NNS . SENT In IN contrast NN , , some DT studies NNS define VV unbiased JJ simply RB as IN some DT sort NN of IN version NN of IN presenting VVG both DT sides NNS of IN the DT story NN . SENT To TO see VV why WRB the DT latter JJ notion NN is VBZ inappropriate JJ , , suppose VVP that IN a DT newspaper NN devoted VVN just RB as IN many JJ stories NNS describing VVG the DT economy NN under IN President NP Clinton NP as RB good JJ as IN it PP did VVD describing VVG the DT economy NN as IN bad JJ . SENT By IN the DT latter JJ notion NN this DT newspaper NN is VBZ unbiased JJ . SENT However RB , , by IN Lott NP and CC Hasset NP s PP notion NN the DT newspaper NN is VBZ unbiased JJ only RB if IN the DT economy NN under IN Clinton NP was VBD average JJ . SENT If IN instead RB it PP was VBD better RBR than IN average NN , , Lott NP and CC Hasset NP as IN many JJ would MD recognize VV as RB appropriate JJ , , including VVG us PP would MD judge VV the DT newspaper NN to TO have VH a DT conservative JJ bias NN . SENT Like IN Lott NP and CC Hasset NP , , our PP$ notion NN of IN bias NN also RB varies VVZ with IN exogenous JJ factors NNS . SENT For IN instance NN , , suppose VVP after IN a DT series NN of IN events NNS liberal JJ conservative NN think VVP tanks NNS gain VV more JJR respect NN and CC credibility NN say VVP , , because IN they PP were VBD better RBR at IN predicting VVG those DT events NNS , , which WDT causes VVZ moderates NNS in IN Congress NP to TO cite VV them PP more RBR frequently RB . SENT By IN our PP$ notion NN , , for IN a DT news NN outlet NN to TO remain VV unbiased JJ , , it PP also RB must MD cite VV the DT liberal JJ conservative NN think VVP tanks NNS more RBR frequently RB . SENT The DT only JJ other JJ paper NN of IN which WDT we PP are VBP aware JJ that IN also RB constructs VVZ a DT baseline NN that WDT controls VVZ for IN exogenous JJ events NNS is VBZ Tim NP Groeling NP and CC Samuel NP Kernell NP s VVZ 1998 CD study NN of IN presidential JJ approval NN . SENT These DT researchers NNS examine VVP the DT extent NN to TO which WDT media NNS outlets NNS report VVP increases NNS and CC decreases NNS in IN the DT president NN s PP approval NN , , while IN controlling VVG for IN the DT actual JJ increases NNS and CC decreases NNS in IN approval NN whether IN reported VVN by IN the DT media NNS or CC not RB . SENT The DT focus NN of IN the DT paper NN , , however RB , , is VBZ on IN whether IN news NN outlets NNS have VHP a DT bias NN toward IN reporting VVG good JJ or CC bad JJ news NN , , not RB on IN any DT liberal JJ or CC conservative JJ bias NN . SENT 8 CD New NP York NP Times NP Executive NP Editor NP Howell NP Raines NP accepting VVG the DT George NP Beveridge NP Editor NP of IN the DT Year NN Award NN at IN a DT National NP Press NP Foundation NP dinner NN , , shown VVN live JJ on IN C NP SPAN NP 2 CD February NP 20 CD , , 2003 CD . SENT 9 CD Paul NP Krugman NP , , Into IN the DT Wilderness NN , , New NP York NP Times NP , , November NP 8 CD , , 2002 CD . SENT 10 CD Al NP Franken NP 2003 CD , , 3 CD Lies NP and CC the DT Lying VVG Liars NNS Who WP Tell VV Them NP . SENT A DT Fair NP and CC Balanced JJ Look NN at IN the DT Right NP . SENT 11 CD Bill NP Moyers NP , , quoted VVN in IN Bill NP Moyers NP Retiring VVG from IN TV NP Journalism NP , , Frazier NP Moore NP , , Associated NP Press NP Online JJ , , December NP 9 CD , , 2004 CD . SENT 12 CD Groseclose NP , , Levitt NP , , and CC Snyder NP 1999 CD argue VVP that IN the DT underlying JJ scales NNS of IN interest NN group NN scores NNS , , such JJ as IN those DT compiled VVN by IN the DT Americans NPS for IN Democratic NP Action NP , , can MD shift VV and CC stretch VV across IN years NNS or CC across IN chambers NNS . SENT This DT happens VVZ because IN the DT roll NN call NN votes NNS that WDT are VBP used VVN to TO construct VV the DT scores NNS are VBP not RB constant JJ across IN time NN , , nor CC across IN chambers NNS . SENT They PP construct VV an DT index NN that WDT allows VVZ one CD to TO convert VV ADA NP scores NNS to TO a DT common JJ scale NN so IN that WDT they PP can MD be VB compared VVN across IN time NN and CC chambers NNS . SENT They PP call VVP such JJ scores NNS adjusted VVN ADA NP scores NNS . SENT 13 CD Importantly RB , , this DT conversion NN affects VVZ congressional JJ scores NNS the DT same JJ way NN that IN it PP affects VVZ media NNS scores NNS . SENT Since IN our PP$ method NN can MD only RB make VV relative JJ assessments NNS of IN the DT ideology NN of IN media NNS outlets NNS e NN . SENT g NN . SENT how WRB they PP compare VVP to TO members NNS of IN Congress NP or CC the DT average JJ American JJ voter NN , , this DT transformation NN is VBZ benign JJ . SENT Just RB as IN the DT average JJ temperature NN in IN Boston NP is VBZ colder JJR than IN the DT average JJ temperature NN in IN Philadelphia NP , , regardless RB if IN one PP uses VVZ a DT Celsius JJ scale NN or CC Fahrenheit NN scale NN , , all DT conclusions NNS we PP draw VVP in IN this DT paper NN are VBP unaffected JJ by IN the DT choice NN to TO use VV the DT 1999 CD House NP scale NN or CC the DT 1980 CD House NP scale NN . SENT 14 CD In IN the DT Appendix NN we PP report VVP the DT results NNS when WRB we PP do VVP include VV citations NNS that WDT include VVP an DT ideological JJ label NN . SENT When WRB we PP include VVP this DT data NNS , , this DT does VVZ not RB cause VV a DT substantial JJ leftward JJ or CC rightward JJ movement NN in IN media NNS scores NNS the DT average JJ media NNS score NN decreased VVD by IN approximately RB 0 CD . SENT 6 CD points NNS , , i NP . SENT e SYM . SENT it PP makes VVZ the DT media NNS appear VVP slightly RB more RBR conservative JJ . SENT Perhaps RB the DT greater JJR affect NN was VBD to TO cause VV a DT media NNS outlets NNS to TO appear VV more RBR centrist JJ . SENT For IN instance NN , , the DT New NP York NP Times NP and CC CBS NP Evening NP News NP tended VVD to TO give VV ideological JJ labels NNS to TO conservative JJ think NN tanks NNS more RBR often RB than IN they PP did VVD to TO liberal JJ think NN tanks NNS . SENT As IN a DT consequence NN , , when WRB we PP include VVP the DT labeled VVN observations NNS , , their PP$ scores NNS respectively RB decreased VVD i NP . SENT e SYM . SENT became VVD more RBR conservative JJ by IN 3 CD . SENT 8 CD and CC 1 CD . SENT 6 CD points NNS . SENT Meanwhile RB , , Fox NP News NP Special JJ Report NP tended VVD to TO do VV the DT opposite NN . SENT When WRB we PP included VVD labeled VVN observations NNS , , its PP$ score NN increased VVN i NP . SENT e SYM . SENT , , became VVD more RBR liberal JJ by IN 1 CD . SENT 8 CD points NNS . SENT We PP think VVP that IN such PDT an DT asymmetric JJ treatment NN of IN think NN tanks NNS ie NP to TO give VV labels NNS more RBR often RB to TO one CD side NN is VBZ itself PP a DT form NN of IN media NNS bias NN . SENT This DT is VBZ why WRB we PP base VVP our PP$ main JJ conclusions NNS on IN the DT non JJ labeled VVN data NNS . SENT 15 CD Groseclose NP , , Levitt NP , , and CC Snyder NP 1999 NP have VHP not RB computed VVN adjusted JJ scores NNS for IN years NNS after IN 1999 CD . SENT One CD consequence NN of IN this DT is VBZ that IN members NNS who WP first RB entered VVD Congress NP in IN 2001 CD do VVP not RB have VH adjusted VVN scores NNS . SENT ocrat NN . SENT Third JJ , , even RB if IN the DT new JJ members NNS were VBD not RB representative JJ , , this DT fact NN alone RB would MD not RB cause VV a DT bias NN in IN our PP$ method NN . SENT To TO see VV this DT , , suppose VVP that IN these DT omitted VVN members NNS were VBD disproportionately RB extreme JJ liberals NNS . SENT To TO estimate VV ADA NP scores NNS for IN a DT media NNS outlet NN , , we PP need VVP estimates NNS of IN the DT citation NN behavior NN of IN a DT range NN of IN members NNS with IN ideologies NNS near IN the DT ideology NN of IN the DT media NNS outlet NN . SENT If IN we PP had VHD omitted VVN some DT extreme JJ liberal JJ members NNS of IN Congress NP , , this DT does VVZ not RB bias NN our PP$ estimate NN of IN the DT citation NN pattern NN of IN the DT typical JJ liberal NN , , it PP only RB makes VVZ it PP less RBR precise JJ , , since IN we PP have VHP less JJR data NNS for IN these DT members NNS . SENT If IN , , on IN the DT other JJ hand NN , , new JJ members NNS behaved VVD differently RB from IN old JJ members NNS who WP have VHP the DT same JJ adjusted JJ ADA NP score NN , , then RB this DT could MD cause VV a DT bias NN . SENT For IN instance NN , , suppose VVP new JJ members NNS with IN a DT 70 CD real JJ ADA NP score NN tend VVP to TO cite VV conservative JJ think VVP tanks NNS more RBR often RB than IN do VV old JJ members NNS with IN a DT 70 CD score NN . SENT Then RB this DT would MD mean VV that IN Congress NP s PP citation NN patterns NNS are VBP really RB more RBR conservative JJ than IN we PP have VHP recorded VVN . SENT This DT means VVZ the DT media NNS s PP citation NN patterns NNS are VBP really RB more RBR liberal JJ relative NN to TO Congress NP than IN they PP appear VVP in IN our PP$ data NN set NN , , which WDT would MD mean VV that IN the DT media NNS is VBZ really RB more RBR liberal JJ than IN our PP$ estimates NNS indicate VVP . SENT However RB , , we PP have VHP no DT evidence NN to TO believe VV this DT or CC the DT opposite NN is VBZ the DT case NN . SENT And CC even RB if IN it PP were VBD , , because IN the DT new JJ members NNS are VBP such PDT a DT small JJ portion NN of IN the DT sample NN , , any DT bias NN should MD be VB small JJ . SENT 16 CD In IN fact NN , , for IN all DT members NNS of IN Congress NP who WP switched VVD parties NNS , , we PP treated VVD them PP as IN if IN they PP were VBD two CD members NNS , , one NN for IN when WRB they PP were VBD a DT Democrat NP and CC one NN for IN when WRB they PP were VBD a DT Republican JJ . SENT 17 CD The DT party NN averages NNS reflect VVP the DT midpoint NN of IN the DT House NP and CC Senate NP averages NNS . SENT Thus RB , , they PP give VVP equal JJ weight NN to TO each DT chamber NN , , not RB to TO each DT legislator NN , , since IN there EX are VBP more JJR House NP members NNS than IN senators NNS . SENT 18 CD Table NN 3 CD , , in IN the DT Estimation NN Results NNS section NN , , lists VVZ the DT period NN of IN observation NN for IN each DT media NNS outlet NN . SENT 19 CD We PP assert VVP that IN this DT statement NN is VBZ more RBR likely JJ to TO be VB made VVN by IN a DT conservative JJ because IN it PP suggests VVZ that DT government NN spending NN is VBZ filled VVN with IN wasteful JJ projects NNS . SENT This DT , , conservatives NNS often RB argue VVP , , is VBZ a DT reason NN that IN government NN should MD lower VV taxes NNS . SENT 20 CD We PP were VBD directed VVN to TO this DT passage NN by IN Sutter NP s VVZ 2001 CD article NN , , which WDT also RB seems VVZ to TO adopt VV the DT same JJ definition NN of IN bias NN that IN we PP adopt VVP . SENT 21 CD Like IN us PP , , Mullainathan NP and CC Shleifer NP 2003 CD define VV bias NN as IN an DT instance NN where WRB a DT journalist NN fails VVZ to TO report VV a DT relevant JJ fact NN , , rather RB than IN chooses VVZ to TO report VV a DT false JJ fact NN . SENT However RB , , unlike IN us PP , , Mullainathan NP and CC Shleifer NP define VV bias NN as IN a DT question NN of IN accuracy NN , , not RB a DT taste NN or CC preference NN . SENT More RBR specific JJ , , their PP$ model NN assumes VVZ that IN with IN any DT potential JJ news NN story NN , , there EX are VBP a DT finite JJ number NN of IN facts NNS that WDT apply VVP to TO the DT story NN . SENT By IN their PP$ definition NN , , a DT journalist NN is VBZ unbiased JJ only RB if IN he PP or CC she PP reports VVZ all PDT these DT facts NNS . SENT However RB , , given VVN that IN there EX may MD be VB an DT unwieldy JJ number NN of IN facts NNS that IN the DT journalist NN could MD mention VV , , it PP also RB seems VVZ consistent JJ with IN the DT spirit NN of IN their PP$ definition NN that IN if IN the DT journalist NN merely RB selects VVZ facts NNS randomly RB from IN this DT set NN or CC if IN he PP or CC she PP chooses VVZ a DT representative JJ sample NN , , then RB this DT would MD also RB qualify VV as RB unbiased JJ . SENT As IN an DT example NN , , suppose VVP that IN , , out RB of IN the DT entire JJ universe NN of IN facts NNS about IN free JJ trade NN , , most JJS of IN the DT facts NNS imply VVP that DT free JJ trade NN is VBZ good JJ . SENT However RB , , suppose VVP that IN liberals NNS and CC moderates NNS in IN Congress NP are VBP convinced VVN that IN it PP is VBZ bad JJ , , and CC hence RB in IN their PP$ speeches NNS they PP state VVP more JJR facts NNS about IN its PP$ problems NNS . SENT Under IN Mullainathan NP and CC Shleifer NP s PP definition NN , , to TO be VB unbiased JJ a DT journalist NN must MD state VV more JJR facts NNS about IN the DT advantages NNS of IN free JJ trade NN whereas IN , , under IN our PP$ definition NN a DT journalist NN must MD state VV more JJR facts NNS about IN the DT disadvantages NNS of IN free JJ trade NN . SENT Again RB , , we PP emphasize VVP that IN our PP$ differences NNS on IN this DT point NN are VBP ones NNS of IN semantics NNS . SENT Each DT notion NN of IN bias NN is VBZ meaningful JJ and CC relevant JJ . SENT And CC if IN a DT reader NN insists VVZ that DT bias NN should MD refer VV to TO one CD notion NN instead RB of IN the DT other JJ , , we PP suggest VVP that IN he PP or CC she PP substitute VV a DT different JJ word NN for IN the DT other JJ notion NN , , such JJ as IN slant NN . SENT Further RBR , , we PP suggest VVP that IN Mullainathan NP and CC Shleifer NP s PP notion NN is VBZ an DT ideal NN that IN a DT journalist NN perhaps RB should MD pursue VV before IN our PP$ notion NN . SENT Nevertheless RB , , we PP suggest VVP a DT weakness NN of IN Mullainathan NP and CC Shleifer NP s PP notion NN . SENT It PP is VBZ very RB inconvenient JJ for IN empirical JJ work NN , , and CC perhaps RB completely RB infeasible JJ . SENT Namely RB , , it PP would MD be VB nearly RB impossible JJ and CC at IN best JJS a DT very RB subjective JJ exercise NN for IN a DT researcher NN to TO try VV to TO determine VV all PDT the DT facts NNS that WDT are VBP relevant JJ for IN a DT given VVN news NN story NN . SENT Likewise RB , , it PP would MD be VB very RB difficult JJ , , and CC maybe RB impossible JJ , , for IN a DT journalist NN to TO determine VV this DT set NN of IN facts NNS . SENT To TO see VV this DT , , consider VV just RB a DT portion NN of IN the DT facts NNS that WDT may MD be VB relevant JJ to TO a DT news NN story NN , , the DT citations NNS from IN experts NNS . SENT There EX are VBP hundreds NNS , , and CC maybe RB thousands NNS , , of IN think VVP tanks NNS , , not RB to TO mention VV hundreds NNS of IN academic JJ departments NNS . SENT At IN what WP point NN does VVZ the DT journalist NN decide VV that IN a DT think VVP tank NN or CC academic JJ department NN is VBZ so RB obscure JJ that IN it PP does VVZ not RB need VV to TO be VB contacted VVN for IN a DT citation NN . SENT Further RBR , , most JJS think VVP tanks NNS and CC academic JJ departments NNS house NN dozens NNS of IN members NNS . SENT This DT means VVZ that IN an DT unbiased JJ journalist NN would MD have VH to TO speak VV to TO a DT huge JJ number NN of IN potential JJ experts NNS . SENT Moreover RB , , even RB if IN the DT journalist NN could MD contact VV all DT of IN these DT experts NNS , , a DT further JJR problem NN is VBZ how WRB long JJ to TO talk VV to TO them PP . SENT At IN what WP point NN does VVZ the DT journalist NN stop NN gathering VVG information NN from IN one CD particular JJ expert NN before IN he PP or CC she PP is VBZ considered VVN unbiased JJ . SENT Even RB if IN a DT journalist NN only RB needs VVZ to TO contact VV a DT representative JJ sample NN of IN these DT experts NNS , , a DT problem NN still RB exists VVZ over IN defining VVG the DT relevant JJ universe NN of IN experts NNS . SENT Again RB , , when WRB is VBZ an DT expert NN so RB obscure JJ that IN he PP or CC she PP should MD not RB be VB included VVN in IN the DT universe NN . SENT A DT similar JJ problem NN involves VVZ the DT journalist NN s PP choice NN of IN stories NNS to TO pursue VV . SENT A DT news NN outlet NN can MD choose VV from IN a DT huge JJ and CC possibly RB infinite JJ number NN of IN news NN stories NNS . SENT Although IN Mullainathan NP and CC Shleifer NP s PP model NN focuses VVZ only RB on IN the DT bias NN for IN a DT given VVN story NN , , a DT relevant JJ source NN of IN bias NN is VBZ the DT journalist NN s PP choice NN of IN stories NNS to TO cover VV . SENT It PP would MD be VB very RB difficult JJ for IN a DT researcher NN to TO construct VV a DT universe NN of IN stories NNS from IN which WDT journalists NNS choose VVP to TO cover VV . SENT For IN instance NN , , within IN this DT universe NN , , what WP proportion NN should MD involve VV the DT problems NNS of IN dual JJ career NN parents NNS . SENT What WP proportion NN should MD involve VV corporate JJ fraud NN . SENT 22 CD Originally RB we PP used VVD Stata NP to TO try VV to TO compute VV estimates NNS . SENT With IN this DT statistical JJ package NN we PP estimate VVP that IN it PP would MD have VH taken VVN eight CD weeks NNS for IN our PP$ computer NN to TO converge VV and CC produce VV estimates NNS . SENT 23 CD However RB , , Hamilton NP also RB notes VVZ that IN CBS NP covered VVD roll NN calls NNS by IN the DT American NP Conservative NP Union NP more RBR frequently RB than IN the DT other JJ two CD networks NNS . SENT Nevertheless RB , , one PP can MD compute VV differences NNS in IN frequencies NNS between IN roll NN calls NNS from IN the DT ADA NP and CC ACU NP . SENT These DT differences NNS show VVP CBS NP to TO be VB more RBR liberal JJ than IN ABC NP and CC NBC NP . SENT That DT is VBZ , , although IN all DT three CD networks NNS covered VVN ADA NP roll NN calls VVZ more RBR frequently RB than IN they PP covered VVD ACU NP roll NN calls NNS , , CBS NP did VVD this DT to TO a DT greater JJR extent NN than IN the DT other JJ two CD networks NNS did VVD . SENT 24 CD Other JJ anecdotes NNS that IN Sperry NP documents NNS are VBP . SENT i NP a DT reporter NN , , Kent NP MacDougall NP , , who WP , , after IN leaving VVG the DT Journal NP , , bragged VVD that IN he PP used VVD the DT bourgeois JJ press NN to TO help VV popularize VV radical JJ ideas NNS with IN lengthy JJ sympathetic JJ profiles NNS of IN Marxist JJ economists NNS . SENT ii NP another DT Journal NP reporter NN who WP , , after IN calling VVG the DT Houston NP based VVN MMAR NP Group NP shady JJ and CC reckless JJ , , caused VVD the DT Journal NP to TO lose VV a DT libel NN suit NN after IN jurors NNS learned VVN that IN she PP misquoted VVD several JJ of IN her PP$ sources NNS . SENT iii NP a DT third JJ Journal NP reporter NN , , Susan NP Faludi NP the DT famous JJ feminist NN characterized VVN Safeway NP as IN practicing VVG robber NN baron NN style NN management NN practices NNS . SENT 25 CD See NP http NN . SENT people NNS press VVP . SENT org NP reports NNS display VVP . SENT php NN 3 CD . SENT ReportID NP 215 CD for IN a DT description NN of IN the DT survey NN and CC its PP$ data NNS . SENT See VV also RB Kurtz NP 2004 CD for IN a DT summary NN of IN the DT study NN . SENT 26 CD This NP comes VVZ from IN the DT estimates NNS for IN the DT Republican JJ coefficient NN that IN they PP list VVP in IN their PP$ Table NN 7 CD . SENT These DT estimates NNS indicate VVP the DT extent NN to TO which WDT a DT newspaper NN is VBZ more RBR likely JJ to TO use VV a DT negative JJ headline NN for IN economic JJ news NN when WRB the DT president NN is VBZ Republican JJ . SENT . SENT 27 CD Sometimes RB even JJ liberals NNS consider VVP NPR NP left JJ wing NN . SENT As IN Bob NP Woodward NP notes VVZ in IN The DT Agenda NP 1994 CD , , p NN . SENT 114 CD . SENT Paul NP Begala NP was VBD steaming VVG . SENT To TO him PP , , OMB NP Director NP , , Alice NP Rivlin NP symbolized VVN all DT that WDT was VBD wrong JJ with IN Clinton NP s PP new JJ team NN of IN Washington NP hands NNS , , and CC represented VVD the DT Volvo NP driving NN , , National NP Public NP Radio NP listening NN , , wine NN drinking NN liberalism NN that IN he PP felt VVD had VHD crippled VVN the DT Democratic JJ Party NN for IN decades NNS . SENT 28 CD To TO test VV that IN NPR NP is VBZ to TO the DT right NN of IN Joe NP Lieberman NP we PP assume VVP that IN we PP have VHP measured VVN the DT ideological JJ position NN of IN Lieberman NP without IN error NN . SENT Using VVG the DT values NNS in IN Table NN 2 CD and CC 3 CD , , the DT t NN test NN for IN this DT hypothesis NN is VBZ t NN 74 CD . SENT 2 CD 66 CD . SENT 3 CD 1 CD . SENT 0 CD 7 CD . SENT 9 CD . SENT This DT is VBZ significant JJ at IN greater JJR than IN 99 CD . SENT 9 CD levels NNS of IN confidence NN . SENT To TO test VV that IN NPR NP is VBZ to TO the DT right NN of IN the DT New NP York NP Times NP , , we PP use VVP a DT likelihood NN ratio NN test NN . SENT The DT value NN of IN the DT log NN likelihood NN function NN when WRB NPR NP and CC the DT NY NP Times NP are VBP constrained VVN to TO have VH the DT same JJ score NN is VBZ 78 CD , , 616 CD . SENT 64 CD . SENT The DT unconstrained JJ value NN of IN the DT log NN likelihood NN function NN is VBZ 78 CD , , 609 CD . SENT 35 CD . SENT The DT relevant JJ value NN of IN the DT likelihood NN ratio NN test NN is VBZ 2 CD 78 CD , , 616 CD . SENT 64 CD 78 CD , , 609 CD . SENT 35 CD . SENT This DT is VBZ distributed VVN according VVG to TO the DT Chi NP Square NP distribution NN with IN one CD degree NN of IN freedom NN . SENT At IN confidence NN levels NNS greater JJR than IN 99 CD . SENT 9 CD we PP can MD reject VV the DT hypothesis NN that IN the DT two CD outlets NNS have VHP the DT same JJ score NN . SENT 29 CD Of IN the DT reports NNS written VVN by IN Matt NP Drudge NN , , he PP cited VVD the DT Brookings NP Institution NP twice RB actually RB once RB , , but CC he PP listed VVD the DT article NN for IN two CD days NNS in IN row NN the DT ACLU NP once RB , , Taxpayers NNS for IN Common JJ Sense NN once RB , , and CC Amnesty NN International NP once RB . SENT On IN June NP 22 CD , , 2004 CD , , the DT Drudge NN Report NP listed VVD a DT link NN to TO an DT earlier JJR version NN of IN our PP$ paper NN . SENT Although IN that DT version NN mentioned VVD many RB think VV tanks NNS , , only RB one CD case NN would MD count VV as IN a DT citation NN . SENT This DT is VBZ the DT paraphrased VVN quote NN from IN RAND NN members NNS , , stating VVG that IN the DT media NN tends VVZ to TO cite VV its PP$ military JJ studies NNS less RBR than IN its PP$ domestic JJ studies NNS . SENT The DT above JJ quote NN from IN PERC NP was VBD not RB in IN the DT earlier JJR version NN , , although IN it PP would MD also RB count VV as IN a DT citation NN . SENT At IN any DT rate NN , , we PP instructed VVD our PP$ research NN assistants NNS not RB to TO search VV our PP$ own JJ paper NN for IN citations NNS . SENT 30 CD Nevertheless RB , , we PP still RB report VV how WRB our PP$ results NNS change VVP if IN instead RB we PP use VVP median JJ statistics NNS . SENT See VV footnotes NNS 34 CD and CC 35 CD . SENT 31 CD The DT year NN 1999 CD was VBD somewhat RB , , but CC not RB very RB , , atypical JJ . SENT During IN the DT rest NN of IN the DT 1990 CD s PP on IN average JJ 17 CD . SENT 6 CD senators NNS received VVD scores NNS between IN 33 CD and CC 67 CD , , approximately RB half DT as RB many JJ as RB would MD be VB expected VVN if IN scores NNS were VBD distributed VVN uniformly RB . SENT See VV http NN . SENT www JJ . SENT adaction NN . SENT org NP votingrecords NNS . SENT htm NN for IN ADA NP scores NNS of IN senators NNS and CC House NN members NNS . SENT 32 RB Another DT possible JJ bias NN involves VVZ the DT fact NN that IN D NP . SENT C NP . SENT has VHZ slightly RB fewer JJR people NNS than IN the DT average JJ House NP district NN . SENT Using VVG 2000 CD population NN estimates NNS source NN . SENT Almanac NN of IN American NP Politics NP , , 2002 CD edition NN , , D NP . SENT C SYM . SENT had VHD 572 CD , , 000 CD residents NNS while IN the DT average JJ House NP district NN in IN the DT country NN had VHD 646 CD , , 000 CD . SENT We PP treat VVP D NP . SENT C SYM . SENT as IN one CD district NN , , whereas IN a DT more RBR appropriate JJ analysis NN would MD treat VV D NP . SENT C SYM . SENT as IN 572 CD 646 CD of IN a DT district NN . SENT Again RB , , this DT will MD bias VV our PP$ results NNS in IN the DT opposite JJ direction NN of IN our PP$ main JJ conclusions NNS . SENT Specifically RB , , this DT will MD cause VV media NNS outlets NNS to TO appear VV more RBR conservative JJ than IN they PP really RB are VBP . SENT 33 CD To TO see VV this DT , , imagine VV a DT state NN with IN three CD districts NNS , , each DT with IN the DT same JJ distribution NN of IN voters NNS . SENT Thus RB , , the DT median JJ voter NN in IN each DT district NN has VHZ an DT ideology NN identical JJ to TO the DT median JJ voter NN of IN the DT state NN . SENT Now RB suppose VVP that IN a DT Republican JJ state NN legislature NN redraws VVZ districts NNS so RB that IN Democratic JJ voters NNS are VBP transferred VVN from IN districts NNS 1 CD and CC 2 CD to TO district NN 3 CD . SENT Suppose VV Republican JJ voters NNS are VBP transferred VVN in IN the DT opposite JJ direction NN . SENT Necessarily RB , , the DT increase NN in IN Democratic JJ voters NNS in IN district NN 3 CD is VBZ twice RB the DT average JJ increase NN in IN Republican JJ voters NNS in IN districts NNS 1 CD and CC 2 CD . SENT Next RB , , suppose VVP that IN the DT expected VVN ideological JJ score NN of IN a DT representative NN is VBZ a DT linear JJ function NN of IN the DT fraction NN of IN Democratic JJ voters NNS in IN his PP$ or CC her PP$ district NN . SENT Then RB it PP will MD necessarily RB be VB the DT case NN that IN the DT expected VVN average JJ ideological JJ score NN of IN the DT representatives NNS in IN this DT hypothetical JJ state NN will MD be VB identical JJ to TO the DT expected VVN average NN before IN redistricting VVG . SENT However RB , , the DT same JJ will MD not RB be VB true JJ of IN the DT median JJ score NN . SENT It PP will MD be VB expected VVN to TO decrease VV i NP . SENT e SYM . SENT to TO become VV more RBR conservative JJ . SENT 34 CD A NP clever JJ alternative NN measure NN , , suggested VVD to TO us PP by IN David NP Mayhew NP , , is VBZ to TO use VV a DT regression NN based VVN framework NN to TO estimate VV the DT expected VVN ADA NP score NN of IN a DT legislator NN whose WP$ district NN is VBZ perfectly RB representative JJ of IN the DT entire JJ U NP . SENT S NP . SENT In IN the DT 2000 CD presidential JJ election NN Gore NP won VVD 50 CD . SENT 27 CD of IN the DT two CD party NN vote NN including VVG D NP . SENT C SYM . SENT . SENT Suppose VVP we PP could MD construct VV a DT hypothetical JJ congressional JJ district NN with IN an DT identical JJ Gore NP vote NN percentage NN . SENT It PP is VBZ reasonable JJ to TO believe VV that IN the DT expected VVN adjusted JJ ADA NP score NN of IN the DT legislator NN from IN such PDT a DT district NN is VBZ a DT good JJ measure NN of IN the DT ideology NN of IN the DT centrist JJ U NP . SENT S NP . SENT voter NN , , and CC this DT appropriately RB adjusts VVZ for IN any DT biases NNS due JJ to TO gerrymandered VVN districts NNS , , exclusion NN of IN D NP . SENT C SYM . SENT voters NNS , , and CC the DT small JJ state NN biases NNS in IN the DT Senate NP . SENT To TO estimate VV this DT , , we PP regressed VVD i NP the DT 1999 CD adjusted JJ ADA NP scores NNS of IN members NNS of IN Congress NP on IN ii NP Gore NP s PP percentage NN of IN the DT two CD party NN vote NN in IN the DT legislator NN s PP district NN . SENT In IN this DT regression NN we PP included VVD observations NNS from IN the DT Senate NP as RB well RB as IN the DT House NP . SENT Remember VV that IN adjusted JJ scores NNS are VBP constructed VVN so RB that IN they PP are VBP comparable JJ across IN chambers NNS . SENT The DT results NNS of IN the DT regression NN were VBD . SENT ADA NP Score NN 46 CD . SENT 48 CD 1 CD . SENT 91 CD Gore NP Vote NP . SENT This DT implies VVZ 49 CD . SENT 53 CD as IN the DT expected VVN ADA NP score NN of IN a DT district NN in IN which WDT the DT Gore NP vote NN was VBD 50 CD . SENT 27 CD . SENT We PP repeated VVD this DT analysis NN using VVG , , instead RB , , adjusted JJ ADA NP scores NNS from IN 1998 CD , , 1997 CD , , 1996 CD and CC 1995 CD . SENT In IN the DT latter JJ three CD years NNS we PP used VVD the DT Clinton NP share NN of IN the DT 2 CD party NN vote NN , , and CC we PP used VVD Clinton NP s PP national JJ share NN , , 54 CD . SENT 74 CD as IN the DT share NN of IN the DT representative JJ district NN . SENT These DT years NNS give VVP the DT following VVG respective JJ estimates NNS of IN the DT ADA NP score NN of IN the DT centrist JJ U NP . SENT S NP . SENT voter NN . SENT 48 CD . SENT 83 CD , , 48 CD . SENT 99 CD , , 47 CD . SENT 24 CD , , and CC 47 CD . SENT 41 CD . SENT The DT average NN of IN these DT five CD measures NNS is VBZ 48 CD . SENT 40 CD . SENT Since IN this DT number NN is VBZ 1 CD . SENT 7 CD points NNS les FW than IN the DT mean JJ based VVN measure NN of IN the DT centrist JJ voter NN 50 CD . SENT 1 CD , , if IN one PP believes VVZ that IN it PP is VBZ the DT more RBR appropriate JJ measure NN , , then RB our PP$ main JJ conclusions NNS based VVN on IN the DT mean JJ based VVN measure NN are VBP biased VVN rightward JJ that IN is VBZ , , the DT more RBR appropriate JJ conclusion NN would MD assert VV that IN the DT media NNS are VBP an DT additional JJ 1 CD . SENT 7 CD points NNS to TO the DT left NN of IN the DT centrist JJ voter NN . SENT Yet RB another DT measure NN is VBZ based VVN on IN median JJ scores NNS of IN the DT House NP and CC Senate NP . SENT The DT average JJ Senate NP median NN over IN the DT five CD years NNS was VBD 58 CD . SENT 19 CD , , while IN the DT average JJ House NP median NN was VBD 40 CD . SENT 61 CD . SENT Again RB , , both CC these DT figures NNS include VVP phantom JJ D NP . SENT C SYM . SENT legislators NNS , , and CC the DT Senate NP score NN is VBZ weighted VVN by IN state NN population NN . SENT The DT midpoint NN is VBZ 49 JJ . SENT 4 LS , , which WDT is VBZ 0 CD . SENT 7 CD points NNS more RBR conservative JJ than IN our PP$ mean JJ based VVN measure NN . SENT If IN one PP believes VVZ that IN this DT is VBZ the DT more RBR appropriate JJ measure NN of IN centrist JJ , , then RB , , once RB again RB , , this DT implies VVZ that IN our PP$ media NNS estimates NNS are VBP biased VVN in IN the DT direction NN of IN making VVG them PP more RBR conservative JJ than IN they PP really RB are VBP . SENT 35 CD If IN instead RB we PP use VVP medians NNS , , the DT figure NN is VBZ 54 CD . SENT 9 CD 36 CD Another DT concern NN is VBZ that IN , , although IN Republicans NPS and CC Democrats NPS are VBP given VVN debate NN time NN nearly RB proportional JJ to TO their PP$ number NN of IN seats NNS , , one CD group NN might MD cite VV think VV tanks NNS more RBR frequently RB than IN the DT other JJ . SENT The DT above JJ reasoning NN also RB explains VVZ why WRB this DT will MD not RB cause VV a DT bias NN to TO our PP$ method NN . SENT 37 CD Despite IN its PP$ name NN , , the DT Federation NP of IN American JJ Scientists NNS is VBZ more JJR of IN a DT lobbying VVG group NN than IN a DT scholarly JJ think NN tank NN . SENT Indeed RB , , like IN most RBS other JJ well RB known VVN lobbying NN groups NNS , , its PP$ address NN is VBZ on IN K NP Street NP in IN Washington NP , , D NP . SENT C SYM . SENT 38 CD Sutter NP 2001 CD similarly RB notes VVZ that IN demand NN factors NNS may MD be VB the DT source NN of IN liberal JJ bias NN in IN the DT newspaper NN industry NN . SENT Specifically RB , , he PP notes VVZ that DT liberals NNS may MD have VH a DT higher JJR demand NN for IN newspapers NNS than IN conservatives NNS , , and CC he PP cites VVZ some DT suggestive JJ evidence NN by IN Goff NP and CC Tollison NP 1990 CD , , which WDT shows VVZ that IN as IN the DT voters NNS in IN a DT state NN become VVN more RBR liberal JJ , , newspaper NN circulation NN in IN the DT state NN increases NNS . SENT 39 CD Perhaps RB an DT even RB more RBR interesting JJ , , in IN Baron NP s PP model JJ news NN consumers NNS , , in IN equilibrium NN , , can MD be VB persuaded VVN in IN the DT direction NN of IN the DT bias NN of IN the DT news NN outlet NN , , despite IN the DT fact NN that IN they PP understand VVP the DT equilibrium NN of IN the DT game NN and CC the DT potential JJ incentives NNS of IN journalists NNS to TO slant VV the DT news NN . SENT 40 CD A NP simple JJ model NN supports VVZ this DT assertion NN . SENT Suppose VVP that IN in IN every DT congressional JJ district NN , , voters NNS have VHP ideal JJ positions NNS that WDT are VBP uniformly RB distributed VVN between IN 1 CD and CC 1 CD , , where WRB 1 CD represents VVZ the DT most RBS liberal JJ voter NN and CC 1 CD represents VVZ the DT most RBS conservative JJ voter NN . SENT Assume VV that IN a DT voter NN is VBZ a DT Democrat NP if IN and CC only RB if IN his PP$ or CC her PP$ ideal JJ position NN is VBZ less JJR than IN 0 CD . SENT Four CD candidates NNS , , two CD Republican NP and CC two CD Democrat NP , , simultaneously RB choose VV positions NNS in IN this DT space NN . SENT Next IN they PP compete VVP in IN two CD primary JJ elections NNS , , where WRB the DT Republican JJ voters NNS choose VVP between IN the DT two CD Republican JJ candidates NNS , , and CC likewise RB for IN the DT Democratic JJ primary NN . SENT Each DT voter NN votes NNS for IN the DT candidate NN that WDT is VBZ nearest IN his PP$ or CC her PP$ ideological JJ position NN , , and CC if IN two CD candidates NNS are VBP equidistant JJ , , then RB the DT voter NN flips VVZ a DT coin NN . SENT This DT assumption NN implies VVZ that DT voters NNS are VBP myopic JJ in IN the DT primary JJ election NN . SENT If IN , , instead RB , , the DT voters NNS were VBD fully RB rational JJ , , then RB it PP can MD easily RB be VB shown VVN that IN the DT candidates NNS will MD choose VV even RB more RBR centrist JJ positions NNS , , which WDT means VVZ that IN even RB more JJR voters NNS will MD consider VV themselves PP more RBR extreme JJ than IN the DT party NN averages NNS . SENT Assume VV that DT candidates NNS maximize VV the DT votes NNS that IN they PP receive VVP in IN the DT general JJ election NN i NP . SENT e SYM . SENT the DT votes NNS they PP receive VVP in IN the DT primary JJ election NN are VBP only RB a DT means NN to TO winning VVG votes NNS in IN the DT general JJ election NN . SENT Then RB this DT setup NN implies VVZ that IN in IN equilibrium NN both CC Democratic JJ candidates NNS will MD locate VV at IN . SENT 5 LS , , and CC both DT Republican JJ candidates NNS will MD locate VV at IN . SENT 5 LS . SENT Each DT winner NN of IN the DT primary NN has VHZ a DT 50 CD chance NN at IN winning VVG the DT general JJ election NN . SENT Once RB this DT is VBZ repeated VVN across IN many JJ districts NNS , , then RB the DT expected VVN number NN of IN voters NNS who WP consider VVP themselves PP more RBR extreme JJ than IN the DT party NN averages NNS will MD be VB 50 CD . SENT 41 CD For IN instance NN , , suppose VVP that DT consumers NNS are VBP distributed VVN uniformly RB between IN 1 CD and CC 1 CD . SENT Suppose VV that IN there EX are VBP 20 CD news NN outlets NNS , , and CC suppose VVP that DT consumers NNS choose VVP the DT outlet NN that WDT is VBZ closest JJS to TO them PP . SENT It PP is VBZ easy JJ to TO show VV that IN the DT unique JJ , , pure JJ strategy NN equilibrium NN is VBZ for IN two CD firms NNS to TO locate VV at IN . SENT 9 CD , , two CD firms NNS locate VVP at IN . SENT 7 CD , , , , and CC two CD firms NNS locate VVP at IN . SENT 9 CD . SENT 42 CD This DT difference NN , , however RB , , is VBZ not RB statistically RB significant JJ at IN the DT 95 CD confidence NN level NN . SENT A DT likelihood NN ratio NN test NN , , constraining VVG Time NP and CC NewsNight NP to TO have VH the DT same JJ score NN gives VVZ a DT log NN likelihood NN function NN that WDT is VBZ 1 CD . SENT 1 CD units NNS greater JJR than IN the DT unconstrained JJ function NN . SENT This DT value NN , , multiplied VVN by IN two CD , , follows VVZ a DT Chi NP Squared VVD distribution NN with IN one CD degree NN of IN freedom NN . SENT The DT result NN , , 2 LS . SENT 2 LS , , is VBZ almost RB significant JJ at IN the DT 90 CD confidence NN level NN , , but CC not RB quite RB . SENT The DT latter NN has VHZ a DT criterion NN of IN 2 CD . SENT 71 CD . SENT We PP obtained VVD similar JJ results NNS when WRB we PP tested VVD , , the DT joint JJ hypothesis NN that IN i NP Newsnight NP and CC Time NP have VHP identical JJ scores NNS and CC that DT ii NN all DT three CD network NN morning NN news NN shows NNS have VHP scores NNS identical JJ to TO their PP$ respective JJ evening NN news NN shows NNS . SENT A DT likelihood NN ratio NN test NN gives VVZ a DT value NN of IN 8 CD . SENT 04 CD , , which WDT follows VVZ a DT Chi NP Squared VVD distribution NN with IN four CD degrees NNS of IN freedom NN . SENT The DT value NN is VBZ significant JJ at IN the DT 90 CD confidence NN level NN criterion NN 7 CD . SENT 78 CD , , but CC not RB at IN the DT 95 CD confidence NN level NN criterion NN 9 CD . SENT 49 CD . SENT Our PP$ hunch NN is VBZ that IN with IN more JJR data NNS we PP could MD show VV conclusively RB that IN at IN least JJS sometimes RB different JJ news NN outlets NNS at IN the DT same JJ firm NN produce VV significantly RB different JJ slants NNS . SENT We PP suspect VVP that IN , , consistent JJ with IN Baron NP s VVZ 20 CD xx NN model NN , , editors NNS and CC producers NNS , , like IN reporters NNS , , are VBP given VVN considerable JJ slack NN , , and CC that IN they PP are VBP willing JJ to TO sacrifice VV salary NN in IN order NN to TO be VB given VVN such JJ slack NN . SENT 43 CD This DT result NN is VBZ broadly RB consistent JJ with IN Djakov NP , , McLiesh NP , , Nenova NP , , and CC Shleifer NP s VVZ 2003 CD notion NN of IN the DT public JJ choice NN theory NN of IN media NNS ownership NN . SENT This DT theory NN asserts VVZ that IN a DT government NN owned VVD media NNS will MD slant VV news NN in IN such PDT a DT way NN to TO aid VV incumbent JJ politicians NNS . SENT If IN so RB , , some DT reasonable JJ theories NNS e NN . SENT g NN . SENT Black NP , , 1958 CD suggest VVP that IN the DT slant NN should MD conform VV to TO the DT median JJ view NN of IN the DT incumbent JJ politicians NNS . SENT We PP indeed RB find VVP that IN the DT slant NN of IN the DT government NN funded VVN outlets NNS in IN the DT U NP . SENT S NP . SENT on IN average NN is VBZ fairly RB close JJ to TO the DT median JJ politicians NNS view VVP . SENT In IN fact NN , , it PP is VBZ closer JJR to TO the DT median JJ view NN than IN the DT average NN of IN the DT privately RB funded VVN outlets NNS that IN we PP examine VVP . SENT See VV Lott NP 1999 CD for IN an DT examination NN of IN a DT similar JJ public JJ choice NN theory NN applied VVN to TO the DT media NNS and CC the DT education NN system NN in IN a DT country NN . SENT 44 CD Geoffrey NP Nunberg NP 2004 CD , , in IN a DT critique NN of IN an DT earlier JJR version NN of IN our PP$ paper NN , , deserves VVZ credit NN for IN first RB noting VVG the DT problems NNS with IN the DT sentence NN level NN data NNS involving VVG Alexis NP de NP Tocqueville NP Institute NP . SENT Our PP$ earlier JJR version NN gave VVD approximately RB equal JJ focus NN to TO i NP estimates NNS using VVG citations NNS as IN the DT level NN of IN observation NN and CC ii NN estimates NNS using VVG sentences NNS as IN the DT level NN of IN observation NN . SENT Partly RB due JJ to TO his PP$ critique NN , , the DT current JJ version NN no RB longer RB focuses VVZ on IN sentences NNS as IN observations NNS . SENT We PP did VVD not RB have VH the DT same JJ agreement NN with IN the DT rest NN of IN his PP$ criticisms NNS , , however RB . SENT See VV Groseclose NP and CC Milyo NP 2004 CD for IN a DT response NN to TO his PP$ essay NN . SENT