images of protest john mccarthy

Upload: franklinramirez8

Post on 14-Apr-2018

217 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/29/2019 Images of Protest John Mccarthy

    1/23

    IMAGES OF PROTEST:DIMENSIONS OF SELECTION BIAS IN MEDIA COVERAGE OFWASHINGTON DEMONSTRATIONS, 1982 AND 1991*

    John D. McCarthyCatholic University of America Clark McPhailUniversity of Illinois. Urbana-Ch ampaign

    Jackie SmithUniversity of Notre DameProtest is now central to politics in W estern demo cracies, but it is know n tocitizens m ainly through portrayals in the med ia. Y et the media cover only asmall fraction of public protests, raising the possibility of selection bias. Westudy this problem by comparing police records of demonstrations in Wash-ington. DC. in 1982 and 1991 with media coverage of the events in TbeNew York Times, The Washington Post, and on three national television net-works. We model the consequences of demonstration form, context, and pur-pose on the likelihood of media coverage. The estimated size of a demonstra-tion and its importance to the current media issue attention cycle are thestrongest predictors of its coverage. Additional analyses support our claimthat heightened media attention to an issue increases the likelihood that pro-tests related to that issue will be covered. Comparing 1982 to 199} suggeststhat television coverage of protests is increasingly subject to the impact ofmedia issue attention cycles.

    Over the past quarter-century protestdem onstrations have become a centralpart of the process of political representationin Western Democracies (e.g., Barnes andK asse et . al 1979; Dalton 1988 ; Tarrow,1988, 1989, 1993 ; Tilly 1983). Tbe dem on-stration became entrencbed in the tacticalrepertoire of citizens* movements with therise of the European labor movetnent (Tilly* Direct all correspondence to John D.McCarthy, Department of Sociology, CatholicUniversity of America, Washington, D.C. 20064(mccarthj@cua,edu). This research was partiallysupported by the National Science Foundation(Grants #SES 91-22691, #SES 91-22732, #SBR93-20488. and #SBR 93-20704),We thank John

    Crist. Bob Edwards, Jeff Goodwin, KevinEverett. Doug McAdam, Sam Marullo, Susan

    1984), and it has become ever more commonand widely legitimate since the middle of tbetwentietb century. Tbe relative w eight of citi-zen preferences is now publicly and sociallycons t ructed wi tb evidence f rom severalsources, including elections, public opinionpolls , and demonstrat ions (Herbst 1993).Media reports of protest signal elites aboutcitizen discontent expressed outside the moredirect and conventional channels of politicalrepresentation. As a result, wbetber and bowtbe media cover demo nstrations play a grow-ing role in structuring democratic outcomes.Tbese same media reports bave become acommon source of evidence for researcberswbo describe and interpret the frequency,form, size, duration, and intensity of protestdemonstrations. Nevertheless, tbere is wide

  • 7/29/2019 Images of Protest John Mccarthy

    2/23

    SELECTION BIAS IN MEDIA COVERAGE OE PROTEST 479plored using methodologically appropriate,systematic evidence. Nor has any previousresearc h emp loyed a source of evidenceabout protest demonstrations independent oftheir media traces.'Our research uses the demonstration per-mit application records of three police agen-cies to con struct a credib le and ob jectiverecord of the population of demonstrations inW ashin gton, D .C. in 1982 and 199 1. Wecompare these records with reports of Wash-ington demonstrations that appeared in localand national print and electronic media in thesame years: The Washington Post, The NewYork Times, and the ABC, CBS, and NBCnightly national television network news-casts.THE PROBLEM OF SELECTION BIASScholars from a variety of perspectives ac-knowledge the centrality of protest demon-strat ions to normal poli t ical processes inWestern democracies. With the increasingfrequency of demonstrat ions, the averagecitizen views this form of protest as commonand acceptable, not as an unusual or deviantform of political behavior (Dalton 1988).During the last several decades we have seenan upsurge of research on demonstrationsand other forms of collective action, drivenin large measure by the growing appreciationof the central role of protest events in demo-cratic politics. A growing community of re-searchers has focused attention on collectiveaction events in much earlier times (Markoff1986; Tilly, Tilly, and Tilly 1975; Tilly1978), upon recent events in the U .S. context(Everett 1992; Jenkins and Perrow 1977;M cA dam 1982, 1983; Olzak 1987; 1989a;1989b; 1992) and upon events around theworld (Feierabend and Feierabend 1966;Gu rr 1968; Paige 1975; Rucht and Ohle-mach er, 1992; Rum mel 1963; 1965; Tanter

    ' Direct observation research of protest demon-strations is very labor intensive and rarely under-

    i966 ; Tarrow 1988; Tilly, Tilly, and Tilly1975). Protest event data sets have been cre-ated from diverse source materials includingofficial archives (Tilly, Tilly, and Tilly 1975)and organizational histories (McCarthy et al.1988). However, because newspapers (andnow electronic news reports) are often theonly source of data on protest events in manyplaces and times, these media sources havebeen rel ied upon heavily by researchers(Franzosi 1987; Olzak 19 89a).The use of such media traces as an indexof protest events is not without problems,and these problems have been identif iedthrough debates among researchers and crit-ics over the adequacy of event data(Danzger 1975 and subsequent response byJenkins and Perrow 1977; Snyder and Kelly1977). These problems include: (I) mediabias in the selection of but a few of themany possible events to observe and report(selection bias), (2) media bias in the de-scriptions of the events they do select to re-port (description bias), and (3) the reliabil-ity and validity of media trace recovery byresearchers ( researcher b ias) (Franzosi1987; Olzak 1989a). Each of these sourcesof bias may lead researc her s to faulty infer-ences about the characteristics of the popu-lation of protest events they hope to de-scribe and understand.'^Because for some protest events newspa-pers are the only source of data, researcherswho use them have tried to systematicallyassess the bias in newspaper usage. As Olzak(1989a) observes, however, "there is rarely away to evaluate these claims directly sincefew alternative sources contain as much in-formation" (p. 128). Two studies have com-pared The New York Times reports of collec-tive action in communities around the coun-try with local community newspaper ac-counts. Jenkins and Perrow (1977) found fewdifference s in the cov erag e of Ca liforniafarm workers collective actions reported inThe New York Ttmes and in The Los Angeles

  • 7/29/2019 Images of Protest John Mccarthy

    3/23

    480 AMERICAN SOCIOLOGICAL REVIEWTimes. But Snyder and Kelly (1977) foundtbat national and local newspaper accountsdiffered in their reports of size, violenceagainst property and persons, and number ofarrests. Moreover, substantive interpretationsof media traces of protest events have typi-cally assumed tbat bias is stable across time(M cA dam 1982; Tarrow 1988). Unfortu-nately, none of tbese analyses or proposedsolutions to tbis problem of selection biasbave identified, let alone adequately esti-mated tbe effects of selection bias on the me-dia traces used to study collective actionevents.

    The appropriate assessment of tbese selec-tion bias issues requires a credible, objectiverecord of tbe population of protest events insome location over several specified periodsof time. Such a record can then be comparedwith media trace records of protest in tbatlocation for tbe same time periods. Multipleand independent media traces of events atdifferent intervals would be idealdifferentmedia sources could be compared for tbe bi-ases tbey introduce in tbe events selected forcoverage as well as the stability of tbose bi-ases (Fran zosi 1987).M E C H A N I S M S O F S E L E C T I O N B IA SA rich and varied body of critical and ana-lytical work on the mass media bas emergedduring tbe last several decades which bas ex-amined tbe mecbanisms of media selectionand description bias (Dan zger 1975; Gan s1980; Herman and Chomsky 1988; Ryan1991; Snyder and Kelly 1977). Media ana-lysts have crafted several general lines of ex-planation for bow story selections are madefrom a vast amount of available Informationto fill a very limited "newsbole," Each oftbese perspectives provides clues for under-standing tbe mecbanisms of selection bias intbe reporting of protest even ts. We briefly re-view tbe more coberent perspectives thathave guided our assessment of the evidence.

    official sources ftlters out many otber possib le source s of pol i t ica l informat ion(Gamson et al. 1992; Kielbowitz and Scbere1986). Deadlines, lead times, staffing, and threlative flow of information also affect the selection of news. We expect, tben. that protesevents fitting established news gatbering routines (e.g., tbose occurring near reporters' assigned locations or on weekends) will bmore likely to gain media attention.Newsworthy "Pegs"Tbe professional incentive structures facedby reporters within large media organizationencourage their reporting of events whicbprovide "news pegs" around wbicb a storycan be constructed (Ryan 1991). The acto rand actions in events that make "good" newpegs typically bave one or more of tbe following characteristics: notorious (e.g.. thactors' faces are famous, the actions taken othe objects of tbose actions are "trendy")consequential (e.g., tbe actors are powerfultheir actions or tbe events tbey make up havewide impact); extraordinary (e.g., tbe actionare spectacular, tbe events are large, or otberwise unusual and tbus of broad human interest); culturally resonant (e.g.. actors, actions, events illustrate, bigbligbt or empbasize that wbich is widely familiar). As reporters compete to get tbeir stories printed andtbereby to advance professionally, tbey musproduce copy tbat stands out. Clever newpegs often suit these purposes.^ Followingtbis reasoning, we would expect that demonstrations distinguished in one or more of tbeabove ways will be more likely to gain media attention.

    ' Reporters are assigned to beats or stories bytheir editors. While reporters have varying degrees of autonomy, editors have the final authority concerning whether a story is printed, wherit appears, with or without byline, and the likeThus, a reporter must figure out what the editowants (or what latitude a particular editor wil

  • 7/29/2019 Images of Protest John Mccarthy

    4/23

    SELECTION BIAS IN MEDIA COVERAGE O F PROTEST 481Corporate Hegemony ^News media in the United Slates and othercapitalist nations are profit-making institu-tions by design, and they are often very suc-cessful ones. They depend at least as muchon advert isers as on consumers for theseprofits. Consequently, it is argued, they canbe expected to select and shape news eventsin ways that do not threaten their own orthei r sponsor 's in terests (Gamson et a l .1992). Other things equal, this leads to theexpectation that demonstrations challengingelite corporate interests (e.g., pro-labor, pro-environment, anti-petroleum industrial com-plex) will less likely be selected for mediacoverage than those that do not.Media Issue Attention CyclesThere is more to the media's story selectionprocess than just news gathering routines,news pegs, and vested corporate interests.There are fluctuations in the amount of at-tention the media give to various issues(Dow ns 1972).^ "M edia i ssue a t tent ioncycle" refers to the sudden ascendance of anissue from previous obscurity to a sustainedprominence (indexed by the number of sto-ries, by column inches, or by minutes andseconds of a telecast) that dominates thenews for a period of time before once againfading from media attention. Some issue at-tention cycles are closely tied to dramaticevents {e.g., the Vietnam War or Watergate),but others (e.g.. poverty, environmental pol-lution, nuclear power) are not so obviouslylinked to objective trends (Gamson andModigliani 1989). Because "issue attentioncycles" imply the movement of issues on andoff the mass media agenda, such cycles mayinfluence the short-term news worthiness ofa particular demonstration event. Most of the

    '* Our statement is a truncated version of the'propaganda model" as outlined by Herman and

    analyses of issue cycles have featured themobil izat ion of media at tention by morehighly organized, wealthy, and media-so-phisticated collective actors than the typicalsocial movement actors who organize dem-onstrat ions (Baumgartner and Jones 1993;Hilgartner and Bosk 1988; McCarthy, Smith,and Zald 1996 ;Ne um an 1990).*Issue cycles may bias story selection inpart because they can push other issues outof the media limelight, but more so becausemedia personnel have a tendency to connectthese cycles with ongoing events, selectingfor stories those events which will receivesubsequent media at tention depending onebbs and flows of media attention cycles.Recognition of this pattern led one mediaconsultant to recommend that corporate me-dia attention seekers "surf" the waves of is-sue attention cycles (Nolan 1985). These ar-guments lead to our expectation that, otherthings equal, demonstrations around those is-sues easily coupled with contemporaneouspeaks in media issue attention cycles willmore likely be selected for coverage thanothers.Washington Demonstrations as NewsWe have noted the pervasive consensus thatdemonstrations are now a central and legiti-mate feature of poli t ical part icipation inWestern democracies. Demonstrat ions arefielded across communities large and smallthroughout the United States, but they are es-pecially likely to take place at seats of statepower, such as municipal commons, statecap itals (Lofland and Fink 1982). and theU.S. National Capital in Washington, D.C.Citizens who seek national recognition oftheir gr ievances are l ikely to choose thenation's Capital as the site for their demon-strations (Etzioni 1970). Between one andtwo thousand protest events occur there eachyear, providing an excellent opportunity toexamine the dimensions of media selectionbias we have just summ arized.

  • 7/29/2019 Images of Protest John Mccarthy

    5/23

    482 AMERICAN SOCIOLOGICAL REVIEWCREATING AN OBJECTIVE RECORDOF WASHINGTON, D.C.DEMONSTRATIONS

    Defining Demonstration EventsOur unit of analysis is consistent witb mostof the criteria of other scholars wbo defmeprotest demonstrations as collective actionevents: gatberings of two or more people inwbicb a visible or audible "claim is madewbich, if realized, would affect tbe interestsof some specific person(s) or group(s) out-side tbeir own numbers" (Tilly 1978:275).^Our unit of analysis is furtber specified as ademonstration with clear assembling and dis-persal phases. Thus, a 72-hour continuousvigil on the Wbite House sidewalk is countedas one demonstration event; a daily two-hourpicket that is repeated for one week is re-corded as seven discrete demonstrationevents. Finally, all recorded demonstrationsinvolve political or religious claims, or mes-sages of protest and/or celebration of someissue, principle, actor, action, or event.

    DataWe developed an objective record of demon-stration events in Washington, D.C. usingdemonstration permit records from the three

    ^ We take e xception to the frequent limitationof co l l ec t i ve ac t i ons (demons t ra t i ons ) tononroutine events and to the exclusion of actionsby members of government. For example. Olzak(1992:54) excluded "anniversary marches or an-nual celebrations since their timing is predeter-mined and not spontaneous," But neither are thecollective actions she included. Large demonstra-tions in the United States arealmost always timedfar enough in advance to permit notification andmobilization of participants; the lag t ime betweennotif icat ion and assembling al lows addit ionalplanning and preparations for the target, place,and form of protest. It is difficult to characterizeas spontaneous any of the demonstrations withwhich we are famil iar , even those that were

    primary policing agencies tbere: tbe NationalPark Service (NPS), tbe U.S. Capitol Police(USCP), and tbe D.C. Metropolitan Police(MPDC). Eacb agency requires tbat allgroups planning to demonstrate witbin its ju-risdiction apply for a permit to do so.^Guidelines have been establisbed, particu-larly in the past two decades, to regulate sucbevents (McCartby. McPbail, and Smith1994). A format permit system in eacb oftbese tbree jurisdictions defines tbe legalconstraints on public gatherings and specifiesdemonstrators' First Amendment guaran-tees.^

    The archival records of permits in tbesetbree policing jurisdictions in Wasbington,D.C. constitute an official record of public

    " The U.S. Capitol Police are responsible fordemonstrat ions on the grounds surrounding theCapitol and the House and Senate office build-ings. The Metropolitan Police are responsible fordemonstrations on the city streets. The NationalPark Service is responsible for demonstrat ions onmonument grounds, the Mall, the Ellipse behindthe White House, and Lafayette Park across thestreet from it as well as most other parks through-ou t the city.^ In the District of Columbia no permit is re-quired to hold vigil or picket on non-U.S. gov-ernment property (e,g.. sidewalks), provided thatthe demonstration does not block pedestrian traf-fic or interfere with entry or exit to adjoining pri-vate or public properties, If demonstrators enterthe street for a procession or march, they mustobtain a permit from the MPDC. Demonstrat ionsthat take place on the U.S. Capitol grounds, in-cluding the House and Senate office buildinggrounds, require permits from the USCP, TheNPS. a division of the US, Department of the In-terior, distinguishes between two types of gather-ings for which permits are required.

    "Demonstration" includes [First Amendmentlspeeches, pickeiing, vigils, etc. and all similar ac-tivity designed to communicate a message of somekind, "Speeial event" includes any presentation.program, or display |e,g,, sports events, pageants,celebrations, historical reenactments, rega ttas, enter-tainments, exhibitions, parades, fairs, festivals andsimilar events, including those presented by the Na

  • 7/29/2019 Images of Protest John Mccarthy

    6/23

    SELECTION BIAS IN MEDIA COVERAGE OF PROTEST 483

    gatherings, some of which are First Amend-ment (demonstration) gatherings and some ofwhich are Special Event gatherings. FirstAmendment gatherings can be further di-vided into political and religious demonstra-tions. In this paper we examine only politi-ca l First Amendment events, excluding reli-gious First Amendment events as well asnonpolitical special eve nts. '" This m akes theanalysis presented bere comparable to the al-tnost exclusive focus in the existing literatureon political collective action events."

    For every proposed demonstration policingagencies solicit extensive information fromapplicants. The records provide details on theindividual or organization sponsoring tbeevent, tbe purpose, location, time, duration,form, activities, equipment, number of ex-pected participants, and wbether the appli-cants anticipate that counter demonstrators

    '"Mindful that NPS might classify as specialevents some gatherings we would consider as po-litical or religious demonstrations, we carefullyexamined and coded permit applicants' statedpurposes for their proposed gatherings accordingto our criteria for demonstrations. For the analy-sis reported here, we excluded religious groupapplicants whose stated purpose was "spreadingthe gospel" or "evangelizing"; we Included reli-gious group applicants whose stated purpose wasadvocating or protesting legislation, electoralcandidates or other political agenda." Throughout this paper we interchangeablyuse the terms demonstration event, protest event,collective action event, and First Amendmentevent. By demonstration we mean those publicgatherings of two or more persons in the samespace and time location whose modal behaviorsare individual or collective, visible or audibleclaims, which protest or advocate some political(or religious) principle, actor, actions or condi-tions (e,g., placarding, leafleting, petitioning, ta-bling, as well as picketing, vigiling, rallying,marching, chanting, gesturing, and the like).Demonstrations may involve two or more re-lated gatherings, as when people assemble for aninitial rally and later march to another locationfor another rally or to join or show support for a

    may appear to possibly disrupt the event.Some of these pertnits are public records andare open to public scrutiny tbrough the Free-dom of Information Act.'^We inspected dem onst ra t ion p er tn i trecords for 1982 and 1991 from tbe NationalPark Service, United States Capitol Police,and the Metropolitan Police of the District ofColumbia, Official constraints prevented usfrom copying tbe original application per-mits, tberefore. we developed forms and pro-cedures tbat enabled on-site coding of rel-evant information embodied in tbe records.'^Many of tbe permit applications represent

    more than one demonstration (e.g., permitsto picket or vigil for several days in a row).Intercoder reliability estimates were ,90 orbigher for tbe variables we created from theseveral permit forms (described below).Validity of the DataThis body of permit records provides a fairlycomprehensive portrait of collective actionevents in Washington, D.C. There are threepotential difficulties witb official records ofsucb events. The first is that some groupswho engage in collective action never applyfor a permit. The NPS does not require a per-mit for protests involving fewer than 25people, although it strongly encourages per-mits as they facilitate the efficient and or-derly public use of national park space. It islikely, then, tbat some smaller protests withinPark Service jurisdiction will not be includedin the permit record. Interviews with ParkRangers responsible for the permitting pro-cess indicate, however, that "unpermittedprotests" constitute no more than 5 percentof total gatherings.Tbe U.S. Capitol Police require permitseven for solo demonsti^ators. Given theirrelatively smaller area of jurisdiction andextensive surveillance, it is our judgementtbat they know about practically all protestson Capitol Hill. Groups protesting without

  • 7/29/2019 Images of Protest John Mccarthy

    7/23

    484 A M E R I C A N SOCIOLOGICAL REVIEWTable 1. Percentage Distributions of the Esti-mated Size of Washington Demonstra-t ions: From Demonstrat ion PermitRecords, 1982 and 1991EstimatedDemonstration SizeLess than 2626-100More than 100Mean sizeMedian sizeMaximt]m sizeTotal number ofdemonstrations

    198260.4%17.0%22.7%

    1,21820

    500,0001,209

    199153.4%20.8%25.8%741

    25100,000

    1,856

    permits are, typically, assigned permits onthe spot.The perm itting of protests within the juris -diction of the MPDC is governed by prevail-ing "Public Forum" law and precedent thatgoverns most public space in the UnitedStates. This does not require a permit forpeaceful protest in traditional public fora oron public sidewalks, as long as they do notinterfere with pedestrian or vehicular traffic(cf. An 1991; Snyder 1985). As a result, itcan be expected that large numbers of pro-tests can occur in MPDC jurisdiction withoutperm its. We estimated the num ber of "unper-m itted " events that occurred in both 1982 and1991 from a logbook kept by an MPDC desksergeant for demonstrations encountered bypatrol officers but for which no permits hadbeen issued. In 1982, 61 percent, and in 1991,68 percent of all protests known to the MPD Cneither applied for nor received permits, andwere only recorded in this logbook,''^Our confidence in the comprehensivenessof the permit record is substantiated by thefact that only a very small number of dem-onstrations that appear in the media recordare not in the permit record. These excep-

    tions are almost always one of the following:a demonstration that took place at the Dis-trict Building (the seat of the local D,C. gov-ernment), which allows demonstrations butdoes not have its own permitting system; ademonstration on a sidewalk in front of a ho-tel or embassy, which do not require a for-mal permit; and an instance of civil disobe-dience, where demonstrators protested by re-fusing to comply with permitting procedures.

    The second potential difficulty with ourdata sources is that a l l permit ted protestevents do not actually take place. However,the effort required to obtain a permit ordi-narily assures that most applicants will carrythrough with their planned demonstrations.U S . Capitol Pol ice records establ ish thatonly 2.3 percent of the events permitted bythe USCP did not take place. Extrapolatingfrom these data, we estimate that 97.7 per-cent of the permitted demonstrations did in-deed take place. We see no reason why thisestimate should be much less for the othertwo police agencies. '^We recognize another validity issue. Dem-onstration size estimates appear on the offi-cial permits in advance of the actual detnon-stration event. Although these are estimates,we believe they are not excessively biased.Agency officials are experienced in anticipat-ing numbers of protesters (especial ly forlarge demonstrations) because they must de-termine and justify personn el ne eds based onthese estimates. In the few cases when these

    officials doubted applicants' claims of num-bers, they entered a revised size estimate onthe permit ; in these instances our coderswere instructed to record agency rather thanapplicant estimates of demonstration size.I ITHE POPULATION OF WASHINGTON,D.C. DEMONSTRATIONSIN 1982 AND 1991We summarize here some important charac-teristics of protests in 1982 and 1991 based

  • 7/29/2019 Images of Protest John Mccarthy

    8/23

    S E L E C T I O N BIAS IN MEDIA COVERAGE OF PROTEST 4851991 , derived from the permits. For eachyear more than half of the protests involved25 or fewer demonstrators, and about three-quarters of them involved fewer than 100people. These data establish that small politi-cal protests have become a daily fixture inWashington: The tactic of "voting with one'sfeet" has been adopted by a wide range ofcauses and movements. Note, too, that thesize distribution as well as the median sizeof p ro tes ts has remained fa i r ly constan tacross the two periods, while the mean sizeis quite different. This derives from the factthat no very large protes t dem onstra t ion(e.g., in the 200,000 range) was held during1991 . Finally, note that the total number ofprotests increased between 1982 and 1991. '^Table 2 shows percentage distributions ofthe other independent variables in our analy-ses of media selection bias. They are arrangedin th ree major ca tego r ies : dem onst ra t ionform, demonstration context, and demonstra-t ion purpose (or issue) . The repertoire ofdemonstration forms includes the familiarrally, march, vigil or picket, plus leafletingand petitioning although these sometimes ap-pear in less familiar combinations. For ex-ample, while many demonstration rallies areself-containe d, ral l ies are also frequentlycombined with marches; sometimes beforeand after, sometimes only before or only af-ter, and sometimes punctuating the march atseveral points enroute . An other form involvesdemonstrators standing and holding placards(here termed a vigil) or walking a rotating oroscillating but fixed route (here termed a

    Table 2. P e r c e n t a g e D i s t r i b u t i o n or D i c h o t o -mous Independen t V ar i ab i es : Wasb ing-ton, D.C, 1982 and 1991

    "" The Asso ciated P ress has assem bled, frompast media reports, a list of 15 mass demonstra-t ions in Washington, D.C. with est imated sizes of200,000 or more between 1967 and 1993 (TheWashington Times. April 26. 1993, p. A6). Twogatherings of this size occurred in 1982: The VietNam Memorial Dedicat ion, est imated from per-mit records at 500,000. and a demonstrat ion hon-oring Viet Nam Veterans the preceding day, est i -mated f rom permi t records a t 250,000. Thesewere not included on the AP list.

    Percent inIndependent VariableForm'

    RallyMarch/RallyVigil/PicketLiterature Table

    Context^Counter-DemonstrationCampaignWeekend

    Purpose (Top 10 Issues)*^Foreign governmentGulf WarLatin America/peaceMiddle East-Lebanon WarSchool prayerAnti-nuclear weaponsEqual Rights AmendmentVeterans' issuesHomelessnessPro-environmentWomen's issuesDemocracy for HaitiLaborHealth careJobs/economyCivil rights

    Total num ber of 1permitteddemonstrations

    1982

    9 .917.351.310.2

    6.274.332.7

    17.0

    13.07.67.35.44.42 .92.62.52.3

    i,209'-

    Percent in1991

    24.213.135,623.6

    3.262.833.9

    19.914.3

    _7.27.05.33.93.32.41.90 .6

    1,856

    " Each form variable is a dummy variable (1 = theform. 0 = not the form). There were only a small

  • 7/29/2019 Images of Protest John Mccarthy

    9/23

    486 AMERICAN SOCIOLOGICAL REVIEWpicket) in the demonstration venue. We havecombined these latter cases into a single cat-egory. Leafleting, and occasionally petitionsignature solicitation, occurs in conjunctionwith most rallies, some marches, and a fewvigils or pickets. Our observations of Wash-ington demonstrations, coupled with the per-mit application evidence we have described,led us to create a demonstration form cat-egory that combines leafleting and petition-ing activities at a stationary (often a cardtable) location. We refer to the form here andsubsequently as a "literature table." Demon-stration context refers to whether a demon-stration was held on a weekend or weekday,whether it was part of an ongoing campaignof demonstrations, and whether or not therewas any indication on the permit that the ap-plicants expected counter-demonstrators tobe present. Demonstration purpose refers toour classification of the applicant's answer toan explicit question on the permit applicationabout the purpose of the demonstration.

    Table 2 also summarizes several changesin the nature of Washington demonstrationsbetw een 1982 and 199 1. For instance , therewere a smaller percentage of demonstrationsin 1991 at which counter-demonstration ac-t ivi ty was expected by permit applicants(three percent as compared to 6 percent in1982). There were also fewer protests thatwere part of a campaign, by which we meana series of demonstrations fielded over mul-tiple days by the same sponsor on the sameissue. One-third of the demonstrations tookplace on weekends in both years.The distribution of forms of protests alsochanged somewhat between the two periods,with rallies and literature tables becomingmore common. There was a decline in thepercentage of protests that were vigil/picketsand a small decline in the percentage thatwere march/rally combinations. Note that arelatively small proportion of demonstrationsare marches, and therefore we reasoned thesewould be more likely to gain media cover-

    in Table 2. While demonstrations took placearound a large number of issues, over 60 per-cent of them focused on I of 10 issues eachyear, and no issue outside of the top 10 ineither year was the focus of more than 1 percent of the demonstrations. The most fre-quent target of demonstrations in both 1982and 1991 was some aspect of a foreigngovernment's policy (17 percent in 1982 and20 percent in 1991). Often many of thesedemonstrators were foreign nationals or immigrants protesting at embassies or on thesidewalk in front of the White House. The is-sue attracting the next largest number ofdemonstrations for both years involved someaspect of U.S. foreign policy (13 percent onCentral American poiicy in 1982, and 14 per-cent on the Gulf War in 1991), as did thethird largest for 1982, the Is raeli-L eba nes ewar. Demonstrations around Veterans' issueswere common in both years, with a fairlydramatic increase in 1991.'^ Women's issueswere a frequent focus of dem onstra tions inboth years. In 1982 these centered around thedebate over the Equal Rights Am endm ent; in1991 they focused on pro-choice and healthissues as well as on the debate over the ap-pointment of Clarence Thomas to the Su-preme Court.

    M E D IA T R A C E S O F W A S H I N G T O N ,D .C . D E MO N S T RA T I O N SDataWe followed stand ard procedu res develop edby other scholars for gathering and reducingthe media traces. We selected media sourcesthat enable comparisons between nationaland local origins as well as between print andelectronic media. The New York Times wasincluded in our study because it is the na-tional print source most widely used by col-lect ive act ion researcbers. We chose The

    " In 1991 the Park Service issued permits for

  • 7/29/2019 Images of Protest John Mccarthy

    10/23

    SELECTION B U S IN MEDIA COVERAGE OF PROTEST 487Washington Post because we expected it toprovide the most extensive print coverage oflocal collective action events. In light of tbecitizenry's increasing reliance on televisionnews rather than newspapers for informationon social and pol i t ical developments(Iyengar and Kinder 1987), we also includednational television news telecasts. We usedth e Vanderbilt Network News Index and Ab-stracts to access the national nightly newstelecasts by ABC, CBS, and NBC.Because we were skeptical tbat The Wash-ington Post index for 1982 was compreben-sive (cf Everett 1992), we read The Wash-ington Post for tbe entire year, copied andcoded all stories and mentions of collectiveaction events in Washington, D.C. The 1982The New York Times Index appeared to pro-vide a reliable summary of stories on Wasb-ington, D.C. collective action events, andwas used to locate relevant articles from1982. We used similar procedures to gatherprint m edia traces for 1991. We also located( through di rect reading of newscas t ab-stracts), copied, and coded the VanderbiltNetwork News Abstracts for ail mentions ofdemonstrat ions in Wasbington, D.C. onABC, CBS, and NBC nightly news telecastsfor botb years. '^For the analysis reported here, we are onlyconcerned witb the presence or absence ofmedia coverage by eacb source for everyevent, rather than tbe content of that cover-age. Tberefore, each news account was reador audited to determine which event(s) it de-scribed, and tbat event record was coded asbeing reported by tbat source.'^

    '"The Vanderbilt Ncrivor*: News Index and Ab-stracts includes a subject index and brief synop-ses of each news item reported during the eveningnational news telecasts of the major networks. Weinitially limited our search to the indices, but laterfound that a direct read of the abstracts provideda more comprehensive list of stories. Our analy-ses are based upon evidence generated through adirect read of both years of the Abstracts.

    The Media Portrayal of WashingtonDemonstrationsAs expected, tbe picture of Washington dem-onstrations portrayed in tbe mass media dif-fers dramatically from that generated fromdemonstration permit records. Table 3 dis-plays and Figure I pictures tbe distributionof Washington demonstrations across six sizecategories for 1982 and 1991 , and the distri-bution of demonstrations noted in any mediasource across those same size categories.For 1982, by far tbe largest proportion ofdemonstrations (60 percent) fall in the small-est size category, consistent with our earlierdiscussion. Despite their greater frequency,demonstrations in the smallest size categoryare the least likely to receive media coverage(2.9 percent). Much larger proportions ofdemonstrations from categories with over1000 part icipants are reported, and evenlarger proportions of the largest demonstra-tions are reportedthere were 2 larger than100,000 in 1982, and both received coverage.A similar pattern is evident in 1991 , altboughit is sligbtly less dramatic.^'* In tbe midsizecategories (e.g. . 26-100, lOl-I .OOO, and1,001-10,000) a smaller percentage of dem-onstrations is covered by any media sourcein 1991 tban is tbe case in 1982.While the media characterization of thepopulation of Wasbington demonstrat ionsclearly under represents tbe vast number ofsmall demonstrat ions, scholars have longrecognized the flawed assumption that everydemonstration bas an equal likelibood of me-dia coverage (cf. Jenkins and Schock 1992;Snyder and Kelly 1977). It is also recognizedthat larger demonstrat ions may representmore widely held views and tberefore aremore likely to be covered. As a result, onemigbt argue tbat a substantially greater pro-portion of citizen demonstrators are encom-passed by media coverage tban the small

    ^"This is largely a result of the fact that muchof the media coverage of 1991 demonstrations (as

  • 7/29/2019 Images of Protest John Mccarthy

    11/23

    488Table 3. Percentages of Annualportion Covered within

    1t 1 _

    Dem onstration Size Less than 2626-100101-1,0001.001-10.00010,001^100,000 rMore than 100,000 'Total percent INumber of demonstrations

    AMERICAN SOCIOLOGICAL REVIEW

    Washington, D.C. Demonstrat ionsEach Size Category by any Media1982 Demonstrations

    PercentPermitted"60,417.015.85.90.80.2

    100,01,209

    PercentReported''2.9

    12.334.747.1

    100.013.0158

    by Demonstrat ion Size and Pro-Source, 1982 and 19911991 Demonstrations

    PercentPermitted"53.420.820.74.11.0.

    100.01.856

    PercentReported^3.37.7

    11,929.837.5

    7.1133

    * Percent permitted represents the percentages of all demonstrations for the year by size category. Thetotals sum to 100 percent,'' Percent reported represents the percentage of all dem onstrations in the size category that w ere rep ortedin any media source. The totals are the percentage of all demonstrations reported in any media source.

    number of actual demonstrat ions that arecovered would imply. To assess this supposi-t ion we summed the permit est imates ofnumbers of demonstrators who took part inthose demonstrations covered by the mediato determine its proportion of the total poolof demonstrators participating in Washingtonprotests in each yearBy aggregating all demonstrators who par-ticipated in demonstrations reported by themedia, we get a somewhat different pictureof the extent of coverage than we get fromexamining the number of protest eventsalone . The percentage of dem onst ra t ionscovered by each media source and the per-centage of estimated number of demonstra-tion participants covered for each year aredisplayed in Table 4,For 1982, med ia coverage appe ars to be farmore comprehensive when judged by theproportion of demonstrators covered in con-trast to the proportion of demonstrations cov-ered. At least 60 percent of demonstrators are

    This 1991 pattern, in contrast to 1982,clearly results from tbe extensive coverageof protests surrounding U.S. participation inthe 1991 Gulf War, and it follows from sev-eral interacting features of those protestsand tbeir coverage. First , the Gulf Warprompted a burst of demonstrations in thefirst several m onth s of 199 1. The Janu arytbrough March period in 1991 witnessed32.4 percent of the year's permitted demon-strations in Washington. Nearly one-half oftbose more than 600 demonstrations werefor or against the Gulf War. But in 1982,only 12.7 percent of the total annual pro-tests occurred in tbe same season, reflectingtbe permitt ing agencies ' common wisdomthat demonstrations are more likely in tbelate spring, summer, and early fall tban dur-ing tbe winter months. Second, few of tbeGulf War demonstrations were large ones,many having been quickly p lanned andfielded. Third, even though tbe annual vol-ume of demonstrat ions increased between

  • 7/29/2019 Images of Protest John Mccarthy

    12/23

    SELECTION BIAS IN MEDIA COVERAGE OF PROTEST 489

    1982Total Permitted = 1,209Reported = 13%

    1991Total Permitted = 1,656Reported = 7%

    a

    Demonstration

    Figure 1. Percent of Demonstrations Permitted and Reported in the Media, by Size of Demonstration:Washington, D.C, 1982 and 1991

    many smaller demonstrations that may havebeen missed in different circumstances.Table 4 also shows the level of coverage ofdemonstrations by media source. Consistentwith the percentage of demonstrators cov-ered. The Washington Post shows more com-prehensive coverage than the New York Timesfor both years, both print sources show re-duced levels of coverage between 1982 and1991, while the national electronic mediashow a consistent level of coverage.

    D E M O N S T R A T I O N I M A G E S A N DPA TTE RN S IN 1982 AND 1991The typical character of demonstrat ions ,their sheer volume, and their representationsin the media changed in several ways be-tween 1982 and 1991. First, there was an in-

    tion of the total annual poo! of Washingtondemonstrators covered by the media declinedfrom 1982 to 1 991. The re is, based on thisevidence, notable instability in the characterof Washington, D.C. demonstrat ions andtheir media representations during these twotime periods. Is this instability reflected inthe structure of media selection bias?Analyses of the Structure ofMedia Selection BiasRecall what we mean by the structure of se-lection bias: the characteristics of protest as-sociated with the likelihood of mass mediacoverage. We see no reason to expect any re-lationship between changes in protest pat-terns and the structure of selection bias inmedia coverage of that protest. Below we test

  • 7/29/2019 Images of Protest John Mccarthy

    13/23

    490 AMERICAN SOCIOLOGICAL REVIEWTable 4. Percentages of Demonstrations and of Demonstration Participants Reported in MediaSourees During 1982 and 1991

    Percent Reported in 1982 Percent Reported in 1991Media SourceThe New York TimesThe Washington PostABC, CBS, NBCAny media source*^

    PermittedDemonstrations"4.17.92.1

    13.1

    DemonstrationParticipants''67.074.762.080.1

    PermittedDemonstrations"1.85.82.17.2

    DemonstrationParticipants''20.329.220.029,8

    'Figures represent the percentage of all demonstrations for the year reported by each source.^ Figures represent the percentage of the aggregate number of demonstrators in each year that were par

    of demonstrations reported in each media source." Any media represents the aggregate coverage across all three media sources.

    All independent variables, with the exceptionof demonst ra t ion s ize , are d ichotomous.Demonstration size is coded into the six cat-egories shown in Figure 1. Table 5 displa ys,in the first and third columns, the bivariateest imates for each independent var iable ,showing its unmediated relationship to cov-erage by any of the media sources for 1982and 1991. To the right of these colum ns arethe full models whicb control for all othereffects. Both the bivariate and the multivari-ate estimates are derived by fitting logisticregression models. The logged parametersassociated witb tbese models and tbe moreinterpretable odds ratios derived from themare presented in Table 5.

    The bivariate estimates are consistent witbmany of the expectations we derived fromtbeoretical interpretation of how media insti-tutions select a few demonstrations for cov-erage from vast numbers tbat take place. Tbecoefficients for demonstration size are posi-tive and significant for both years, reflectingtbe patterns we have already discussed. Thedemonstrations tbat made up ongoing cam-paigns of protest are less likely to be coveredin both years. More dramatic protest forms,like marches/rallies, attract more coverage;

    otbers to be covered in each yeartbe Is-raeli/Lebanese War in 1982 and tbe Gulf Warin 1991. And , in 1991 , protests against tbepolicies of foreign governments, in spite oftbeir bigb frequency, were less likely to gainattention. Tbe pattern of coverage suggestsless attention in tbe media to tbose protestissues tbat cballenge important elite inter-ests , sucb as Hai t ian democracy, labor ,bomelessness, and tbe environment, althoughtbese coefficients are not generally signifi-cant (witb tbe exception of tbe protestsagainst U.S. policy in Latin America in 1982and those against bomelessness in 1991). Ineacb year, demonstrations pertaining to atleast one domestic issue yield significant pa-rameters. The associated odds ratios sbownimply that in 1982 civil rights demonstra-tions were more tban tbree times as likely tobe covered . and dem ons trations on bebalf ofveterans were more tban twice as likely to becovered, while demonstrations on economicissues were almost five times as likely to becovered in 1991.

    Tbe mult ivariate est imates for the ful lmodel in Table 5 present a clearer pattern ofresults. Witb a few minor exceptions, themost consequential, statistically significant

  • 7/29/2019 Images of Protest John Mccarthy

    14/23

    S E L E C T I O N B I A S IN M E D I A C O V E R A G E O F P R O T E S T 491Ta ble 5 . CoefTicients from th e Logistic Regress ion of M edia Selection Bias in Co vera ge in Any M ediaon Selected Independent

    Independent VariableContext

    Demonstration SizeCounter-demonstrationCampaignWeekend

    Form'March and rallyVigil/picketRallyLiterature table

    Purpose (Top W tssue.^}Foreign governmentpolicyLalin America/peaceMiddle East-Lebanon WarSchool prayerAnti-nuclear weaponsER AVeterans ' issuesLaborJobs/economyCivil rights

    Constant-2 log likelihoodChi-square

    1BivariateEstimates

    L 1 9 5 ' "(3,305)1.040'"(2.830)

    - l , 7 0 9 " '( 1 8 1 ),M 0( I . I 1 6 )

    1.686"'(5.399)- 1 . 3 6 5 "(.255). 9 7 0 " '(2.638)

    - 1 . 3 4 9 "(.259)

    ,162(1,176)- .997(.369)1,078'"(2,939)

    - 2 , 6 3 6 "(,072), 8 3 5 "(2,304).176

    (1.192). 8 6 8 '(2.381).253(1.288),293(1,341)1,188"(3,280)

    Variab les : Wash ington , D.C. Demonstra t ions , 1982 and 1991982

    FullModel

    1.150'"(3.158).647(1.910)

    - .683*(.505)- , 6 3 6 "(.529)

    - .973*(.378)

    - 1 . 3 5 9 " *(.257)

    -1.020*(.360)

    .326(1.385),644(1,903)

    2 , 8 6 6 " '(17,568)- . 3 4 3( 7 0 9 )

    .961*(2,614)L 2 9 1 '

    (3,637)- , 1 4 5(.865),287(1,332)

    - .117(,889).672(1.959)

    -3.324***577,413

    2 6 6 . 1 5 9 ' "

    PurposeForeign governmentpolicyGulf WarVeterans ' issuesHomelessnessPro-environmentWomen's issuesDemocracy for HaitiLaborHealth careJobs/economy

    1991BivariateEstimates

    . 7 4 7 "*(2,111)- , 0 8 0(.923)- .705** '(.494)

    ,139(1,149)

    1 ,461"*(4,312),267(1.113)

    - . 0 5 0(.951)- 2 , 6 1 1 * "(,078)

    - 1 . 2 9 5 * "(.274)2 . 8 0 0 * "(16.413)

    -6 .722(.001)

    -1 .237*(.290)

    - 5 . 7 0 0(.003)- , 2 9 6(,743)- , 4 2 8(.652)

    - 1 , 2 4 0(,289).198(1.219)

    L599*(4,946)

    FullModel

    ,824**'(2.279),240(1,272)

    - , 0 2 3( 9 7 7 )- ,545*(.579)

    1.205(3.335),228(1.256)

    - . 0 3 8(.962)

    ,336(.714)2 , 9 9 0 "*(19,885)

    - 5 . 6 7 3( 0 0 3 )- , 4 8 3(.617)

    -6 ,067( 0 0 2 )1,341

    (3,825),125(1.132)- . 1 5 3(,858)2 , 1 8 1 "(8,854)1,769*( 5 8 6 7 )

    - 5 , 5 7 2 * "533.487

    2 8 6 . 3 1 7 "*

  • 7/29/2019 Images of Protest John Mccarthy

    15/23

    492 AMERICAN SOCIOLOGICAL REVIEW(10,001 to 100,000) are (3.158)4 or 99 timesmore likely to be covered than those in thesmallest size category (1 to 25) for 1982 and(2.279)4 or 27 times more likely to be cov-ered in 1991.The next largest effect in both years resultsfrom a demonst ra t ion 's purpose . In 1982.Middle East demonstrat ions were over 17times more likely to be covered than otherdemons t r a t ions , and ERA demonst ra t ionswere 3 times more likely to be covered. In1991, demonst ra t ions about the Gulf Warwere over 19 times more likely to be cov-ered, those pertaining to health issues almost9 times more likely to be covered, and thoserelated to the economy almost 6 times morelikely to be covered. These are large effects,but they are dwarfed by the consequences ofsize on media coverage.

    Other effects of demonstration context inthese analyses include those of weekend in1982 and 1991. In both years demonstrationsthat occur on weekends are somewhat larger;however, when the effects of size are takeninto account, weekend demonstrat ions areabout one-half as likely to be covered asother demonstrat ions. Based on the "newsroutine" perspective we had anticipated, in-correctly, that the larger size of the weekendnewshole would favor coverage of weekenddemonstrat ions. The fact remains that week-end demonstrat ions that are small and ad-dress Issues low on the media agenda are ap-parently neglected, and this may be a func-tion of the routinely smaller weekend staff ofmedia organizations.With the except ion of the vigi l /p icketform in 1982, there remain no statisticallysignificant effects of demonstration form onmedia coverage. Vigils and pickets are lesslikely than are literature tables to be coveredin 1982, but the odds ratios imply that theeffect is not substantively signif icant . In1991, the coefficient for the march and rallyform approaches statistical significance (p =.06) and shows that this form is more than

    Tahle 6, Coefllcients from the Logistic Regression of Media Selection Bias (Coveragein Any Media) on selected IndependenVariabies: Washington, D.C. Demonstrations, Full Model for Each Source1982

    Dependent /VariableContext

    DemonstrationsizeCounter-

    demonstratioiCampaignWeekend

    FormMarch and rallyVigil/picketRallyLiterature table

    TheVew YorkTimes

    1.037"(2.820)-.052

    1 (,949).180(1,197)-560(.571)

    -.794(-452)-.399(,671)-,4I1(.663)

    ~Purpose (Top 10 Issues)

    Foreign govern-ment policy

    Latin America/peaceMiddle East-

    Lebanon WarSchool prayerAnti-nuclear

    weaponsER AVeterans'

    issuesL^borJobs/economyCivil rights

    Constant

    -,251(.778)-1,042(.353)

    2.088"'(8,071)-5,912(.003)

    -,041(.960).717(2,048)

    -,453(.636)-7,210(001)-,074(929)

    -1.232(3,429)-5.076**"

    Media SourcesTheWashingtonPost

    , 941"*(2,564),870'

    (2,386)-,101(.904)-.516(.597)

    -772(.462)

    -1.083*(.338)-.434(.648)

    ,202(1.224)-,383(,682)1,678*"

    (5,353)-311(.732),601(1.825),803(2.231)

    -1,525(.218).422(1,526)

    -869(.419)1,037(2.821)

    ' -3,847"'

    NationalTelevisionNews

    1.076'"(2.933)1.275

    (3,577)-1,147(.317)-.124(.883)

    -294(.745),574

    (1,775),789

    (2.202)

    -7.778(.000).493(1,637)

    -,519(595)-6,998

    (,00I),687(1.988)-.327(.721)

    ,203(1,226)-8.566(,000)

    .388(1,474)

    .578(1.783)

  • 7/29/2019 Images of Protest John Mccarthy

    16/23

    SELECTION BIAS IN MEDIA COVERAGE OF PROTEST 493Tab le 7. Coefficients from the Log istic Regres-sion of Media Selection Bias (Coveragein Any Media) on selected IndependentVariables: Washington, D.C. Demon-strations, Full Model for Each Source,1991

    Dependent JVariableCon lex r

    DemonstrationsizeCounter-

    The\ew YorkTimes

    1.096"(2.993)- 7 . 4 6 5

    demonstration (.001)CampaignWeekend

    FormMarch and rallyVigil/picketRallyLiterature table

    .853(2.347)- 1 . 7 6 2 "(.172)

    .208(1.232)- .644(.525)- .207(.813)

    Purpose (Tap 10 Issues)Foreign govern-ment policyGulf War

    -5 .251(.005)4 . 7 5 9 "(116.651)

    Veleran.s"issuesHomelessnesKPro-environmentW o m e n ' sissuesDemocracyfor HaitiLaborHealth careJobs/economy

    Constant

    -4 .889(.007)

    - 6 . 0 0 8(.002)

    - 5 . 3 1 3(.005)2.957(19.248)

    -6 .143(.002)3.317*(27.589)

    -4 .719(.009)

    - 6 . 6 0 5(.001)

    -8.930**

    Media SourcesTheWashingtonPost

    . 835"*(2.304)

    .672(1.958)- .056( 9 4 6 )- .524(.592)

    2.339*(10.367).760(2.138).527

    (1.694)

    - .616(.540)3 . 3 5 4 " '(28.625)

    -5 .722(.003).272(1.312)

    -6 .301(.002)

    -6 .221(.002).551(1.735).158(1.171)

    2.099*(8.162)1.290

    (3.631) - 6 . 8 5 8 " "

    NationalTelevisionNews

    .381(1.463)-7 .332(.001)- .334(.716).179(1.196)

    .528(1.696)

    .399(1.490)- .373(.689)

    .229(1.2.58)1.683*"(5.382)

    -6 .381(.002)

    -6 .870(.001)

    -6 .501(.001)1.810*(6.108)

    -7 .089(.001).885(2.423)

    1.432(4.187)1.796

    (6.025)- 5 . 1 8 3 * "

    more likely to be reported in both 1982 and1991. And demonstrat ions associated withsome prominent domestic policy issues, suchas the Equal Rights Amendment in 1982 andhealth care in 1991 , were also more likely tobe covered. This fmding suggests that selec-tion by the media of demonstrations for cov-erage may be influenced by some externalassessment of what issues merit media atten-tion. It suggests also that demonstrations onissues central to the current media agendawill more likely be covered than those thattry to introduce neglected issues onto themedia agenda. We further explore this inter-pretation and its implications in our discus-sion.Media Source ComparisonsTable 6 presents coefficients for each of thethree media sources for 1982. Analyses areequivalent to the full model presented inTable 5. Table 7 presents the same for 1991.What is striking about the coefficients inthese two tables is how similar each is to themodels' coefficients based on coverage ag-gregated across media sources shown inTable 5; and, also, how similar each is to theother (with a few minor exceptions). In 1982,demonstration size shows constant, statisti-cally significant effects across media sources(note the relative similarity of the odds ra-tios for size across media sources), but thereare no statistically significant effects on net-work TV coverage of any of the protest pur-poses.In 1991 , the most substantively impo rtantresult is the lack of a statistically significanteffect of size on the likelihood of coveragefor network TVin that year, a Gulf Wardemonstration purpose is the only significanteffect. The size of the effect for network TVis smaller than for the other media sourcesGulf War demonstrations are only about 5times more likely to be covered, comparedto more than 25 times and 116 times morelikely in The "Washington Post and The New

  • 7/29/2019 Images of Protest John Mccarthy

    17/23

    494 AMERICAN SOCIOLOGICAL REVIEWSUMM ARY AN D INTERPRETATIONNearly three decades have passed sinceMichael Lipsky's (1968) insightful discus-sion of the relationship between the massmedia and political protest.

    To the extent that successful protest activitydepends upon appealing to. and/or threatening,other groups in the community, the communi-cations media set the limits of protest action. Ifprotest tactics are not considered significant bythe media, or if newspapers and television re-porters or editors decide to overlook protesttactics, protest organizations will not succeed.Like the tree falling unheard in the forest, thereis no protest unless protest is perceived andprojected, (P, 1151)Our comparison of off icial records ofW ashington dem onstra t ions for 1982 and1991 shows significant variation in how theywere covered by the media for these twoyears, as well as in several of their character-istics. Their typical forms, temporal patterns,and purposes changed; their numbers in-

    creased. Some of these changes may repre-sent trends, and some may be episodic.^'These d i f ferences notwi thstanding, thestructure of media selection bias appears tobe re la t ively s table between 1982 and199 J 22 -pj^g yggj majority of demonstrationsare ignored by the mainstream media; thevery large ones are covered. Demonstrationsize is, by far, the most important character-istic determining the likelihood of mediacoverage.After the effects of size are taken into ac-count, the next most important correlate ofcoverage is being in the right place at theright time in a media attention cycle. While

    ^' Using available archives, we compared theNPS permit records from 1973 with those of 1982and 1991 in an effort to disentangle trends frominstability in the 1982 versus 1991 comparison.This truncated comparison showed substantialannual volatility in the volume of demonstrations

    social movements may, at times, be centralto the shape of a media attention cycle, dem-onstrators are probably more often at themercy of large cycles upon whose trajecto-ries they can have little impact. The peaks ofissue cycles, then, can be thought to providewindows of opportunity for demonstratorswho are unwilling or unable to mount largeprotests. When the volume of coverage for anissue is large, pertinent protests, even smallones, are more likely to attract coverage. Sohere, the purposes of newsmakers and pro-testers mesh.^^

    We explored this interpretation by using aNexis search to cons truct an inde pend entmeasure of media coverage of the issuesaround which protests occurred.'^'* We com-bined the total number of stories from TheWashington Post and The New York Times byissue for 8 of the top 10 dem onstration is-sues shown in Table 5 for 1982 for eachthree-month period (quarter) in each year;we did the same for stories in TimeNewsweek, and U.S. News and World Re-port.-^^ The correlations for 1982 betweenthese two independent, aggregated measuresof issue coverage and the likelihood that ademonstration on that issue was covered'^^ by

    " A s several reviewers and colleagues pointedout to us, the lack of such an independent mea-sure makes our interpretation highly speculative,however plausible,-''The Nexis search included all The New YorTimes, The Washington Post, Newsweek, U.S.News and World Report, and Time Magazineporis in 1982 and 1991 and the quarters preced-ing and following those years. The ranking ofdemonstration issues did not vary substantiallyacross newspapers and newsweeklies, respec-tively, so the sources were pooled for thesearches. The issues "economy" and "foreigngovernment" proved to be too broad to yield us-able comparative results, so they were excludedfrom the analyses.^ The Spearman rho correlation between thecombined newspaper and newsmagazine storycounts by issue was .69 for 1982 and ,89 for

  • 7/29/2019 Images of Protest John Mccarthy

    18/23

    SELECTION BIAS IN MEDIA COVERAGE O F PRO TEST 495any media source appear in Table 8. Tbequarter before coefficients represent tbe re-lationsbip between tbe number of stories intbe tbree-montb period preceding tbe quarterin wbicb demonstrations occurred; the cur-rent quar ter represents tbat re la t ionsbipwithin tbe same tbree-montb period in wbichtbe detnonstrations occurred; and tbe quarterafter represents tbe three-montb period fol-lowing tbe quarter in wbicb tbe demonstra-tions occurred. For botb aggregated sourcesof media coverage, the number of stories intbe quarter preceding tbe quarter in whichprotests occur is significantly correlated witbtbe rate of coverage. And, so is tbe numberof current quarter stories for tbe newspapers.In otber words, tbe media's attention is al-ready focused on an issue before demonstra-t ions on tbese issues themselves becomemore likely to be tbe subject of media atten-tion.^^ Tbis evidence strongly supports ourinterpretation tbat issue-attention cycles af-fect tbe likelibood of coverage of demonstra-tions.

    Notice that we have not presented {al-thougb we completed) similar analyses for1991. Tbis is because tbe vast majority (76percent) of demonstrations covered tbat year(amon g tbe top 10 dem onstration issues) per-tained to tbe Gulf War.^^ Tbirty-three percentof tbe Gulf War demonstrations, tbe mostcommon demonstration issue in 1991 (mostdemonstrations occurred in tbe first quarterof tbe year), were covered by some mediasource, while only 2.5 percent of demonstra-tions on any of tbe otber 10 issues were cov-ered. Tbis pattern even more strongly sup-ports our supposition tbat media coverage ofdemonstrations is a consequence media-at-tention cycles.Tbese patterns indicate tbat media agendasetting processes are a key to understandingbow public dissent is selected for reporting.used because the distributions were severelyskewed in each year.^ This does not necessarily demonstrate that

    Table 8. Spe arm an Rho Coefficients Showingthe Correlation between Newspaperand Newsmagazine Coverage of Dem-onstration Issues and the Likelihoodthat Demonstrations on those IssuesWere Covered in the Media: Washing-tun, D .C , by Calendar Qu arter, 1982Media Source

    The Washington Time/Calendar Post/The NewsweeklQuarter, 1982 New York Times U.S. NewsQuarter before .44Current quarter .51(p = .OQuarter after .31(p = .1Number of quarters 28

    .49

    .18(p = .36)

    .21(p = .28)28

    Few analysts of agenda setting processes intbe media attribute mucb general influence toprotest movements. Ratber, officials, botbelected and ap pointed, well-organ ized inter-est groups, and tbe internal workings of me-dia organizations tbemselves are seen asmore important in accounting for normal pat-terns of media issue attention (Baumgartnerand Jones 1993). The likelibood that a pro-test will be reported by tbe mass media issbaped by forces mostly beyond tbe controlof most protest groups, unless they are ca-pable of generat ing mass part icipation indemonstrations.Tbe factors of demonstration size and me-dia issue cycle in concert, tben, provide areasonably good account of tbe structure ofmedia selection of Washington, D.C. demon-strations in tbese data we bave examined,providing a ratber straightforward answer totbe question of the structure of selection bias.Nevertbeless, tbe sbarp escalation of newscoverage surrounding tbe Gulf War in tbefirst montbs of 1991 was apparently capableof muting what appear to be tbe typically

  • 7/29/2019 Images of Protest John Mccarthy

    19/23

    496 AMERICAN SOCIOLOGICAL REVIEW52 times more prevalent in Newsweek, TimeMagazine, and U.S. News and World Reportthan was coverage of the least reported pro-test issue (pro-environment) in the samequarter. In contrast, during 1982, the topdemonstration issue (war in Lebanon) wasonly 8 and 5 times more likely to be coveredin those sources, respectively, than was theleast reported issue (school prayer) in thesame quarter.

    Important questions remain concerning thegeneral izabi l i ty of the pat terns we haveshown . As we noted ab ove, Washington is thenational capital, and probably a majority ofthose who participate in large demonstrationsthere come from elsewhere in the UnitedStates. As a result, media organizations mayview dem on stratio ns in W ashington differ-ently than those that occur elsewhere.

    Whether or not the selection processes de-fming local media organizations' coverage oflocal demonstrations mirror national mediapatterns remains to be seen. For 1982, wewere able to develop an 1982 estimate ofcoverage for one local television network af-filiate in Washington (results not shown),marginally apropos to this problem. That lo-cal electronic coverage was more extensivethan any of the other media sources by a no-ticeable margin, but the structure of selectionmirrored the patterns we have already de-scribeddemonstrat ion size remained themost important factor in accounting for thelikelihood of coverage. The only noticeabledifferences on the likelihood of local cover-age were the important positive effect ofdemonstrat ions targeting economic issuesand the significant negative effect of demon-strations concerning the policies of foreigngovernments. Neither effect is evident incoverage by the nat ional media sources .These results suggest that national and localissue attention cycles may diverge from oneanother, differentially affecting national andlocal coverage of demonstrations.Demonstrations in Washington in 1982 and

    A few unruly and unpermitted demonstra-tions were covered by the mediatypicallysmall sit-ins resulting in confrontations andarrests. As a result, we are unable to ad-equately assess the independent effects ofdisruptive versus routine protest demonstra-tions. Extrapolating from White's (1993) evi-dence showing higher rates of media report-ing of Nor thern I re land conf l ic ts whendeaths were involved, we would be surprisedif demonstration unruliness was not an im-portant factor influencing the media selec-tion from a pop ulation of de m on stratio nswith any variation on this characteristic (cfSnyder and Kelly 1977).What have we learned by comparing me-dia attention across several media sources?First, there is a strong correspondence be-tween the aggregate results we have de-scribed and the separate analyses for TheWashington Post and for The New YorkTimes, which is the usual print media sourcefor past studies on collective action events inthe United States. This suggests that thestructure of bias is rather stable across thesetwo sources, even given the wide variationbetween them in the extent of coverage. Se-lection bias is similarly stable across time, inspite of high levels of volatility in the char-acteristics of populations of demonstrationsand in the patterns of media coverage. Inshort, judging from our comparison, the na-tional print media provides an amazinglystable portrait of the churning mixture ofprotest forms, purposes, and contexts inWashington, D.C, during 1982 and 1991.The same cannot be said for the aggregatednational television network newscasts. Thedisappearance of size as a statistically sig-nificant factor in the likelihood of televisionnetwork coverage in 1991 by the televisionnetworks suggests somewhat less stabilityover time in the structure of bias for televi-sion coverage of Washington demonstrations.The intense focus on Gulf War issues and theattendant protests markedly altered the struc-

  • 7/29/2019 Images of Protest John Mccarthy

    20/23

    SELECTION BIAS INMEDIA COVERAGE OF PROTEST 497the extent that this contrast between print andelectronic media is a trend, rather tban anaberration, there are enormous implicationsfor how protest is mediated in the UnitedStates. To the extent that media issue cyclesbecome the dominant factor in accountingfor media coverage of demonstrations in theelectronic media, this increasingly "legiti-mate" channel by which citizens may regis-ter their policy preferences is undermined. Ifmedia issue attention cycles come to play amore significant role than do the form, con-text, substance, or size of citizen protests indetermining which demonstrations are se-lected for media coverage, then protest inmodern democracies will have become me-diated to a greater extent than even MichaelLipsky's (1968) prescient observation pre-supposed.John D. McCarthy is Ordinary Professor of So-ciology and a Member of the Life Cycle Instituteat The Catholic University of America in Wash-ington. D.C. He continues his research on pro-test events, the policing of protest and the role ofsocial movement organizations in the mobiliza-tion of citizen action. He spent the 1995-1996academic year as a Senior Fulbright ResearchScholar at the Wissenschaftszentrum in Berlin.He is co-author (with Jim Castelli) o / P o w e r Or-ganizing (Henry Holt, forthcoming) and is co-edi-tor (with Doug McAdam and Mayer N. Zald) ofComparat ive Perspect ives on Social Movements(Cambridge University Press. 1996).Clark McPhail is Professor Sociology at the Uni-versity of Illinois at Urbana-Champaign. Hisprincipal interests are purposive individual andcollective action. He is completing a monograph.Act ing Together : The Social Organizat ion ofCrowds. The current "description bias " phase ofhis collaboration with John McC arthy comparessystematic observations of collective action inWashington demonstrations with the descriptionsin the electronic and print media. For The Mythof the Madding Crowd (Aldine De Gruyter. 1991)he received the 1994 distinguished scholarshipaward from the ASA sec tion on Collective Behav-ior and Social Movements.

    is co-editor (with Ron Pagnucco and CharlesChatfield) of Sol idar i ty beyond ihe Sta t e :Transnational Social Movements and World Poli-tics (Syracuse University Press, forthcoming).

    REFERENCESAn, Mildred. 1991. "Free Speech, Post OfficeSidewalks and the Public Forum Doctrine."Harvard Civil Rights-Civil Liberties Review26:633-48 .Barnes. Samuel H. and Max Kasse, et al. 1979.Political Action: Mass Participation in FiveNations. Beverley Hills. CA: Sage.Baumgartner. Frank R. and Bryan D. Jones. 1993.

    Agendas and Instability in American Politics.Chicago. IL: University of Chicago Press.Berk, Richard A. 1983. "An In t roduc t ion toSample Selection Bias in Sociological Data. "American Sociological Review 48 : 386 - 9 8 .Dalton, Russell J. 1988. Citizen Politics in West-ern Democracies. C h a t h a m , NJ: ChathamHouse.Danzger, Maurice H. 1975. "Validating ConflictData." American Sociological Review 40 : 5 7 0 -84.Downs . An thony . 1972. "Up and Down withEcologyThe Issue At tent ion Cycle ." ThePublic Interest 28:38-50 .Everett, Kevin. 1992. "Professional izat ion andProtest: Changes in the Social Movement Sec-tor , 1961-1983." Social Forces 70:951-76.Etzioni. Amitai. 1970. Demonstrating DemocracyNew York: Gordon and Breach.Feierabend, Ivo K. and Rosaland L. Feierabend1966. "Aggressive Behavior within Poli t ies.1 9 48 - 1 9 62 . " Journal of Conflict Resolution10:249-71.Franzosi, Roberto. 1987. "The Press as a Sourceof Socio-Historic Data: Issues in the Method-ology of Data Collection from Newspapers."Historical Methods (Winter) 20:5-15.

    . 1990a. "Strategies for the Prevention. De-tection and Correction of Measurement Errorin Data Collected from Textual Sources." So-ciological Methods and Research 18:442-72.- . 1990b. "Computer-Assisted Coding ofTextual Data" Sociological Methods and Re-search 19:225-57.Gamson, Wil l iam A., David Croteau, Wil l iamHoynes and Theodore Sasson. 1992. "Media

  • 7/29/2019 Images of Protest John Mccarthy

    21/23

    498 AMERICAN SOCIOLOGICAL REVIEW

    News, Newsweek and Time New York: Vin-tage,Gurr, Ted R. 1968, "A Ca usal M ode! of CivilStrife: A Coniparalive Atiaiysis Using New In-d i c e s . " American Political Science Review62 :1104-24 .Herbst , Susan . 1993. Numbered Voices: HowOpinion Polling Has Shaped American Poli-tics. Chicago. IL: University ot Chicago Press,Her tnan . Edward and Noam Chomsky, 1988.Manufacturing Consent. New York: Pantheon,Hierich, Max. 1971. The Spiral of Conflict: Ber-keley, 1964. New York: Columbia UniversityPress.Hilgartner, Stephen and Charles L, Bosk. 1988,"The Rise and Fall of Social Problems: A Pub-lic Arenas Model," American Journal of Soci-ology 94(iuly):53-7S.Iyengar, Shanto and Donald R, Kinder, 1987,News That Matters: Television and AmericanOpinion. Chicago, !L: University of ChicagoPress,Jenkins. J, Craig and Charles Perrow, 1977, "In-surgency of the Power less : Farm WorkerMovemen ts (1946-1972) . " American Socio-logical Review 42 :249-68 .Jenkins, J. Craig and Kurt Schock. 1992. "Global

    Structures and Political Processes in the Studyof Dom estic Po litical Conflict," Annual Reviewof Sociology 18:161-85.Kie lbow itz. Richard B, and Clifford Sch erer.1986. "The Role of The Press in the Dynamicso f Soc i a l Movemen ts , " Research In SocialMovements, Conflict and Change 9:71-96.Klan derm ans. Bert and Dirk Omeg a, 1987. "Po -tentials. Networks. Motivations and Barriers:Steps towards Participation in Social Move-m e n t s . " American Sociological Review 52:5 1 9 -3 1 ,Lipsk y. Michae l. 1968. "Protest as a Political Re-source . " American Political Science Review62:1144-58,Lofland. John and Michael Fink, 1982. SymbolicSit-ins: Protest at the California Cap itol.Washington, DC: University Press of America,M acC anne ll. Dean, 1973. "Nonviolent A ction asThe ater : A Dram aturgical A nalysis of 146Demonstrations." Monograph # 10, Haverford.PA: Haverford College Center for NonviolentConflict Resolution Monograph Series,Markoff, John, 1986, "Literacy and Revolt: Some

    Baker, and Elaine Mosakowski . 1988. "TheFounding of Social Movement Organizations:Local Citizens Groups Opposing Drunk Driv-ing. " Pp. 71-84 in Ecological Models of Or-ganizations, edited by G. R. Carrol l . Cam-bridge, MA: Ballinger.McCarthy. John D., Clark McPhail . and JackieSmith. 1994. "The Institutional Channeling ofProtest: The Emergence and Development ofU.S. Protest Management Systems," Presentedat International Sociological Association 23rdWorld Congress, July 14, Bielefeld, Germany,McCarthy. John D., Jackie Smith, and Mayer N.Zald. 1966. "Accessing Public, Media, Elec-toral and Governmental Agendas ." Pp . 291-311 in Com parative Perspectives o n SocialMovem ents: Political Opportunities, Mobiliz-ing Structures and Cultural Eramings. editedby D. McAdam, J. McCarthy and M. N. Zald.New York: Cambridge University Press.McManus, Kevin . 1993, "At the Wal l , SalesCome Under Fire." The Washington Post, June20, p . B l .McPhail, Clark. 1991. The Myth of the MaddingCrowd. New York: Walter De Gruyter.Molotch, Harvey and Marilyn L ester. 1975. "A c-cidental News: The Great Oil Spill as LocalOccurrence and National Event ," AmericanJournal of Sociology 81:235-60,Neuman. W. Russel l . 1990, "The Threshold ofPublic Attention." Public Opinion Quarterly54:159-76.Nolan, Joseph T. 1985. "Political Surfing WhenIssues Break ." Harvard Business Review63(January-February) :72-8] .Olzak, Susan. 1987, "Causes of Ethnic Protestand Conflict in Urban America, 1877-1889."Social Science Research 16:185-210.

    . 1989a. "Analysis of Events in the Studyof Collective Action," Annual Review of Soci-ology 1 5 : 1 1 9 ^ 1 .-. 1989b, "Labor Unrest. Immigration, andEthnic Conflict in Urban America, 1880-1915.American Journal of Sociology 94 :1303-33 .-. 1992. The Dynam ics of Ethnic Comp eti-tion and Conflict. Stanford, CA: Stanford Uni-versity Press.Paig e, Jeffrey. 1975. Agrarian Revolution. NewYork: Free Press.Rucht, Dieter and Thomas Ohiemaeher. 1992."Protest Event Data: Uses and Perspectives,"ed-

  • 7/29/2019 Images of Protest John Mccarthy

    22/23

    SELECTION BIAS IN MEDIA CO VERAGE OF PROTEST 499

    Ryan, Char lo t te . 1991. Prime Time Activism.Boston. MA: South End Press.Snyder, David and William R, Kelly. 1977. "Con-flict Intensity, Media Sensitivity and the Va-lidity of Newspaper Data," American Socio-logical Review 42 :105-23 ,Snyder, John V, 1985. "Forum over Substance:Cornelius v, NAACP Legal Defense and Edu-cation Fund," Catholic University Law Review35:333-70.Tanter. Raymond. 1966, "Dimensions of ConflictBehavior within and between Nations. 1958-1960," Journal of Conflict Resolution 10 :41 -64.Tarrow, Sidney. 1988. Democracy and Disorder:Politics and Protests in Italy, 1965-1975. NewYork: Oxford University Press.. 1989. Struggle. Politics and Reform: Col-lective Action, Social Movements, and Cyclesof Protest. Ithaca, NY: Cornell University Cen-

    ter for Institutional Studies., 1993. "Mo dular Collect ive Action andthe Rise of the Social Movement." Politics andSociety 2\{March).69-90.Tilly, Charles, 1978, From Mobilization to Revo-lution. Reading, MA: Addison Wesley.. 1983, "Speaking Your Mind wi thoutElect ions , Surveys , or Social Movements ,"Public Opinion Quarterly 47 :461-78 .. 1984, "Social Movements and NationalPolitics," Pp. 297-317 in Statemaking and So-cial Movements, edited by C, Bright and S.Harding, Ann Arbor, MI: University of Michi-gan Press.Tilly. Charles, Louise Tilly, and Richard Tilly.1975. The Rebellious Century: 1830-1975.Cambridge, MA: Harvard University Press.White, Robert W, 1993, "On Measuring PoliticalViolence: Northern Ireland, 1969 to 1980."American Sociological Review 58 :575-85 .

  • 7/29/2019 Images of Protest John Mccarthy

    23/23