NBER WORKING PAPER SERIES

Transcription

1 NBER WORKING PAPER SERIES INTEREST GROUPS, INFORMATION MANIPULATION IN THE MEDIA, AND PUBLIC POLICY: THE CASE OF THE LANDLESS PEASANTS MOVEMENT IN BRAZIL Lee J. Alston Gary D. Lbecap Bernardo Mueller Workng Paper NATIONAL BUREAU OF ECONOMIC RESEARCH 1050 Massachusetts Avenue Cambrdge, MA Aprl 2010 We thank Sarah Anderson of UCSB for detaled comments, as well as A. Mushfq Mobarak, Mchael Oppenhemer, Anna Rubnchk-Pessach, and partcpants at the followng workshops and meetngs: XXIII Encontro Braslero de Econometra Salvador; Unversdade Católca de Brasíla; Department of Economcs of the Unversty of Brasíla; Latn Amercan Meetngs of the Econometrc Socety ; Insttute for Behavoral Scence Unversty of Colorado; Unversty of Colorado, Envronmental and Resource Economcs Workshop. We acknowledge the followng nsttutons for support: Alston, the STEP Program at the Woodrow Wlson School at Prnceton; Alston and Mueller, NSF Grant # ; Mueller, CNPQ; and Lbecap, the Internatonal Center for Economc Research (ICER), Turn, Italy. The vews expressed heren are those of the authors and do not necessarly reflect the vews of the Natonal Bureau of Economc Research. NBER workng papers are crculated for dscusson and comment purposes. They have not been peerrevewed or been subject to the revew by the NBER Board of Drectors that accompanes offcal NBER publcatons by Lee J. Alston, Gary D. Lbecap, and Bernardo Mueller. All rghts reserved. Short sectons of text, not to exceed two paragraphs, may be quoted wthout explct permsson provded that full credt, ncludng notce, s gven to the source.

2 Interest Groups, Informaton Manpulaton n the Meda, and Publc Polcy: The Case of the Landless Peasants Movement n Brazl Lee J. Alston, Gary D. Lbecap, and Bernardo Mueller NBER Workng Paper No Aprl 2010 JEL No. D23,D72,D78,O13,Q15 ABSTRACT We extend the lterature on nterest group behavor and polcy outcomes by examnng how groups wth lmted resources (votes and campagn contrbutons) effectvely nfluence government by manpulatng meda nformaton to voters. Voters n turn lobby poltcans to mplement the group s preferred polces. In ths manner nterest groups can secure favorable government actons beyond ther sze and wealth. Ths s an mportant contrbuton because of the ncreased role of the meda n the nformaton age and because ths lnkage better explans observed government polces. We develop a mult-prncpal, mult-task model of nterest group behavor and generate the characterstcs of nterest groups that would be most successful usng publcty to secure ther polcy objectves. We apply the model to the Landless Peasants Movement n Brazl. We detal how the Landless Peasants Movement molds nformaton; show the general voter response; and examne the reacton of poltcans n changng the tmng and nature of polcy. Lee J. Alston Program on Envronment and Socety Insttute of Behavoral Scence Department of Economcs Unversty of Colorado at Boulder Boulder, CO and NBER Lee.Alston@colorado.edu Bernardo Mueller Departamento de Economa Unversdade de Brasíla Campus Darcy Rbero, ICC Norte Brasíla-D.F Brasl bmueller@unb.br Gary D. Lbecap Bren School of Envronmental Scence and Management and Economcs Department Unversty of Calforna, Santa Barbara Bren Hall 4412 Santa Barbara, CA and ICER and also NBER glbecap@bren.ucsb.edu

3 I. Introducton. In ths paper we examne how an nterest group wth lmted votes and campagn funds nfluences poltcal polcy through the control of meda nformaton to voters. Voters respond to the meda by lobbyng poltcans to adopt the group s desred polces. In ths manner such groups can be more effectve than ther sze and wealth could suggest. Ths s a valuable nsght because of the ncreased role of the meda n affectng polcy and because ths lnkage better explans observed government polces. We buld on the nterest group lterature that generally argues that cohesve, artculate, wealthy, votng groups are most lkely to secure favorable government transfers (Peltzman, 1976; Ornsten and Elder, 1978; Becker, 1983; Rothenberg, 1992, 1-5; Grossman and Helpman, 2001; 2002, 2-3, 44). We also draw from the related lterature on nterest groups as transmtters of nformaton (Austen-Smth, 1999; Grossman and Helpman, 2001, 2002, 2-3; Lohmann, 1998 and Van Wnden, 1999 and 2003 for surveys). In ths work, nformaton flows from the group to poltcans to drectly nfluence ther actons and to moblze voter support through poltcal endorsements. 1 Baron (1994, 1989) ncorporates the meda nto ths process. Canddates chose among both partcularstc and collectve polces to maxmze contrbutons, and use the meda to generate favorable voter support. Baron (2005) examnes how competng nterests (ndustry and actvsts ) employ the meda to nfluence publc sentment on envronmental polces. Besley, Burgess and Prat (2002), Besley and Burgess (2001) and Stromberg (2004) model the role of the meda n makng government more responsve to ctzens. Rothenberg (1992, 11-43, 251) adds low labor costs n descrbng the role of organzed actvsts (Common Cause), who use volunteers, along wth campagn funds, and 1 See McKelvey and Ordeshook (1985), Lupa (1992). Accordng to Reuben (2002), Most of the nformaton lterature has concentrated on the nformaton transmsson between the nterest group and the polcymaker. It has neglected to look nto the nformaton flow among and wthn the nterest groups. Some exceptons are Grossman and Helpman (1999, 2001) and Cameron and Jung (1995). 2

4 meda clout to promote legslaton. Brewer and Lbecap (2008) also model actvst envronmental groups as havng below market wages n selectng to go to tral rather than settlng dsputes n conflcts over natural resources. We contnue ths lne of nvestgaton by defnng the characterstcs of nterest groups that successfully nfluence government polcy whle lackng both campagn contrbutons and votng power. We detal how groups utlze and show how they affect government offcals n moldng the tmng and nature of polcy. The group also captalzes on the nablty of voters to evaluate the polcy s effectveness. To llustrate our model we focus on the Landless Peasants Movement (Movmento Sem- Terra) or MST and ts efforts to nfluence domestc land reform polcy n Brazl. The MST represents a comparatvely small and poor sector of the Brazlan electorate, yet t has been remarkably effectve n promotng ts agenda. Although we drect attenton to the MST, our analyss can be generalzed to nterest group behavor n other contexts. 2 We examne Brazl because t has land dstrbuton problems smlar to those facng other Latn Amercan socetes and because ts nterest group poltcs are comparable to those found elsewhere. We analyze the poltcs of land reform n Brazl through a model developed n Secton II and detaled n the Appendx wth a untary government, the Presdent or executve. 3 The framework consders Presdental constrants n determnng resource allocaton. Although there are well-organzed, wealthy consttuents, large property owners, who oppose land reform, we show how the MST successfully counters by affectng voter perceptons through skllful use of the meda. Our model captures three mportant aspects of the relatonshps among nterest groups, the electorate, and the government: ) moral hazard from nformaton asymmetres between 2 Tetlock and Oppenhemer (2008) argue that envronmentalsts were able to create a taboo amongst voters concernng any drllng n the Arctc Natonal Wldlfe Refuge (ANWR). 3 We use a untary government because t s reflectve of the Brazlan settng, but t also s smlar to Parlamentary democraces (Alston, Melo, Mueller and Perera, 2006). 3

5 polcy makers and nterest groups; ) exstence of multple groups smultaneously pressurng the government; and ) ablty of some groups to dsrupt the flow of nformaton from others to the government, thereby ndrectly changng equlbrum polcy effort. From the model we derve the characterstcs that make an nterest group specalze n meda manpulaton rather than contrbutons or votes. Four mportant characterstcs emerge: 1) the nterest group has low labor costs; 2) the nterest group has a hgh productvty of effort to nfluence voters through the meda; 3) the ssue of the nterest group has hgh salence and concerns voters; and 4) polcy costs are low enough to make the polcy response feasble. Secton III provdes background nformaton. In Secton IV we argue that the MST matches the characterstcs from the model, and we provde evdence lnkng voter preferences to MST actons, meda responses, government reactons, and land reform. Although ths seems to suggest that the MST helps voters get more of a polcy they want, we show that ths may not necessarly be the case. II. A Multprncpal, Multtask Model of Interest Group Behavor. Our model extends Dxt (1996, pg ), whch s a combnaton of the multtask model of Holmström and Mlgrom (1991) and the multprncpal model of Bernhem and Whnston (1986). The major nnovaton s to make the sze of the nformaton asymmetry between groups and the government endogenous. We allow each group to dedcate resources to nformng or confusng other groups. We descrbe the model n the context of land reform n Brazl, leavng a formal presentaton of a general verson n the appendx. The agent n the model s a untary government (Executve) that mplements a set of polces, ncludng land reform. The three prncpals are the MST, landowners, and voters, each nterested n a specfc polcy (task). The MST desres expropraton of prvate land and creaton of settlement projects. Landowners want subsdes and oppose land reform. Voters 4

6 seek polces that delver more employment, ncome, and securty, whle also sympathzng wth land reform. The government s efforts can be substtutable or complementary across tasks, and the prncpals offer both campagn contrbutons and votes to nfluence the allocaton. The prncpals do not observe the government s effort for each task but rather have lmted nformaton on outcomes. These nformaton asymmetres allow the agent to skmp on effort, thereby earnng an nformatonal rent. To see the operaton of the model, consder a benchmark where the prncpals observe the government s effort and act cooperatvely. The optmal level of effort n ths scenaro would be a frst-best soluton where the margnal cost of effort n each task equals the margnal socal beneft. The second step s to relax the assumpton of observable effort, but retan cooperaton. In ths stuaton contracts between the prncpals and poltcans are contngent on outcomes. The optmal effort now devates from the frst-best. The sze of the wedge between these frst and second-best stuatons s a functon of the sze of the nformaton asymmetry as well as the rsk averson of the government. The thrd step s to ntroduce non-cooperatve behavor among the prncpals. Ths nvolves fndng the Nash equlbrum of the game where each prncpal provdes separate nducements to the agent and strategcally takes nto account the actons of others. Ths equlbrum s even further from the frst-best. Because competng nducements cancel, the greater the number of prncpals the lower the government s effort. Ths creates a thrd-best outcome, a drect consequence of the multprncpal multtask nature of the problem. The fnal step allows the three prncpals to nfluence polcy through nducements (contrbutons and votes) and by nfluencng the nformaton avalable to other partes, especally voters, about the government s efforts n each task. Ths s a novel contrbuton of our model. In ths context, each prncpal must decde how to allocate resources between 5

7 nducements and nformaton. In ths decson, the prncpal consders both that nformaton manpulaton s costly and that the other prncpals may engage n ths acton. The soluton s a Nash equlbrum, and t mples that the sze of resultng nformaton asymmetres between the agent and the prncpals s endogenous. The optmal levels of government effort contnue to be thrd-best, but the actons of the prncpals to mold nformaton may lead to changes n the supply and allocaton of effort across tasks. Consder what happens to optmal ncentves for poltcal acton wth a change n the level of nformaton avalable. Each prncpal wll adjust nducements for a partcular polcy when more nformaton becomes avalable regardng the government s effort. Those who beneft from the polcy wll provde addtonal nducements now that they have a better noton of what they are recevng from poltcans n exchange. Those who oppose wll provde fewer nducements as they see that the returns from opposton are lower than prevously perceved. These reactons may be reversed, dependng on the relatve values of the cost and beneft functons. Whatever the case, each nterest group can strategcally calculate how much and n whch drecton to affect nformaton so as to pursue ts polcy preferences. 6

8 Implcatons for the choce of nstrument for the nterest groups In order to llustrate the mplcatons of the model for land reform polcy n Brazl, consder the government s ncentve for choosng how much effort to place on land reform. Voters wll offer votes and contrbutons commensurate wth ther understandng of outcomes based on (thrd best) levels of government effort. The second prncpal, MST desres more land reform and would gan f voters ncreased ther nducements to poltcans. Accordngly, MST devotes effort to provdng more nformaton to voters about government actons. The new nformaton leads voters to push for more land reform. How much effort MST wll choose n ths regard s determned by the margnal gans and costs nvolved. Interest groups wll vary n ther profcences to nfluence government drectly through votes and contrbutons, or ndrectly through nformaton. The costs and benefts facng each group depend upon ts characterstcs and the polcy t s pursung. Our model suggests four key characterstcs that provde an advantage for an nterest group n the provson of nformaton: 4 ) The hgher the margnal cost of nfluencng nformaton (G e ) the lower wll be the j optmal level of such effort chosen by that group, ceters parbus. If the margnal cost s suffcently hgh, then the group wll not nfluence nformaton on that task. Some nterest groups may have relatve advantages through lower labor costs and therefore be better able to nfluence the nformaton flow. Note that low opportunty costs are not a suffcent condton for successful moblzaton as other complementary characterstcs are also necessary. The model shows that groups wth hgh costs of nfluencng nformaton prefer to pressure poltcans through votes or contrbutons. 5 4 The notaton refers to the formulas n the appendx. 5 The low labor cost may also arse due to the avalablty of workers that are motvated by non-fnancal objectves as n the not-for-proft lterature (see Francos and Vlassopoulos (2008) for a survey). Another mpedment to moblzaton, whch s partcularly relevant to the case of landless peasants, s the possblty of represson and volence from other groups or the government. One of the reasons for the success of the MST s 7

9 ( ( e, e jj j j e ) The hgher the productvty of effort to mold nformaton receved by voters j ) ), the more effort that wll be employed by the group, ceters parbus. Low productvty for some groups may be due to ther lack of credblty among voters. Accordngly nterest groups that work through the nformaton channel wll tend to have reputaton advantages and lower costs of beng notced. ) The ablty of an nterest group to affect polcy through nformaton depends on the preferences of all other prncpals (b), that s, the salence of the ssue. If voters care strongly about a gven polcy, ether favorably or unfavorably, then changes n the level of nformaton they receve can have mportant mpacts on the government s effort level for that polcy, ceters parbus. 6 v) Overall polcy costs and whether t s a complement or a substtute for those sought by the other prncpals (C) affect the success of an nterest group. A group whose task s a substtute (complement) to other tasks of other groups wll encounter negatve (postve) ncentves from the other groups. In the model the cost of each polcy s common knowledge, preventng an nterest group from tryng to change perceptons of how the task affects the costs of others. 7 III. Background: Land Reform n Brazl. 8 Brazl has long had one of the most concentrated land ownershp structures n the world. Approxmately 45% of the agrcultural land s held by the largest 1% of farm owners, and large tracts of ths land are not used. The Gn coeffcent of 0.85 n 1985 was the 9 th precsely ts ablty to turn ths volence to ts own advantage by publczng t to sympathetc voters who then press for more government acton (Ondett, 2006). In dong so the MST reduces the gans to landowners of engagng n volence and thereby reduces the level of volence. The Catholc Church also plays an mportant role n protectng the landless peasants from retalaton from landowners. See Alston and Mueller (2010) the role of the Catholc Church n land conflcts n Brazl. 6 We later argue and provde evdence that the publc cares about the landless. The same holds for the envronment, especally f the costs to voters appear to be low. 7 In prncple, however, the model could be extended to nclude ths addtonal nformaton asymmetry. 8 For greater detals on land reform n Brazl see Mueller et al. (1994). 8

10 hghest n the world (FAO/UN, 2004). 9 In terms of the sze of the populaton affected, t s certanly among the most problematc cases wth an estmated 4 mllon landless peasants. 10 Concentrated land holdngs grew out of the Brazlan colonal experence, and snce the 19 th century there have been repeated efforts by the central government to substtute small holdngs for latfunda (Dean, 197, 624). But lttle of consequence has happened. The Gn coeffcent has barely budged. In 1960 t was 0.84; n 1978, 0.85 and n 1998, Very large farms of over 1,000 hectares also have contnued to domnate land holdngs. In % of the farms held 48% of farmland; n % held 44%; and n % held 45%, and n % held 45%. 11 Despte ths lack of acton, the general electorate has long been sympathetc to the noton of land reform, a natural reacton gven such salent nequalty. 12 Ths broad consttuency for land reform, however, s unorganzed, heterogeneous and has only lmted nformaton regardng how much land reform s beng carred out. Rural property owners have steadfastly opposed t. 13 Hstorcally, they were well organzed and provded support to poltcans through contrbutons and votes. They were represented by several organzatons, such as the Brazlan Rural Socety (Socedade Braslera Rural) and especally by a large group of Congressmen from varous partes, known as the rural bench (bancada rural), that unted to promote the nterests of land owners and agrcultural producers. As a result untl 1993, the pattern was for poltcans to call for aggressve land reform durng electoral 9 The ten hghest Gn coeffcents n the FAO dataset ( are Barbados (0.94, 1989 data), Paraguay (0.93, 1991), Guam (0.88, 1987), Vrgn Islands (0.87, 1987), Panama (0.87, 1990), Bahamas (0.87, 1994), Peru (0.86, 1994), Span (0.86, 1989), Brazl (0.85, 1985), Argentna (0.83, 1988). 10 Ths s the estmate typcally gven by the Landless Peasants Movement. The Lula government s II Natonal Plan for Agraran Reform (Mnstéro do Desenvolvmento Agráro, 2004) puts the demand for land reform at 3.1 mllon, or 5 mllon f one counts those who have land but n an nsuffcent amount. The Food and Agrculture Organzaton estmated the demand for land reform n Brazl at 2.5 mllon famles n the md-nnetes (Romero et. al. 1994). 11 Gn coeffcents presented here are from the Natonal Insttute for Colonzaton and Agraran Reform, INCRA (2001), whch tres to deal wth several of the methodologcal dffcultes n the calculaton of ths ndex. 12 We provde evdence to back ths statement n Secton IV. 13 See Mueller (1998) for an ncdence analyss on the groups affected by land reform, and econometrc measurement of the mpact of nterest groups on land reform polcy. 9

11 campagns, and for lttle to be mplemented once the electon was over. So long as large landowners could delver more support than could landless peasants, and so long as voters had lttle nformaton on the actual state of land reform, poltcans devoted few resources to t. The pattern began to change when the MST (organzed n 1985) took advantage of the new Consttuton of 1988 that mandated the federal government to exproprate and redstrbute unproductve propertes. Complementary legslaton was passed n By the md 1990s, the MST had honed ts strategy of nvadng unproductve propertes wth elaborate press coverage to demonstrate the plght of the landless poor. Ths new publc relatons effort galvanzed voters and spurred the government to act on land reform. 15 As the numbers of nvasons multpled, urban voters were remnded of the task at hand, and land reform moved to the forefront of poltcal debate wth more resources devoted to t. The MST has become regarded as one of the most successful grassroots movements n the world and s frequently held as a model of nterest group effectveness even though t lacks drect votng power and funds to contrbute to poltcans. 16 IV. The Informatonal Role of the MST va Manpulaton of the Meda IV.1. The Argument. In order to understand the mpact of the landless peasant movement t s useful to explore the mplcatons of the model for land reform poltcs n Brazl wth and wthout the MST by comparng the perods before and after the group became actve throughout the country. 14 Although the Land Statute of 1964 already provded the legal bass for expropratons, the new Consttuton broadened the scope for the use of ths nstrument, made t easer to use and, most mportantly, sgnaled the dsposton of government and socety that land reform should be pursued wth hgh prorty. 15 For the hstory of the MST and an analyss of ther organzaton, thnkng and mpact see Wrght and Wolford (2003). 16 Noam Chomsky was cted by the Economst (2004) as statng that the MST s the most exctng popular movement n the world today. 10

12 In the pre-mst perod, there were two prncpals: landowners and urban voters. 17 The task for landowners was to ether block expropraton or f some land reform were to occur, to lobby for addtonal acton that benefted them as well, such as government credt and other subsdes. Both of these polces reduced the resources avalable for land reform. The second group, urban voters, sympathzed wth land reform because of hgh levels of land and wealth concentraton and averson to general nequaltes n the country. 18 But voters had lttle noton of the costs of land reform or of the government s efforts toward t. 19 Land reform was not a central preoccupaton of urban voters, who were more concerned about unemployment, nflaton, health, and crme. Consequently, they were ratonally mperfectly nformed about government land reform actons. Pror to the MST, voter nformaton was based on government self reportng. Wth lmted nformaton to generate poltcal pressure the government announced ambtous targets and set up a vsble bureaucratc structure yet wth lttle actual mplementaton. Wth the MST, ths changed. Sensatonal, hghly-publczed meda events--farm occupatons, marches, nvasons of governmental offces, roadblocks, and accusatons that the government was stallng--caused voters to reassess the commtment of government towards land reform, and to call for more nterventon. Poltcans responded by ncreasng land reform. IV.2. Evdence. The relatonshp between MST nvasons/occupatons and concrete government efforts n terms of famles settled on exproprated farms and government resources devoted to land dstrbuton s shown n Graphs 1 and 2. They show that crca 1993 the MST became more 17 The landless peasants and rural workers could also be consdered prncpals, but because they lacked organzaton they had lttle power to affect government polcy. 18 Ondett (2008) notes that land reform n Brazl s seen as a ltmus test of a presdent s progressveness. 19 Expropraton for the purpose of land reform n Brazl s, by consttutonal mandate, compensated at far value, though much of t n Ttles of the Agraran Debt, so that there are hgh costs for obtanng land as well as the expendtures to settle and mantan the benefcary famles. In 2004 the budget for INCRA, the federal land reform agency, was R$ 2.5 bllon (approxmately US$ 833 mllon. A study commssoned by the government and the FAO for the perod of found that the average drect cost of settlng a famly n a settlement project was US$ (Marques, 2007). 11

13 actve, ncreasng the number of nvasons and occupatons of prvate farms and publc land (Graph 1). There were correspondngly larger budgets for land reform (Graph 2) and greater numbers of landless famles settled (Graph 1). 20 The model specfes the channels through whch ths objectve s acheved, hghlghtng the role of nformaton asymmetres as well as nterest group characterstcs and preferences. The causal relatonshps lnk voters preferences, MST acton, the meda, government, and land reform. These lnks are shown n Fgure 1, wth the upper lne ndcatng where the nsghts from the model apply. We now present evdence of each lnk n the chan. Lnk 1 - Voters care about land reform. In the model ths lnk s represented by the vector b, a successful nterest group pursues a task that s desred (or abhorred) by other key prncpals. Ths stuaton allows the group to generate support among the other prncpals by alterng the nformaton they receve regardng the government s actons ( ). If the other prncpals are ndfferent to the objectve, nformaton adjustments wll have lttle payoff, and the group would do better by pressurng through another channel. In the case at hand, because voters care about land reform, MST has been able to extract more effort from the government through publcty. 21 The hgh salence of land reform n Brazlan poltcs s not a new phenomenon. It has been a central poltcal ssue snce the 1960s and remans so today despte currently hgh levels 20 Hereda et al. (2002), n a large sample of settlement projects, found that 96% of settlement projects emerge from some form of conflct rather than government ntatve. They also show that although there are several other groups, the MST s by far the most actve. The drop n occupatons and settlements after 1999 can be attrbuted to several factors. By that tme so many famles had been settled that the MST s man focus shfted to pressurng the government to make good on ts pledges of credt to those famles rather than obtanng more land for new famles. Ths s mportant for the MST as t receves 1-4 % of all the credt provded to settled famles. Also, n 1999 the economy passed through a severe crss nvolvng a large devaluaton of the real. In the followng years the government successfully dealt wth ths crss by pursung strct fscal restrant (Alston, Melo, Mueller and Perera, 2006), whch severely constraned the governments land reform effort. Fnally, n the year before the 2002 presdental electon, the MST purposefully reduced the number of nvasons so as not to harm the electoral chances of Lula, who they preferred to the more conservatve canddate. 21 An alternatve hypothess for why so much land has been redstrbuted n the past decades could be that the Presdent pursues land reform as a matter of publc nterest or deology rather than for Downsan motves. We wll show n ths and the followng subsectons that the full set of the evdence fts our hypothess better than alternatve hypotheses. 12

14 of urbanzaton, comng to symbolze more than the mere redstrbuton of land. Several publc opnon polls have been conducted over tme to gauge socety s poston towards land reform. Almeda (1998) revews eght opnon polls from 1962 to 1998 and shows that there has consstently been broad support towards land reform. In 1998, for example, a poll conducted by IBOPE revealed that 80% of those ntervewed were n favor of land reform. A survey n 2006 asked respondents whch reform the government should address frst (Fundação Perseu Abramo, 2006). When the queston was left open-ended, land reform was at the top of the lst, When a lst of reforms was proposed land reform fell below labor, socal securty, fscal and poltcal reform, but was nevertheless stll chosen by 45% of respondents as one of the top three prortes. It s mportant to make a dstncton between the sympathy of voters for land reform and ther vew of the MST. Durng most of the perod we are analyzng the dstncton between the MST and land reform was often blurred to the general publc wth the sympathy for land reform rubbng off on those who were helpng to promote t. The group s popularty ganed a boost from two hghly vsble confrontatons n 1995 n Corumbara n the state of Rondona and n 1996 n Eldorado dos Carajas n the state of Pará, where several landless peasants were klled. These events gave the MST natonal and nternatonal exposure and pushed land reform to the center of poltcal debate. As noted by the Economst (1997), the MST won the ultmate accolade: sympathetc portrayal n a prme-tme soap opera on Globo, Brazl s leadng televson staton. Ondett (2008) explans that (s)nce telenovelas are a natonal entertanment nsttuton on a par wth soccer, ths was the ultmate form of cultural recognton of the salence of the land queston to the Brazlan publc. More recently the approval of the MST has started to decouple from that of land reform. By 2006 the landless ssue may not have seemed as pressng as t had once been gven that an area greater than France and Portugal together had been redstrbuted. An opnon poll 13

15 by Ibope (2006) showed that 56 percent of Brazlans thought the MST brngs more negatve than postve results to land reform and 53 percent thnk the government should use the polce to evct the MST from evaded farms. Wth fewer unproductve farms left to occupy, the MST has taken to nvadng productve propertes, dentng ther mage wth much of the Brazlan electorate. We note that our argument does not requre the approval of voters of the actons of the MST and nvasons of prvate farms. All that s requred s that: 1) the MST s contnually able to elct meda coverage; 2) voters approve of land reform; and 3) the MST s actons mpart nformaton to voters on the government s efforts towards land reform. Durng most of the perod we are dealng wth n ths paper, ncludng n the regresson analyss below, not only land reform, but also the MST were well seen by the electorate. The analyss n ths paper refers to the perod up to 2007 and may be less suted for the subsequent perod. 22 Lnk 2 MST acton: occupatons, nvasons, marches. Graph 1 shows the number of occupatons by the MST and settlement projects created by the government. The data show that after 1993 the number of occupatons and the number of settlement projects started rsng, markng a clear break wth the past. Graph 2 shows that government expendtures rose smultaneously wth conflcts and settlement projects. Together the data n Graphs 1 and 2 are consstent wth our story of greater acton by the MST leadng to responses by the government n the form of greater expendtures on land reform and settlement projects. We perform Granger causalty tests to see f the data are compatble wth causaton from MST acton to government reacton. Table 1 panel A shows the descrptve statstcs for 22 Wth the large number of landless peasants that have been settled and gven land n the past ffteen years, the stock of actual landless peasants, that s, those that really have apttude to work the land rather than smply beng poor, has fallen and t may become harder for the MST to recrut n the future. Grazano (2004) argues that there are no longer any true landless n Brazl and that current MST occupatons are flled mostly wth poor, unemployed people lvng n bad condtons n the ctes. Smlarly, the stock of unproductve land avalable for expropraton has plummeted, not only because of the scale of past land reform, but also because of the rse of commodty prces snce 2003 and the rse of bofuels. 14

16 the data and Table 2 show the tests. Unfortunately these data are only avalable annually, from 1989 to 2005 so there are only 17 observatons, thus results are only suggestve. Because unt root tests have low power wth small samples we choose to do the causalty tests both n levels and n frst dfferences and then compare the results. The tests are run both between each par of varables as well as condtonally, addng lags of the thrd varable on the rght-hand sde. The results show an ntmate relatonshp between occupatons, expendtures and settlements. In addton the results show that greater effort by the government (expendtures) encourages the MST to pursue even more occupatons. 23 One of MST s strengths s that t has low labor costs (G e ), enablng a strategy of j nfluence that s labor ntensve. Most MST members are extremely poor wth few alternatves and are motvated by the opportunty for land. 24 The nvason and occupaton of a farm and turnng t nto a land reform settlement requre that large contngents of people are moblzed for long perods of tme. Between 30 to 500 people wat near a property for the rght tme to act. The nvason may occur peacefully or nvolve volent conflct. 25 If there are evctons, the group wll return to a provsonal camp to renvade the same property or start over elsewhere. Years can go by before progress s made. To mantan order the MST mposes strct dscplne n ts camps to overcome free-rdng, a necessary element n nterest group success (Olson, 1965 and Stgler, 1971). 23 Ths result s consstent wth Alston, Lbecap and Mueller (1999a, 2000). The results hold for both levels and frst dfferences, reducng concerns of spurous correlaton, especally n the lght of all the other qualtatve evdence we present about the nature of Brazlan land reform. 24 In ths regard Wrght and Wolford (2003: 54) cte a settler s recollecton of hs days n an MST occupaton: We lost what lttle we had when we went to the encampment. We could take lttle even of those few thngs that we owned nto the new encampment, the only thng we took was our (wood-burnng) cook stove. What lttle savngs we had were soon gone, because we were earnng nothng. We had no house nor land to return to, no household goods, hardly any clothng, very few of our tools everythng was lost. And there was no way to go back and be the same person agan to the old neghbors, the frends on the outsde. In other passages the authors also document several postve memores that settlers held from the occupaton days, n partcular the camaradere and the sense of empowerment from partcpatng n the movement. 25 From 1985 to 2003 there were 13,524 conflcts for land n Brazl, most of whch nvolved an nvason and the subsequent resstance to evcton (Comssão Pastoral da Terra, 2004). The number of conflcts follows closely the number of occupatons n Graph 1. 15

17 As noted by Becker (1983) what matters n competton between nterest groups s relatve poltcal pressure. In the competton for the attenton of voters va the meda, landowners are at a consderable dsadvantage despte havng fnancal resources because they have hgh opportunty costs and the publc s not partcularly senstve to ther nterests. Lnk 3 The Meda covers MST actons. There s competton for meda coverage, yet the MST has contnually proved to be a master n publc relatons, creatng newsworthy stuatons as part of ts strategy. These nclude nvadng farms belongng to hgh profle fgures, such as ex-presdent Cardoso (nvaded n 2002) or mega-nvestor Danel Dantas (2009), even whle knowng that those propertes were productve and could not be legally exproprated; stagng marches by landless peasants across the country; blockng roads and occupyng enterng government offces; and havng charsmatc leaders make bombastc statements to the press. Ondett (2008) provdes evdence that these actons by the MST have actually kept the land reform ssue contnually n the meda. He quantfed the volume of coverage of land reform by a the country s major weekly magazne (Veja) and by a major newspaper (Folha de São Paulo) from the early 1990s to the md 2000s and shows that although there s varaton over tme, reflectng specfc events such as the massacres and the marches, the ssue has remaned n the publc eye throughout the perod. There s also evdence that ths volume of coverage s not concdental or an unntended sde effect of the MST s acton, but rather has been purposefully sought as a means to pursue ther objectve. There s qute a szeable lterature dedcated to studyng the relatonshp between the MST and the meda, often n the journalsm and communcatons area (Berger, 1998; Vargas, 2006; Melo, 2007; Cruz and Varela, 2009; see several addtonal ctatons n Hammond, 2004). Berger (1998) dscusses a MST document from 1995 called For an MST Communcaton Polcy where the need to engage drectly wth the meda s explctly 16

18 recognzed. In Vargas (2006) the author spent a week n the MST s communcaton offce n Brasla and accompaned ts efforts to get the movement s actons covered by the press. Durng that week the MST was promotng ts yearly red Aprl event when the MST ncreases the number of occupatons n remembrance of the massacre n Aprl of She ntervewed not only the MST but also the reporters of major newspapers and concluded that the MST was hghly successful n puttng the ssue on the meda s agenda, though only partly successful n nfluencng the nterpretaton or readng gven by the meda to the events they were coverng. Hammond (2004) has smlarly found that the MST has been successful n attractng meda coverage. He analyzes the content of the meda s portrayal of the MST and concludes that the treatment of the MST n the Brazlan meda s dverse, presentng a mxture of sympathy and hostlty (Hammond, 2004:83). The ssue of the treatment gven by the meda to land reform brngs up the queston of the preferences and ncentves faced by the meda. We have treated the meda as beng a neutral player that smply passes on the nformaton created by the MST to the general publc. But clearly the meda can have ts own nterests both related to land reform and to other governmental polces, so t could be modeled as another prncpal n the game. 26 Especally at the local level landownng nterests concde wth ownershp of the meda. We choose, nevertheless, to not explctly model the meda as a prncpal. Although there s consderable concentraton n televson and the prnted press t s not so hgh that a sngle vew prevals and sgnfcant competton exsts n terms of ponts of vew. Also the qualty of most Brazlan meda s hgh compared to world standards wth good levels of objectvty and ndependence. As noted n much of the lterature cted above, the MST and land reform have gotten far and often sympathetc portrayals (Hammond, 2004; Berger, 1998; Vargas, 2006; Gohn, 2000), though the movement and other analysts would not agree. Smlarly there does not appear to be 26 See Besley, Burgess and Prat (2002) and Besley and Burgess (2001) for analyses of the motvatons and mpacts of the meda on polcy responsveness to preferences of voters. 17

19 underreportng of the ssue, n great extent thanks to the explct efforts of the MST for coverage. Therefore t s a reasonable smplfcaton to treat the meda as exogenous. Lnk 4 The voters are nformed by the meda. The productvty of effort n nfluencng nformaton ( ( e, e jj j j e j ) ) requres both attractng meda attenton and changng the perceptons of voters. The evdence we have presented ndcates that voters demand land reform, the MST ngenously create newsworthy content that the meda, whch has to compete for audence, systematcally chooses to keep broadcastng. One would not expect these actons to persst f voters were smply gnorng what the meda transmts about land reform. The man message receved by voters s not so much whether the MST s good or evl, but nstead that the government sn t movng forward wth land reform, nformaton whch s fltered nto ther subsequent electoral decsons, as we show n the next lnk. Lnk 5 The Presdent s pressured by the voters for more land reform. To emprcally support ths argument, we estmate a model of presdental popularty, addng to the usual specfcaton of economc and poltcal explanatory varables other varables that measure farm occupatons by the MST. The dea s to show that voters not only care about land reform, but effectvely send the message through to the Presdent. If we fnd that more MST occupatons reduce the Presdent s popularty, ths wll be evdence of the MST s effectveness n manpulatng perceptons of voters to ndrectly force the government to ncrease land reform, after all popularty s a poltcan s compass of what to do to assure poltcal survval. A negatve mpact of occupatons on presdental popularty could also be nterpreted as punshment by the voters for the Presdent beng soft on the MST and not stoppng them from breakng the law and dsrespectng property rghts. If ths were the case one would expect Presdents and presdental canddates to frequently adopt a hard-lne aganst 18

20 the MST n ther actons and rhetorc so as to captalze on ths supposed sentment of voters. But what we see s the opposte. Whenever tensons flare up n land-related confrontatons the Presdent typcally reacts by promsng more land, more credt and more effort. The voters default s to always assume that poltcans are evadng the promse to redstrbute land, for ths has hstorcally been the case, and the presdents reflex s to reterate the promse and up the ante whenever events cast a doubt on ther resolve. An example took place n May 2005 when the MST organzed a march of 12,000 landless peasants from dfferent parts of the country that met n Brasla where they held a large demonstraton. Instead of gnorng or repressng the event Presdent Lula receved the protesters n the governmental palace, as Cardoso had done n the 1997 march. Lula turned the event nto a photo op and promsed: 1) to send a presdental decree to Congress by the end of the month to provde an addtonal R$ 700 mllon for settlement projects; 2) to hre 1,300 new staff members for INCRA, the land reform nsttute; and 3) to tghten the crtera whch determnes how productve farms have to be so that they are mmune from expropraton (Estado de São Paulo, May 17, 2005; The Economst, May 19, 2005). Smlarly n Aprl 2004 the movement announced an ntensfcaton of ts occupatons to greet the new Lula admnstraton. MST leaders spun ths as an effort to create a red Aprl, a play of words on the red of the MST flag and the possblty that blood would be splled. The government reacted to ths threat not by rebukng the MST, but by expropratng 34 farms n Aprl, sgnfcantly more than average (Ornagh, 2004). It s therefore safe to nterpret a negatve mpact of occupatons on popularty as a punshment for not dong enough land reform. There s a large lterature testng the determnants of presdental popularty by regressng opnon poll data aganst a seres of varables that capture the state of the economy and poltcal events (Prce and Sanders, 1993; Edwards, 1991; Erkson, 1989; Markus, 1988; Monroe, 1984, Mueller, 1973). Our dependant varable s the percentage of the electorate that 19

21 fnds the Presdent s performance very good / good n perodc publc opnon polls performed by Datafolha Instuto de Pesqusas. 27 As explanatory varables we use: monthly data on nflaton; nterest rates; exchange rates; GDP growth n the last 12 months; unemployment; and lagged popularty, the latter varable to control for the nertal aspect of popularty and thus avod problems wth autocorrelaton. 28 In addton to these economc varables we added controls for poltcal and socal events that may have an mpact on popularty. The frst of these s a dummy ndcatng months when Presdent Lula was n offce (January 2003 onwards) wth Presdent Cardoso s terms n offce as the baselne. Another dummy controls for the Mensalão scandal (June to December 2005), where Presdent Lula was accused of purchasng votes n Congress. Gven the mportance of soccer n Brazl and the percepton that success can have economc and poltcal consequences (Falter, Pérgnon and Vercruysse, 2008; Ashton, Gerrard and Hudson, 2003), we added dummes for the World Cups of 1998, 2002 and 2006 (months of June, July and August). 29 The key to ths test s to add to the lst of economc, socal, and poltcal varables a measure of the MST s efforts at callng attenton to land reform. If ths measure reduces the Presdent s popularty ceters parbus, we wll have found evdence for our hypothess that the MST mpacts the government by ncreasng voters awareness of the ssue. We use two measures of the MST s efforts. The frst s the number of occupatons promoted by the MST and other landless groups lagged for three months. Occupatons are the man nstrument used by the landless to pressure for land reform and typcally yeld hgh vsblty n the meda. The second varable s a dummy for the red Aprl months (see Graph 3). These are the months n 27 Our perod of analyss s January 1997 to December The ntal date s constraned by the avalablty of monthly data on land occupatons ( There are data on Presdental approval rates for most months n the perod. When no poll was performed that month, we repeat the value of the prevous month. Usng only very good as dependant varable yelds smlar results. 28 Usng a measure of ncome nstead of GDP growth gave smlar results. 29 In 1998 Brazl fnshed n second place, n 2002 Brazl was the world champon and n 2006 t dd not reach the sem-fnals. 20

22 whch the MST ncreases ts publcty campagns as a trbute to the past massacres. 30 If ths dummy varable detects a fall n Presdental popularty n the subsequent month (May) of each year, ths result s consstent wth our hypothess that the MST, through the meda affects the popularty of the Presdent. Graph 3 shows the presdental popularty and the occupaton data from January 1997 to December 2007, hghlghtng the red Aprls and the terms of Cardoso and Lula. Because we are usng tme-seres data we have to ascertan whether the varables are statonary before regressng them, at the rsk of nterpretng spurous correlatons as meanngful relatonshps. Descrptve statstcs are n Table 1 Panel B. In Table 3 we present augmented Dckey-Fuller tests for unt roots for all the contnuous varables, specfyng the number of lags, sample sze and use of a trend. The second column shows that the null hypothess of a unt root s not rejected for all varables at a confdence level of 1%, though t s rejected for nflaton at 5%. In the last column we show the ADF tests after frst dfferencng the data. For all varables n frst-dfferences the null hypothess of a unt root s rejected. These results together suggest that all varables are I(1) and do not come from statonary processes, so that our regresson wll only make sense f the varables are found to be contegrated, whch wll be tested. Typcally unt root tests do not provde a defntve assessment of the order of ntegraton, so we wll also run the regresson on frst-dfferenced data that s statonary and can be consstently estmated by OLS. We present results n Table 4. In Columns I, II and III we show the results for data n levels under dfferent specfcatons to test the robustness of the results. In all three specfcatons the regresson s run usng OLS and the resdual s tested for contegraton. Gven that all varables are assumed to be I(1), f the resdual s found to be I(0) the results can be taken as meanngful rather than spurous and can be nterpreted as the long-term relatonshp 30 Actvtes nclude coordnated mass occupatons throughout the country, marches, festvals, and ntervews, wth specal care to elct meda attenton (Vargas, 2006). 21

23 among the varables. In all three specfcatons the resdual s tested wth a Phllps-Perron test and an augmented Dckey-Fuller test usng the approprate more strngent crtcal values (see bottom of Table 2). In addton a Johansen contegraton test s performed to test the null of no contegraton. The test statstcs and approprate crtcal values are presented n the bottom of Table 2. All the tests reject the hypothess of no contegraton, ndcatng that there s a nonspurous relatonshp between presdental popularty and the determnants we specfy. The frst column shows that lagged popularty has a coeffcent slghtly below 1 and s hghly sgnfcant, ndcatng that popularty tends to have hgh nerta. Of the economc varables GDP growth and nterest rates have a postve and sgnfcant effect on popularty, whle unemployment, as expected, has a negatve and sgnfcant mpact. Exchange rate and nflaton are found to have negatve, but not statstcally sgnfcant coeffcents. The poltcal varables also have the expected effects. The results show that Lula ceters parbus had hgher levels of popularty than Cardoso and that the Mensalão corrupton scandal dd dent that popularty. World Cups are found to have boosted the Presdent s approval rates, but only the 1998 cup s statstcally sgnfcant. 31 The key results are the effect of occupatons and the red Aprl dummy on popularty. The estmated coeffcent for occupatons s negatve and sgnfcant at the 1% confdence level. The mpact of a one standard devaton ncrease n occupatons n prevous months s to reduce the presdent s popularty by 1.4% when all varables are set at ther means wth dummes equal to zero except the Lula dummy. Ths comes from a fall n approval ratngs from 35.2% to 33.8%. Ths mpact s slghtly larger than the mpact of a one standard devaton ncrease of the other 31 A possble nterpretaton of ths result s that the Brazlan electorate matured as democracy consoldated and was less and less nfluenced by such poltcally perpheral ssues. Note that Brazl was second n 1998, won the Cup n 2002 and was ffth n

24 economc varables: GDP 1.2%; exchange rate -0.9%; nflaton -0.9%; nterest rates 1.0% and unemployment -0.9%. 32 The red Aprl dummy also provdes evdence that MST actvty reduces presdental popularty. There s a statstcally sgnfcant (10%) drop n popularty every year on the month after the wave of MST demonstratons. Wth all the varables at ther means (dummes=0, Lula=1) approval ratngs fall 1.7%, from 35.2% to 33.5%. The adjusted R-squared for ths regresson s Ths hgh value s partly due to the lagged popularty on the left-hand sde, but even wthout ths auto-regressve term the adjusted R-squared was stll In column II we estmate a specfcaton where the only economc varable s GDP growth. The reason for ths exercse s that the economc varables are naturally nterdependent, whch may rase concerns over multcollnearty, even though the correlatons are all under The estmated coeffcents of the occupaton and red Aprl stll retan ther sgnfcant mpact on popularty. In column III we test whether the mpact of occupatons vary n the Cardoso and the Lula admnstratons. Ths s done by addng an nteracton term of the Lula dummy and the occupaton varable. Because Lula s supposedly more deologcally algned wth land reform and the MST, one mght expect occupatons to be a more effectve nstrument wth hm. The results show that the opposte s true. For the Cardoso admnstraton the elastcty of popularty wth respect to occupatons s and for Lula t s consderably lower, , both statstcally sgnfcant at 1%. A possble explanaton s that Lula s left-wng credentals are suffcently strong to neutralze publcty about occupatons. A possbly complementary explanaton s that the publcs vew on land reform and the MST has gradually changed over tme, becomng less senstve to the nformaton conveyed through the occupatons. If ths s 32 Comparng the economc varables and occupatons n logarthms so that the estmated coeffcent s a measure of the elastcty of popularty wth respect to those varables, we fnd that the mpact of occupatons s somewhat lower: occupatons , exchange rate , nterest rate and unemployment