Downloaded By: [Trinity College] At: 10:11 19 February 2008 171ips04.qxd 07/08/2002 08:50 Page 59 Page 08:50 07/08/2002 171ips04.qxd RS PLTCL TDE O.7 O1 20) PP.59–73 (2002) NO.1 VOL.17 LONDON S, CAS RRANK BY STUDIES PUBLISHED POLITICAL IRISH RNSP, the Politics at of Professor Visiting Paris.Rellow and Research Senior work on *Michael this Laver’s article was completed while he was a Government of Ireland odce fr xml b te nleta Mnfso eerh Group as Research texts, Manifesto influential political the of by coding example for hand conducted intensive the beyond labour well very analysis content traditional, science political computing take in improvements These huge power. recent by possible made been have New in developments text computational analysis within political science LOCATING TDs IN POLICY SPACES: POLICY IN LOCATINGTDs h tcnqe s sd o nls al 8 nls lnug speeches language English 58 all analyse to used is technique the article, this In analysis. empirical and development theory for both of benefits huge with actors, political individual of of positions policy the estimation benefits systematic the allows This The record. parliamentary the legislature. as such text, of sources a electronic daunting and complex time, first in the for made analyse, to ability the from come analysis text computerised speeches of the to manifestos, analysis party analyse to used previously texts, political of analysis computerised the for technique new a adapts article This neetn qetos bu te oiin o idvda Irish space. policy interesting substantively a in individual debate 1991 the in of computerised speakers all locate of and speeches legislative of use positions context the in analysis the vindicate the results The 1991. about in legislators questions interesting when substantively answer analysis to second, text and, speeches computerised legislative to applied of validity was, the purpose examine The to dimension. first, anti-government’ versus ‘pro- a on speakers individual the of one every locate to debate the in spoken words the use to was task the The coalition. of Ráil–PD Rianna future incumbent the on debate confidence 1991 October the in made WORDSCORING DÁIL SPEECHES DÁIL WORDSCORING ihe Lae* n Kneh Benoit Kenneth and aver* L Michael Department of PoliticalScience of Department TrinityDublin College Introduction BSTRACT AB Downloaded By: [Trinity College] At: 10:11 19 February 2008 171ips04.qxd 07/08/2002 08:50 Page 60 Page 08:50 07/08/2002 171ips04.qxd MG, o te oprtv Mnfso Poet CP (Budge, (CMP) Project Klingeman 1982, Manifestos Budge, and Laver 1987; Hearl, Comparative and Robertson the now (MRG), oiia pris Hr or ou i o txs eeae b individual by generated texts on is focus our Here parties. political issued manifestos election of scoring the beyond goes that texts political approach analysis. to text difficult very reliable and effective be cheap, fast, of form some to would access having without that ones including substantive new – tackle problems to therefore and texts, political of forms to other coding computerised apply to possible becomes it however, coding, becomes profession the As computer of them. methodology the in experienced and issued confident increasingly that parties the those and with texts of results the positions policy comparing the scoring of methods by traditional more using obtained techniques new the of validity new, relatively are texts political analyzing the assess to sense made has it for methods computerised Rinally,because standpoint. practical a from these simply gains coding enormous promises manifestos of coding of computerised for enterprise method the successful a resource-intensive, extremely is hand by because manifestos Second, scientists. political to value research great of therefore are and parties, political of positions policy the of statements substantive important considered party are manifestos project, CMP the motivate that reasons important the all several for for Rirst, reasons. manifestos, party of analysis the on focused has Bara, 2001; Mansergh, and Giannetti 2001). Vries, de 2001; Garry, 2001; Garry,and (Laver authors of number Pennings,and Kleinnijenhuis 2000; replicating a by published been recently analysis, have sources, data independent completely text computerised of otherwise. doing implementations in Successful involved costs huge the of because simply dataset existing an to resort to need no with analyst, specific a of needs research the to according documents of recoding and coding effective and fast of the of all recode to documents involved needed in a manner more suitable to been the application at hand. post-war have would in that effort positions phenomenal policy the party of because been part large in has on this recently Until Europe. western data time-series sought have support widespread found not has within the profession, the CMP data have been widely used by that many who competition party of theory’ Budge 1994; hs sn a e poaiitc wr-crn’ ehd o computerised for method ‘word-scoring’ probabilistic new a using this We debates. legislative the in do made speeches of form the in legislators In this paper we present an application of computerised analysis of analysis computerised of application an present we paper this In texts political of coding computer the on work published all Nearly prospect the offers now however, analysis, content coded Computer et al et ., 2001). While grounded in a very specific ‘saliency specific very a in grounded While 2001). ., RS PLTCL STUDIES POLITICAL IRISH 60 et al. et , Downloaded By: [Trinity College] At: 10:11 19 February 2008 171ips04.qxd 07/08/2002 08:50 Page 61 Page 08:50 07/08/2002 171ips04.qxd n sbtniey cnldn b daig esn fr uue ss of speeches. legislative uses in future for lessons computerised textanalysisininvestigatingthepolicypositionscontained drawing by concluding substantively, and then present and discuss the results of our analysis both methodologically We analyse. we texts the describe briefly we Next manifestos. party than rather speeches legislative to this applying from arising issues discuss and the both opposition. the and within government coalition splits potential for looks to specifically and legislator, the has individual the of level it the at Ireland in differences where intra-party and inter- context a explore in to us allows Substantively, it payoffs. coding huge generate to potential computer of use the Irish evaluate individual of positions the Methodologically,to us space. allows policy this common a in estimate legislators to is aim Our government. coalition Ráil–PD Rianna incumbent the of future the on 1991, October in held confidence, of debate acrimonious and long a during TDs Irish by made speeches the of analysis the to this applying (2002), Garry and text analysis that has been developed and found effective by Laver, Benoit prah cmue cdn te M dt ad plig hs o the Budge see itself, CMP the to by area this in this dictionary-based applying and alternative data Penningsin found be can parliament, Ror (2002). European essays recent CMP Vries, An the de coding (2001). by computer approach, declarations, Mansergh government and Irish Giannetti as and well German as for manifestos, implemented Italian and Dutch been for and (2001) Garry by manifestos subsequently party Norwegian has and (2000), Garry and Laver by described is Ireland and Britain in manifestos party of positions policy social and economic the estimating to approach of this application successful recent A analysis. under texts the of positions to subjected policy the of estimates derive to then order in technique scaling of form some are categories particular into fall to of observed frequencies words relative The positions. policy particular be with to associated exploration, empirical by guided judgement a as subjective analyst, of the matter by priori a deemed words key of lists are dictionaries the frequencies of words found in predefined coding These ‘dictionaries’. count essentially analysis text computerised of techniques ‘Traditional’ to developing an appropriate coding dictionary upon which to ground to which upon dictionary coding appropriate remains an it developing to technique computerised a is it labour-intensive in that very time considerable and effort must be applied that fact the despite Rirst, In what follows we first outline the word-scoring technique we use we technique word-scoring the outline first we follows what In While While this works technique well it has two paradoxical disadvantages. The Word-ScoringThe TextComputerised toApproach Analysis OAIG D I PLCY SPACES Y POLIC IN TDS LOCATING 61 et al. et (2001). Downloaded By: [Trinity College] At: 10:11 19 February 2008 171ips04.qxd 07/08/2002 08:50 Page 62 Page 08:50 07/08/2002 171ips04.qxd S eeec txs Ti soe s n fet cniinl siae f the of estimate word conditional dimension a on text effect any of in position is score This texts. reference dictionary based computer coding techniques, in which the development the which in techniques, coding computer based traditional dictionary to contrast stark no in with is seconds This whatsoever. of intervention matter human a in texts reference the from calculated be of reference texts on policy dimension dimension policy on texts reference of set the of positions known the and probabilities conditional these Using e qatt i te odsoig prah Ti i te conditional the is This approach. word-scoring probability the in quantity key of the calculation the allowing text, in each used words of all frequencies is analyst the which in texts ‘virgin’ interested. new of analysis computerised the calculate to used matrix of is ‘word scores’ that replaces traditional coding dictionaries in the but information, substantive new no provides texts reference the of analysis computer The role. same the fulfil would – analysis content hand-coded prior for even – or data confident survey is mass example analyst the which in estimates independent any but Laver, investigation. under surveys, expert from estimates independent dimensions these take Garry and Benoit policy the on texts reference technique The 1997. in requires that there manifestos be independent estimates of the policy positions of the party Irish of analysis an for texts in in 1997, party and manifestos Irish 1992 party manifestos as reference British of positions policy the of analysis an for texts reference as 1992 well-known with texts in manifestos party British use (2002) Garry and Laver,Benoit example, ‘reference’ of set a Ror interested. is analyst the which in dimensions policy the on a positions is there of Rirst, analysis following. the preliminary involves essentially and Benoit but Laver, (2002) in Garry fully described is technique This researcher. the by calls judgement subjective no uses and dictionaries predefined use not first principles a probabilistic technique for coding political text that does from developed have (2002) Garry and Benoit Laver, coding, computer coding computer the comprise that the lists in dictionary. word freedom considerable the has of typically construction analyst the that sense the in subjective ultimately remains technique numerical highly this Second, inappropriate. dictionary coding given any render my context and time the in analysis, a in situation which changes in the lexicon across political wd Given Given the power of modern computers, the matrix of word scores can hs rlmnr aayi o rfrne et osre te relative the observes texts reference of analysis preliminary This of benefits full the realise to attempt an in problems these Addressing on dimension dimension on w . P wr that theanalystisreadingreferencetext d RS PLTCL STUDIES POLITICAL IRISH to every word word every to 62 d gvn ht h aayt s reading is analyst the that given , w d , it is possible to assign a ‘score’ a assign to possible is it , n h wr uies o the of universe word the in r , given word word given , w . Downloaded By: [Trinity College] At: 10:11 19 February 2008 171ips04.qxd 07/08/2002 08:50 Page 63 Page 08:50 07/08/2002 171ips04.qxd estimated positionofvirgintext The knowledge. prior no has analyst the which about text, ‘virgin’ any available. become investigation under dimensions policy the or appropriate, whenever improved estimates of the positions of these on deemed is texts reference of set new a whenever recalculated instantly be to scores word of matrix the allows technique word-scoring The analyst. the of part the on time of weeks requires that effort research a is scratch from dictionary coding computer new To a stage. test every and develop research task,involvingsubstantivejudgementstomadebytheanalystat human time-consuming and major a is dictionary coding computer a of et o te iesos ne ivsiain Te eeec texts, of is reference analyst human The more. much and – The dictionary coding a by done investigation. previously everything under do positions their of estimates of the dimensions with combined set the a on with texts dictionary these of coding positions policy the of estimates of matrix a and computer texts reference traditional the replaces that are scores then applied to motorcycle party word manifestos. In a nutshell, our derive new approach to or used be books not should Travel example, for texts. manuals, repair reference the as meaning, general same the with lexicon same the having of sense the in type’, same the ‘of that means This texts. reference independent expert advice is the needed to ensure that the reference texts are from derived scores word using drawn be can texts virgin the of positions the about inferences valid that so analysis, under texts virgin the for scores word of sources appropriate political conventional within possible been analysis. text science not has that determine to something analyst significant, the statistically are texts allows between differences This estimated whether texts. virgin and the reference of the in estimates words of patterns provides the on based score, time text virgin the each of first uncertainty that the is scores) for CMP technique the computerised (e.g. techniques coding traditional over method this of the advantage considerable of very and positions final A the texts. of reference estimates independent the as units same the in to produce estimates of the positions of virgin texts that are denominated estimates these rescale to necessary is it texts, reference the in appear not word of patterns overlapping do that words use may texts virgin that fact the and texts, Given between usage authors.) the from available are analysis and Benoit Garry,Laver, this 2002. Necessary in computer software and the raw method text files analysed the replicate of description relative to full the their consult wishing should by (Readers weighted occurrence. text, virgin of the frequency in used words scored the Using the derived matrix of word scores, it is now possible to analyse to possible now is it scores, word of matrix derived the Using In all of this it is very important to ensure that the reference texts are texts reference the that ensure to important very is it this of all In OAIG D I PLCY SPACES Y POLIC IN TDS LOCATING v on dimension 63 d is sumofthescores Downloaded By: [Trinity College] At: 10:11 19 February 2008 171ips04.qxd 07/08/2002 08:50 Page 64 Page 08:50 07/08/2002 171ips04.qxd rsn we sitn tx aayi fo oe otx t aohr but another, to context one from analysis text shifting party when of present ever- are analysis issues These the speeches. legislative of from analysis the to analysis manifestos text computerised of techniques inter-party and intra- feasible. hitherto been of have than politics legislative analyses detailed and sophisticated more far the policy positions of individual legislators, opening up the possibility of n sacal rcr a te oss f h Oraha website: the of Houses the at record . searchable a in on publication houses both of their the Oireachtas since the in foundation of the state is spoken now available word following recorded every example, Ror analysis websites. legislative computerised highly to become have amenable speeches these the record, to parliamentary written is the application obvious One of part as verbatim preserved Always speeches. coders. parliamentary of analysis human for daunting too simply are that tasks on take can analysis text computerised where areas in work to it put to is task next the demonstrated, been has manifestos party to applied when technique word-scoring the of efficacy the Once troublesome very the been analysts speak. not do they that language a in content even and policy, social of dimension liberal–conservative previous for had what on even analysed, they that texts virgin the as of positions the of elections prior from manifestos reference texts. They were able to replicate utterly estimates using independent manifestos, party Irish and German British, of analysis the to successfully very approach this applied have (2002) Garry and Laver,Benoit positions. policy reference known with texts to usage, word of patterns their given virgin probabilistically, match texts to is does technique the what Intuitively, texts. reference and virgintexts,independentestimatesofthepolicypositions The analyst. the by in reference both of frequencies word are the patterns required data only understood not can languages technique to the applied analysis, be therefore content other any unlike and, written are analysis under texts the which in language the of whatsoever knowledge techniques. computer-coding possible best or hand- traditional using when done the be to have that those than and analyst texts reference possible estimates of the positions of these, jobs far more to appropriate an expert best the out seeking towards redirected are efforts her or his but with, dispensed not course ao ise ms b rsle, oee, f e ih o migrate to wish we if however, resolved, be must issues Major Perhaps the most feature remarkable of our approach is that it uses no Migrating Word-ScoringMigrating Speeches to Manifestosfrom RS PLTCL STUDIES POLITICAL IRISH 1 64 This allows estimates to be made of made be to estimates allows This Downloaded By: [Trinity College] At: 10:11 19 February 2008 171ips04.qxd 07/08/2002 08:50 Page 65 Page 08:50 07/08/2002 171ips04.qxd Manifestosare • following: the include purposes our for distinctions Key form. explicit very a in them confront to us forces analysis text computerised order and procedure, and so on. However, aside from these knockaboutthesefrom However,aside on. so procedure,and and order of points behaviour, disorderly for ejected being occasionallymembers details,including gory interruptions, its all insults, in general said, mêlée, was interventions that everything from of the account chair, verbatim a is HousestheOireachtasof website (seeabove). Thetranscript thedebateof investigationunderdebatedownloadedthethe wasfrom of text full The pehs f h Tosah n Lae o te poiin wih we which Opposition, the respectively. of anti-government and pro- definitively be to priori Leader a assume and as the speeches of speeches certain select set-piece the uncontroversially notably – scores word derive to which from texts reference can we that advantage methodological the has This debate. this in government anti- or pro- as party themselves expressed legislators which to extent the the estimate to here out of We each set words. 167,000 over by just of record written speeches a generating leaders, including speeches, set-piece 59 with • Manifestos and speeches use use speeches and Manifestos • Speeches tend to be • aiets r pbihd n a in published are Manifestos • is well worth attempting since the potential returns are so great. so are returns potential the since attempting worth well is it Nonetheless manifestos. party of analysis the that problematic more be will speeches of legislative the analysis therefore, respect, In every almost ofdne n h Iih oenet cnutd vr h tre as of days three 1991. October 16–18 the over conducted government, Irish the in confidence of motion a on debate major a in legislators individual of positions the estimate to out sets below reported analysis the direction, desired the in Estimating the Positions of Irish Legislators on Legislators PositionsIrish the of Estimating justifythecomparison differentof speeches theinsame analysis. to are we ifspeeches, of contextpoliticalestablishing the in taken be allows one manifesto to be compared to another; much more care must eios I tu ses iey ht h aayi o mnfso and manifestos of analysis the text. reference of types that different require will speeches likely seems thus It lexicons. oiy sus sece tn t b rsrce t a iie nme of number limited subjects. a to restricted be to tend speeches issues; policy In order to minimise some of these problems and yet take a first step first a take yet and problems these of some minimise to order In analyse,ourstatistical confidence intheresults islikely tobereduced. OAIG D I PLCY SPACES Y POLIC IN TDS LOCATING encyclopaedic 2 much shorter than manifestos This has the advantage that it was a major debate major a was it that advantage the has This different language registersdifferentlanguage documents dealing with a wide range of rangewide a documentswithdealing Data 65 lal-eie pltcl context politicalclearly-defined A Priori A . With fewer words to PolicyScales and different and that Downloaded By: [Trinity College] At: 10:11 19 February 2008 171ips04.qxd 07/08/2002 08:50 Page 66 Page 08:50 07/08/2002 171ips04.qxd anti-governmentdimension using the method described above. versuspro-the speakers otheron 55 positions estimateofthepossible to all.Havingincalculated wordscoresreferencefromthe then texts,wasit debate in at least one of the reference texts – thea indifferenttotal usedwordsall forscores calculation ofword allowedthe of2,856 different words referenceouras textswithindependently knownreference positions. This Democratic Left).Thusthespeechesofthesethreepartyleadersweretaken Rossa, then leader deof Prionsias the of Workers’ that anti-government – Party be (most axiomatically to assumedof which was and was soon anti-government to become be to axiomaticallyassignedreferencea position assumedof–1.0. The speech ofone other party leader was Bruton, John Thespeech of the Rine Gael leader of the day and leader of the opposition, a reference position of +1.0 on the pro- versus anti- governmentgovernment, dimension. was assumed axiomatically the of Taoiseach, leaderto thebe as pro-government of speech self-evident. The be and assigned to priori a party leadersonthe‘pro-versusanti-government’dimensionwereassumed takenasbeing theessence ofthedebate. Thereference positions ofcertain ‘pro-versusanti-government’speakersapositionthe theon of dimension thedebate under investigation and convert these into text files for analysis. difficultto extract the set-piece 59 speeches made by different legislators in theirnamestheparliamentaryon recordhaving as spoken. wasthusItnot party,sametheothersfromput with chancetoallowing people the more speechwithinbenchmemberssharetheirallottedthis.Backoftenagreedtotime single long- a to made according and orders, period standing time and conventionsestablished enforced strictly a allotted was speak elements,thedebate wasvery tightly structured. Each legislator allowed to treated as virgin texts, the positions of which were to be estimated. be to were which of were positions the texts, debate virgin as treated the in spoke who TDs other all of speeches The position. pro-government wholehearted than less a of his indications whether showed speech assess to important substantively considered was it this of light the in and Ráil, Rianna with coalition leave to shortly were PDs the because was This estimated. be to position a with text virgin a as treated O’Malley, a government minister during the 1991 confidence debate, was Des leader PD of speech the Similarly a disposition. pro-government more of hints giving was leader Labour the time, that at whether, assess to with important considered government was it this of coalition light the in a and 1992, in Ráil into Rianna go to was Labour because was This interest. substantive of matter a as estimated be to was which of position parties, that of Labour leader Dick Spring was treated as a virgin text, the These texts were analysed using the word-scoring method to establishword-scoringtothemethodanalysed using weretexts These unn frt o h sece o te edr o te te mi Dáil main other the of leaders the of speeches the to first Turning RS PLTCL STUDIES POLITICAL IRISH 66 3 Downloaded By: [Trinity College] At: 10:11 19 February 2008 171ips04.qxd 07/08/2002 08:50 Page 67 Page 08:50 07/08/2002 171ips04.qxd o nweg o te at ht ta aayig ina ál ministerial Ráil Rianna analysing itwas scores.estimatingthesespeecheswhen that fact the of knowledge no the of statistical productpattern of word usage in their speeches, since the computer ahad entirely ministersis Ráil Rianna of clustering the Thus speaker.the of affiliation party or identity the of whatsoever knowledge referencescoresderivednothewordfromtexts,butthe speech and each in frequencies word the using technique, the by ‘automatically’ entirely is It validity. face strong important populated very to note in this context scale that the scores in entirely the Appendix 1 are derived gives almost This while backbenchers). Ráil is scale, Rianna stray scale two or one the (plus TDs theopposition Labour and of Gael Rine by of end end pro-government anti-government the the at together clustered speeches. legislative to applied as technique word-scoring the of vindication remarkable a are results The debate. the in speech their from estimated score the to according pro-government to anti- from ranked aor7–.59002 06 0.38 –0.62 0.0220 –0.3599 7 Standardised Standardised Labour SD Raw mean Raw N Group closely speeches, their in TD. anti-government Green sole the by followed systematically most the were TDs Workers’Party and Gael Rine Labour, scale, the of end other At speeches. in their the Riannapro-government less TDs on Ráil average with debate, the in speakers pro-government most the overwhelmingly were expect, might we as ministers, Ráil Rianna speaker. of category by dimension, anti-government versus pro- the on scores mean gives which res1–.48––.3– 0.53 0.73 – 1.24 – –0.59 –0.46 –0.43 0.66 0.41 –0.21 0.0306 0.0423 – 0.50 –0.3580 –0.3501 – 0.0721 21 1.15 –0.3488 2 – –0.2999 1 –0.3360 10 0.0383 –0.2947 1 RG WP –0.2571 1 Greens Independent 12 RR Minister PD Ministers RR observations observations to allow comparisons to be made more clearly. 55 all over standardised and form raw in both 1, Appendix in given are debate confidence 1991 the in speakers non-reference 55 all for Scores Very striking indeed is the pattern in which all Rianna Ráil ministers are The patterns in Appendix 1 are summarised in Table 1 and Rigure 1, Rigure Table and in 1 summarised are 1 Appendix in patterns The 91 ORDNE EAE N PO VRU AT- GOVERNMENT’ ANTI- VERSUS ‘PRO- ON DEBATE CONRIDENCE 1991 EN A AD TNADSD CRS R PAES IN SPEAKERS OR SCORES STANDARDISED AND RAW MEAN OAIG D I PLCY SPACES Y POLIC IN TDS LOCATING IESO, Y AEOY R TD OR CATEGORY BY DIMENSION, AL 1 TABLE Results 67 enSD mean 4 The The TDs are Downloaded By: [Trinity College] At: 10:11 19 February 2008 171ips04.qxd 07/08/2002 08:50 Page 68 Page 08:50 07/08/2002 171ips04.qxd scores. The width of each box is proportional to the number of speakers in each category. each in speakers of number the to proportional is box each of width The scores. Note: te hn, ik pigs peh crd ey oil i te anti- the in solidly very scored speech this in words his from whatsoever indication no with camp, government Spring’s Dick hand, other the On lukewarm. a distinctly as estimated as been have would out government picked the for support his been though minister, this Ráil Rianna have a in not was who used not speaker words would the O’Malley of basis debate, the confidence on Simply territory. same the in less was were Burke and Brennan Rlynn, ministers Ráil Rianna though colleagues, O’Malley Des of position The than staunchly pro-government that of most of his Rianna Ráil ministerial questions. these to give answers (a 1 Appendix in reported O’Malley scores The partner). coalition Des Ráil Rianna of positions Tanaiste future a and also but leader party opposition prominent a (then the Spring Dick leader Labour design, of and leader) PD and research minister government individual our of by positions open relative deliberately left questions the substantive two the to first We turn speakers. about conclusions substantive Table 1 gives us to some use encouragement the to scores generated draw O PO O SADRIE SOE O SEKR I 19 CONRIDENCE 1991 IN SPEAKERS OR SCORES STANDARDISED OR PLOT BOX The remarkable face validity of the scale reported in Appendix 1 and 1 Appendix in reported scale the of validity face remarkable The

oe idct te ein ad neqatl rne o ec gops standardised group’s each of ranges interquartile and medians the indicate Boxes Standardised Score EAE N PO VRU AT- OENET ‘DIMENSION, GOVERNMENT ANTI- VERSUS ‘PRO- ON DEBATE (Box width proportional to number of speakers) of number to proportional width (Box RS PLTCL STUDIES POLITICAL IRISH Y AEOY R TD OR CATEGORY BY IUE 1 RIGURE 68 Downloaded By: [Trinity College] At: 10:11 19 February 2008 171ips04.qxd 07/08/2002 08:50 Page 69 Page 08:50 07/08/2002 171ips04.qxd motn tig a be t miti a la sne f appropriate of sense clear a The maintain speeches. to been parliamentary has to thing applied important as technique word-scoring these manifestos, language-blind the of party vindication considerable of a as seen analysis be must the results from away step first a Taking Durkan.ConnaughtonOwen,and TDs Rine Gaelfrom came all of Yatesspeechesanti-governmentPeterviolentlyBarry, andmost the while less hostile to the government than those of their colleagues includedmuch Ivan were speeches whose TDs Gael Rine Other minister. government Garret Taoiseach former from came a of thatlikelookedspeech his inRitzGerald,usage word pattern of who speech maverick main the side, Gael Rine the On TDs. Gael Rine of those from dimension government gave anti- pro-versus Daverne the on indistinguishable and been have Roche would that speeches McDaid, as such others while government, pro- avidly most the among being Cullimore, and Nolan as such some, in debate this in government the on negotiations. coalition future of anticipation ‘soft’ being was he that debate when deriving the estimated positions reported in Appendix 1 and Tableand 1 Appendix in reportedpositions estimated the deriving when language English the of knowledge any of made was whatsoever use no sense itthe‘languagetechnique blind’,inrememberedisthe thatbe should Rirst, applied. be might word-scoring computerised suggests analysis material. raw available easily from datasets new exciting generate to easily and quickly potential considerable have does really technique the carefully, applied that, implies This parliament. in utter they words the upon solely based the for allowing first timethelocationofindividuallegislatorsinacommon policyspace, speeches, parliamentary in of analysis Party the to Labour manifestos the and of party analysis the from here methodological. migrated has government technique word-scoring The are drawn be in located. to conclusions PDs main the but convincingly the opposition, of be position could the debate the about questions intriguing some answer to English- us Substantively,allowed in this other 54 the speakers which on validity,language face good very with scale a of estimation the allowed it that in off, paid conservatism However,this positions. anti-government pro-and for texts reference confidence as leaders a party analysing in conservative was debate and in taking the set-piece speeches of government and opposition study current the sense, this In estimated. be to scales the on positions their and texts reference ina ál ak ec Ts ee mc mr vre bnh with bunch varied more much a were TDs bench back Ráil Rianna In this context it is worth taking note of the types of data to which this OAIG D I PLCY SPACES Y POLIC IN TDS LOCATING Conclusions 69 Downloaded By: [Trinity College] At: 10:11 19 February 2008 171ips04.qxd 07/08/2002 08:50 Page 70 Page 08:50 07/08/2002 171ips04.qxd 4. The fact that all raw scores are negative is entirely an artifact of the fact that two anti- two that fact the of artifact an entirely is negative are scores raw all that fact The 4. his because excluded be to had Dukes, Alan leader Gael Rine former speaker, One 3. or actual an with faced is government the when Ireland, in convention legislative The Irish the having 2. from arising issues constitutional interesting the side one on set We 1. the that feel we development.intellectualitsfurther investmentin considerable merits this similar, something or Given here, used have available. we approach are sources text appropriate a into sideways only range of different cultural not contexts, appropriate but backwards analysisas far back into time dataas systematic that of reach the provided extends Rome, approach this words, other or In available. were texts virgin Greece and reference ancient in politicians positionsofpolicyestimatethe to used beeasily as justcontrast,could in degreeanyreliabilityof validity.or techniqueThe demonstrated we here, retrospectestimateappliedpresentinwithpolicypositionscannot be but to used be can They prospect. in only work surveys) expert and studies the for conventionalmethodsestimatingforpositions such example(for election breakthrough considerable into another converted the ofpolitical Manyactors.positionspolicyof the systematic analysisof be is can This or form. in era, available politicalelectronic either any are in the these generated texts that extends toprovided applied vastly be can it technique the since technique Second, researcher.comparative serious the the of armoury methodological of feature empowering allow to tremendously a is language,itself in This calculated. be same to scores word appropriate the in were texts reference the that provided 1. The speeches analysed could have been delivered in any language at all, oenet pehs ee sd o aclt te od crs bt ny n pro- one only but scores, word the calculate speech. to government used were speeches government Irish had scores such available. been texts calculate reference language to possible perfectly been have would it Nonetheless words. language Irish for instance this in calculated be could scores no that so speeches ‘reference’ in their Irish use not did leaders party the while in Irish, entirely was speech 1994. December in was election an without government another by replaced was legislature the of confidence the lost have to deemed government a that time only The one. call and legislature the election, an always almost practice of dissolve to obliged constitutionally not is president the circumstances the these matter in although a and as lost is been there has resigned motion has confidence government a Once resign. must motion confidence a converts this into a confidence motion. Under the Constitution, a government that loses threatened motion of no confidence proposed by the opposition, is that the government government. Irish the of that of subset a as website main legislature’s RS PLTCL STUDIES POLITICAL IRISH Notes 70 Downloaded By: [Trinity College] At: 10:11 19 February 2008 171ips04.qxd 07/08/2002 08:50 Page 71 Page 08:50 07/08/2002 171ips04.qxd Budge, Ian, David Robertson and Derek Hearl, eds. 1987. 1987. eds. Hearl, Derek and Robertson David Ian, Budge, lnean Hn-itr Rcad ofret In ug, as ea, Torbjorn Keman, Hans Budge, Ian Hofferbert, Richard Hans-Dieter, Klingemann, Kleinnijenhuis, Jan and Paul Pennings. 2001. ‘Measurement of Party Positions on the Basis the PartyPositions on of ‘Measurement PaulPennings. 2001. and Jan Kleinnijenhuis, ar, on 20. Te optr oig f oiia Txs Rsls rm Britain, from Results Texts: Political of Coding Computer ‘The 2001. John. Garry, Bara, Judith. 2001. ‘Tracking Estimates of Public Opinion and Party Policy Intentions in Intentions Policy Party and Opinion Public of ‘TrackingEstimates 2001. Judith. Bara, Pennings, Paul. 2002. of ‘The Dimensionality the EU policy space’, Policy in Parties Political ‘Placing Garry. 2002. John and Benoit Kenneth Michael, Laver, Laver, Michael and Ian Budge, eds. 1992. 1992. eds. Budge, Ian and Michael Laver, Policy ‘Estimating 2001. Mansergh. Lucy and Giannetti Daniela Miranda, Vries, De Tannenbaum. Eric and Bara Judith Volkens, Anrea Klingemann, Hans-Dieter Ian, Budge, miia etmto o te rfrne o pltcl cos Rcn pbiain i these in publications Shepsle) Recent A. Kenneth actors. (with include political fields of preferences the of estimation empirical current was His Galway. research interests College and are in theories of party competition and government formation and in the University Dublin at Sociology College Trinity and Politics at of Professor Science previously Political of Professor is LAVER MICHAEL Texts’. Political from Positions Policy ‘Estimating 2000. Garry. John and Michael Laver, eevd i PD rm avr Uiest i 19. drs: eatet f Political of Department Address: 1998. in . E-mail: University Ireland. 2, Dublin Trinity RosterPlace, Science, 1 College, Harvard from PhD his received and 1996) oiia Sine Tiiy olg, nvriy f uln Hs eerh neet are interests research His Dublin. Eastern European of politics. He has published in a variety of journals, including the University methodology, College, research elections, comparative systems, and electoral and party comparative Trinity Science, Political of Department the in Studies Graduate of Director and Lecturer a is BENOIT KENNETH . E-mail: +353-1-677-0546. fax Department of Political Science, Trinity College, Dublin 2, Ireland. Tel.: +353-1-608-2036; Journal of PoliticalResearch of Journal Bergman, Rrançois Pétry and Kaare Strom. 1994. 1994. Strom. Kaare and Pétry Rrançois Bergman, EstimatingthePolicy Positions Politicalof Actors ed, Laver, Michael in VoterPerceptions’, and Coverage Media Programmes, Party of Politicalof Actors ed., Laver, Michael in Norway’, and Germany, Ireland Actors USA’, Laver,ed., the Michael and in Britain hne Sail nlss f otWr lcin rgams n 9 Democracies 19 in Programmes Election Post-War of Analyses Spatial Change: Boulder, CO: Westview.Boulder, CO: the Italy, from Results Texts: ed., PoliticalActors Laver, Political Michael of in Ireland’, Coding and Netherlands Computer the from Positions Press. University Oxford Oxford: 2001. Press. University Cambridge Cambridge: Spaces. Unpublished paper.Trinity Unpublished Dublin. Spaces. College 3:1, pp.59–80. 3:1, Politicalof AmericanJournal Science Macmillan. London: (London: Routledge), pp.217–36. Routledge), (London: apn Plc Peeecs Pris Eetr ad oenet: 1945–1998. Governments: and Electors Parties, Preferences: Policy Mapping Estimating the Policy Positions of Political Actors Political of PositionsPolicy the Estimating (London: Routledge), pp.193–216. Routledge), (London: OAIG D I PLCY SPACES Y POLIC IN TDS LOCATING (London: Routledge), pp.183–92. Routledge), (London: , Electoral Studies Electoral References 44:3, pp.619–34. 44:3, Making and Breaking Governments Breaking and Making 71 , and the the and , at Plc ad oenet Coalitions Government and Policy Party Estimating the PolicytheEstimatingPositions Political of (London: Routledge), pp.162–82. Routledge), (London: siaig h Plc Pstos of Positions Policy the Estimating Journal of Theoretical PoliticsTheoretical of Journal ate, oiis n Democracy and Policies Parties, Estimating the PolicyPositionsthe Estimating dooy Srtg ad Party and Strategy Ideology, (Routledge, 2001). Address: 2001). (Routledge, European Union Politics (Cambridge, European . He . . . . Downloaded By: [Trinity College] At: 10:11 19 February 2008 171ips04.qxd 07/08/2002 08:50 Page 72 Page 08:50 07/08/2002 171ips04.qxd aenR 036 .10–.41543489.0 384 85.4 1,534 375 93.3 –0.74 1,401 91.2 252 0.0110 85.6 –0.71 232 88.0 730 –0.3668 662 0.0120 625 88.2 475 87.9 3,320 –0.3647 –0.71 573 2,224 –0.68 –0.63 394 0.0150 87.5 88.9 –0.69 2,574 0.0180 0.0080 1,611 –0.3646 440 492 0.0090 –0.61 –0.3633 –0.3601 –0.60 2,062 89.1 0.0090 2,086 –0.3638 0.0110 85.5 –0.55 88.0 575 –0.45 93.4 –0.3592 –0.3585 0.0100 RR 415 0.0100 260 2,683 292 89.7 –0.3555 1,445 RR –0.3497 764 –0.42 1,038 582 –0.43 0.0090 84.9 –0.38 –0.35 3,314 88.9 0.0130 RG –0.3480 590 0.0130 Davern Lab 0.0170 –0.21 532 RR –0.3488 91.7 McDaid 3,546 Lab –0.3458 89.2 86.1 0.0080 –0.3441 2,352 G Reynolds 578 88.8 –0.19 RG 254 269 –0.3360 88.4 MD RG Higgins –0.18 3,241 0.0090 377 815 970 Howlin 0.0090 748 0.03 –0.3348 1,451 RR RG Roche 90.9 3,465 –0.3344 –0.04 0.12 0.06 0.0080 Noonan 613 Greens 0.14 0.0150 Lab Boylan –0.3222 0.0130 0.0150 2,789 81.4 0.0080 D –0.3263 Ahern RG –0.3167 –0.3202 0.30 689 RG Creed –0.3156 3,758 Ind Garland 0.0090 82.4 89.0 87.4 Quinn 0.16 –0.3063 90.8 703 RG 634 Browne 529 RG 0.0080 401 3,557 2,917 89.3 Kenny 2,068 –0.3144 1,571 0.39 Blaney 0.37 593 89.7 RR 0.70 85.4 HigginsJ 87.5 0.92 RR Lab 2,818 0.0080 0.0090 WP 488 Rlaherty 712 0.0110 Minister 611 RG 0.0120 0.50 –0.3010 –0.3020 1,963 O’Donoghue 4,178 –0.2833 3,250 RR –0.2702 Lenihan 1.10 0.0090 87.8 0.82 RG 0.93 Minister Minister Gilmore –0.2947 0.0110 742 0.0080 88.1 90.1 87.1 Stagg 0.0080 RR RR –0.2600 4,249 Yates –0.2758 791 261 743 –0.2697 Minister 1.00 85.2 Burke 92.6 87.3 4,155 669 3,697 PD Barry 754 Minister 393 0.0080 1.28 1.33 674 Minister Brennan 1.80 4,440 RG 1,238 –0.2659 3,219 0.0080 84.8 RR 0.0090 Rlynn RR RR – 1.71 – 88.4 0.0200 2.92 1.48 O’Malley –0.2495 763 –0.2462 873 –0.2194 Minister 0.0080 RitzGerald 1536 0.0150 0.0090 3,944 1181 6,226 – ORourke 4,474 RR 4,375 –0.2245 –0.1542 Minister RR Minister 2.15 –0.2377 Cowan 2.14 (Jr) Minister 1617 0.0090 RR Daly RR – % 6,711 – Minister 0.0080 OKennedy RR –0.1990 Unique –0.1991 Hillery RR Total – OHanlon Minister Standard- – Woods Minister – RR Leyden RR –1.0000 –1.0000 RR RR SE Collins – Cullimore 1.0000 A Reynolds Leader Leader Wilson Raw DL Taoiseach Nolan RG VirginTexts RR Rossa de Position Bruton Haughey Party TextsReference Speaker Raw and Standardised Scores of Speakers in 1991 Confidence Debate on Debate Confidence 1991 in Speakers ofStandardised Scoresand Raw oeie eghwrs words words* length ised sore ‘Pro- versus Anti- Government’ Dimension Government’ Anti- versus‘Pro- RS PLTCL STUDIES POLITICAL IRISH Appendix 1: Appendix 72 cr nwrsscored words in score Downloaded By: [Trinity College] At: 10:11 19 February 2008 171ips04.qxd 07/08/2002 08:50 Page 73 Page 08:50 07/08/2002 171ips04.qxd uknR 040 .10–.9882289.1 262 82.4 868 542 85.2 86.2 88.2 –1.49 3,012 285 87.1 372 0.0130 342 –1.05 946 508 1,243 –0.4104 1,191 0.0080 85.0 88.2 2,414 –1.14 –1.02 –0.3845 –1.03 641 90.8 361 0.0130 –1.00 0.0120 0.0140 86.4 3,031 0.0090 1,273 248 –0.3898 –0.3827 –0.3834 90.0 924 –0.97 –0.96 –0.3820 730 0.0090 361 0.0120 6,396 85.2 (2002). Garry and Benoit Laver, per as computed are –0.96 errors Standard 1,438 text. –0.3800 the –0.92 –0.3797 in words of number 347 total the to relative RG texts reference 0.0150 the from scorable 89.3 words 0.0060 (non-unique) –0.85 total the to refers scored 1,188 % 90.0 –0.3792 RG 0.0110 537 –0.3770 Note: –0.82 Unique 438 –0.3730 2,460 0.0130 Total Lab RG 1,839 Durkan RG –0.76 Standard- –0.3711 RG –0.79 Connaughton 0.0090 0.0100 Owen –0.3679 WP SE RG Deenihan –0.3693 Rerris RG Lab Deasy Rabbitte Raw RG Currie RG Rinucane Spring Lab Ahearn Position RG BrutonR TaylorQuinn Party Shea O’ Speaker *Unique words for virgin texts refers to scored words only. words scored to refers texts words virgin of for words percentage The *Unique oeie eghwrs words words* length ised sore OAIG D I PLCY SPACES Y POLIC IN TDS LOCATING Appendix 1 Appendix 73 (Cont’d) cr nwrsscored words in score