/Users/andrea/_magisterarbeit/korpus/clean/trainkorpus/10/Lfile2.html NN ----------------------------------------- : 1 LS 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