Critical Masses in the Decollectivisation of Post-Soviet Agriculture

Transcription

1 Citical Masses in the Decollectivisation of Post-Soviet Agicultue Matin Petick Leibniz Institute of Agicultual Development in Cental and Easten Euope (IAMO), Halle (Saale), Gemany and Michael R. Cate Depatment of Agicultual and Applied Economics, Univesity of Wisconsin-Madison, USA Decembe 2008 Abstact Decollectivisation in post-soviet agicultue has been slow except fo single islands of complete individualisation. We pesent a theoy that intelinks two types of citical mass phenomena to explain this outcome. A sufficient numbe of efom pionees is equied to eap the positive netwok extenalities fom individualisation. On the othe hand, manages, by limiting hoizons, exploit the pefeences of wokes fo confomity to cement thei own powe. Pola efom equilibia ae geneated that ae consistent with the obseved spatial pattens of estuctuing. Empiical suppot is povided by a theshold egession based on a unique dataset of egional efom outcomes in Moldova. Keywods Agicultual tansition; fome Soviet Union; social inteaction effects; netwok extenality; fam estuctuing. JEL classification D23; O18; P32; Q15. Acknowledgements The authos ae gateful to Heinz Hockmann, Andey Nedoboovskyy, Ebehad Schulze, the efeees of this jounal, and semina paticipants in Belin, Budapest, Halle, Kiel, Madison and Stuttgat fo helpful comments. The usual disclaime applies. Financial suppot by Deutsche Foschungsgemeinschaft (DFG) is gatefully acknowledged. petick@iamo.de; Phone: ; Fax:

2 Citical Masses in the Decollectivisation of Post-Soviet Agicultue Abstact Decollectivisation in post-soviet agicultue has been slow except fo single islands of complete individualisation. We pesent a theoy that intelinks two types of citical mass phenomena to explain this outcome. A sufficient numbe of efom pionees is equied to eap the positive netwok extenalities fom individualisation. On the othe hand, manages, by limiting hoizons, exploit the pefeences of wokes fo confomity to cement thei own powe. Pola efom equilibia ae geneated that ae consistent with the obseved spatial pattens of estuctuing. Empiical suppot is povided by a theshold egession based on a unique dataset of egional efom outcomes in Moldova. 1 Intoduction Afte moe than one and a half decades of agicultual tansition in the fome Soviet Union, economists continue to be puzzled by the lack of change in faming oganisation in many impotant successo counties (Leman et al. 2004). This pesistence is supising to those who expected that pivatisation was the key to efficiency and income impovements in agicultue. But equally supising is a shap vaiation in the extent of decollectivisation even within elatively naow geogaphic aeas. Fo example, collectives wee completely dismantled in some of the Soviet successo counties, such as Albania o the Tans-Caucasus. In Russia and Moldova, thee emeged islands of full pivatisation in a secto dominated by collective fam successos. Standad explanations fo the lack of estuctuing ely on a coe set of objective economic factos. Some authos ague that the existing faming technology was elatively capital-intensive and was not suitable fo individualisation, so that it was highly dependent on a web of input- and output-channels and high disoganisation costs wee to be expected. Futhemoe, decollectivisation would have meant the end of subsidies and govenmental potection, and would thus have inceased the isk fo pivate fames (Swinnen and Rozelle 2006). While these explanations ae no doubt impotant, they cannot explain vaiation within elatively homogenous counties o egions whee these conditions do not vay. In this aticle, we suggest that a full undestanding of decollectivisation equies a model that admits multiple equilibia. Based on ecent evidence on the fam estuctuing pocess, we ague that the intelinkage of two types of citical mass phenomena can povide such a model. The fist is the existence of a netwok extenality. It captues the idea that the development of a suppotive political and economic famewok, fo example in tems of public sevices and the emegence of supplies and pocessos geaed to small-scale poduces, equies a sufficient numbe of efom pionees in ode to become viable. At the same time, mutual leaning and skill acquisition educes uncetainty fo efoming fames. Ou second agument is elated to the often ovelooked pesonal inteests of the collective fam diecto. Diectos benefit fom the pe-efom status quo, because it assues them access to income, local powe and pestige. Assuming that wokes have pefeences fo behaving in confomity with pees, manages may find it expedient to manipulate thei wokes in a way that they eject any oganisational

3 2 change as being not confom with the nom. Following Schaffne s (1995) analysis of Latin Ameican land efom, we call this keeping the hoizon of fam wokes limited. In the sequel, we motivate ou hypotheses moe thooughly (section 2) and develop a fomal model of the two counteacting netwok effects (section 3). It applies to counties with edistibutive de-collectivisation policies, as opposed to counties that opted fo estitution to fome ownes. 1 A egional equilibium is deived in which eithe all copoate fams in a homogeneous agicultual aea emain intact o all fams ae dissolved (section 4). We use theshold egession methods based on a unique dataset of egional efom outcomes in Moldova to povide empiical suppot fo ou popositions (section 5). Moe conclusive testing between the two mechanisms will equie additional empiical wok, as we discuss in the final section of the pape. 2 Netwok effects and limited hoizons in the decollectivisation of post-soviet Agicultue Evidence on the pesence of netwok effects comes fom one of the few egions in Russia whee an almost complete tansition to pivate faming took place. As descibed by Pallot and Nefedova (2007, ), the Lysye Goy distict in Saatov oblast has been dominated by pivate fams since As an exception to the ule, the fome state fam diecto was the initiato of the fame movement thee, as he had not been on favouable tems with the local Communist paty in the 1980s. Togethe with the shaes of othe family membes, he set up the fist pivate fam. Two futhe specialists fom the fome state fam followed suit in establishing independent entepises. All wee able to exploit thei pevious contacts to obtain inputs and loans. Faming pactices wee adjusted to technological and economic constaints, which meant the widespead abandonment of commecial animal husbandy. Ceeal yields on pivate fams soon equalled o exceeded those of the fome state fam and new bick-built famhouses wee eected in the middle of the fams. In tun, seventy fome employees of the state fam followed these examples and set up peasant fams. Pallot and Nefedova (2007, 182) comment on these developments as follows: The withdawal of key pesonnel fom kolkhozi and sovkhozi shifted the balance fom the old-style Soviet collectivist entepises to new fam businesses until a tipping point was eached when new fam fomation became a mass movement. In the ealy days, wokes wee eluctant to quit agicultual entepises fo pivate fams because of loyalty to thei paent fam and, pobably moe impotantly, because they lacked confidence that the efoms would not be evesed. By the middle of the 1990s as pivate fams bedded down and agicultual entepises defaulted on wages, these inhibitions began to be ovecome and wok on pivate fams became sought afte. In the absence of a sufficient numbe of efom pionees, pivate faming will neve become a mass movement. Howeve, if thee is such economic potential in individualised agicultue, why is this citical mass so aely emeging? 1 In a simplifying view, asset edistibution, at least fomally, povided a faily lage but well-defined goup of ual esidents the option to appopiate a shae of fomely socialised assets. It was pactised in all counties of the Commonwealth of Independent States (CIS) and in some Cental Euopean counties, notably in Albania. Restitution, on the othe hand, esticted the goup of eligible ecipients and intoduced fome and possibly absentee ownes o thei heis as additional stakeholdes. This mode of estuctuing pevailed in most othe Euopean tansition counties (Leman et al. 2004, 85-93).

4 3 To addess this puzzle, ou second agument is elated to a stategy of limiting hoizons of fam wokes. Manages keep away the image of an independent peasanty fom thei wokes minds in ovet and subtle ways, fo example by withholding infomation concening pivatisation ights, peventing political oganisation of efomes, not allowing outsides to invade the village o stat businesses with defecting wokes, and by stessing the necessity of collective solutions to poblems. A ecent monogaph by Allina-Pisano (2008) on decollectivisation in Ukaine and Russia agues that collective chaimen signalled appaent compliance with efom legislation to the national govenment but locally shielded wokes fom any meaningful change. The study povides extensive evidence fo what we call limiting of hoizons. The autho epots how efom-willing fam wokes wee delibeately excluded fom public meetings o denied meeting space (p. 67), how agenda setting by manages in these meetings pevented discussions of independent faming as a viable altenative (p. 69), how land owneship cetificates wee kept in the collective fams safes athe than handed out to wokes (p. 71), and how manages intevened though back oom deals with local state officials to pevent land distibution to individuals (p. 86). The autho descibes how chaimen of collective fam successos and othe local authoities wee attempting to tun public opinion against pivate faming by launching citical aticles in the local pess, publicly belie it as something stange, suspicious and wothy of idicule, up to acts of open violence against family membes of pivate fames. Futhemoe, it is shown how authoities used the notion of the stange and outside as an instument to dive a wedge between the (loyal) ual population and pivate fames. The fact that many of these independent fames emeged fom maginal goups of the ual society, such as single women o membes of ethnic minoities, was publicly denounced, and epesentatives wee insulted and called Gypsies (pp ). In addition, Allina- Pisano (2008) mentions the difficulties national govenments o intenational agencies had in administeing efom-elated suveys to fam wokes, as manages wee attempting to diect the enumeatos to those who wee instucted to give coect answes. The autho heself was egulaly denied to talk to wokes about thei access to individual land shaes: one fam diecto categoically objected because he did not want membes of his collective to get any ideas (p. xxii). Given this evidence, the cental hypothesis of the aticle is that successful decollectivisation depends on whethe o not a citical mass of efom willing wokes exits fom the collective. This citical mass has a mateial and a psychological implication: it allows newly independent fames to eap netwok extenalities and establishes a new confomity inducing nom. While the mateial netwok effect establishes the necessay conditions fo lage-scale individualisation, it is not sufficient to set this pocess in motion. Wokes must also be willing to ente it, which they ae unlikely to do if the fam diecto systematically limits thei hoizons in ode to alienate them fom the efom altenative. 3 A model of oganisational choice in post-soviet agicultue This section puts fowad a fomal model in which the stategic choices of collective fam manages inteact with wokes loyalty decisions to detemine the oganizational stuctue of agicultue. The model captues two foces that make wokes decisions intedependent, one mateial o objective, and the othe pefeence-based o subjective. While these two foces ae edundant in a modelling sense, evidence indicates that both

5 4 ae opeative in eality. We show that the intedependence ceated by these foces can ceate multiple egional equilibia, one chaacteized by a pedominance of lage scale collective fams, and anothe chaacteized by small scale family fams. We use the model to develop specific testable hypotheses that we empiically exploe in section Collective fam loyalty and social utility Intospection and casual empiical obsevation confim that a human tendency fo confomity is ubiquitous in eveyday life. Accoding to Aonson (1992, 13-33), confomity is einfoced if the majoity of the goup has an unanimous opinion, if the othe goup membes ae impotant and compaable to the individual, o if the individual feas social punishment by pees. Schaffne (1995, 249) hence agues that it is paticulaly stong in ual communities whee the individuals wok, kinship, social and eligious goups ae almost coincident. In addition, people have a tendency fo confomity if the envionment of the individual becomes inceasingly uncetain, so that the behaviou of othes povides guidance on what is the ight thing to do (Aonson 1992, 28). In the tansitional context of fam estuctuing in the CIS, this is likely to be a elevant facto. It is theefoe not supising that a numbe of studies have found evidence in favou of deeply ooted pefeences fo goup confomity in ual aeas of the fome Soviet Union. Leonad (2000) gives a summaising account of the tadition of communal land owneship and management in Pe-Soviet Russia. The stong social consensus on the ules of the contempoay Russian village is descibed by Paxson (2002), who mentions the moal obligation to wok togethe and to help each othe in the village community; and a geneally stong emphasis of ecipocity. She also epots that the subodination of one s own will to that of the goup is a vitue explicitly endosed by villages. 2 Despite its intuitive plausibility, economists have only ecently paid attention to the fomal modelling of social inteaction effects (e.g., Bock and Dulauf 2001). One appoach is to split the utility function into a standad mateial component and an additively sepaable social component, whee the latte ewads confomity (penalizes deviance) fom socially nomal behaviou. Futhemoe, it is commonly assumed that deviations fa fom nomal o goup aveage behaviou ae penalised moe stongly. Wokes in ou model must make a binay choice whethe to emain loyal to the collective fam ( λ = 1) o to withdaw thei esouces fom the collective and ente pivate, small-scale faming ( λ = 0 ). Wokes ae assumed to act to maximize thei expected well-being as given by the composite utility function V: 2 V u( y( λ) ) αv( ( λ λ) ) 2 whee u(.) is conventional individual mateial utility, and ( λ λ ) ), (1) v is social utility. The tem λ measues aveage o nomal behaviou in the individual s social efeence goup. We assume that u ', v' > 0, and the paamete α [0..1] denotes how much the individual caes about goup confomity. Individual income, y( λ ), is given by: 2 Schmemann (1997, 314) wites, based on expeiences in a cental Russian village: The communal mentality of the peevolutionay countyside was only stengthened by collectivization, and those peasants who stayed on the land stayed thee pecisely fo the sense of collective secuity offeed. To gab a lage piece of land fo oneself and to milk it fo money was to spit in the face of the collective and to lose its potection.

6 5 wif λ = 1 y( λ) =, (2) πλ ( ) if λ= 0 whee w is the cetain wage eceived as a collective fam woke, and π is the stochastic net income the individual eceives as a pivate, small-scale fame. As will be discussed below in section 3.3, π depends on λ, the faction of local fam wokes who emain loyal to the collective. Equation (1) fomalises the idea that if α > 0, individuals have pefeences fo confomity with thei pees and in doing what is the nomal thing to do in a given social efeence goup. Both inceasing positive o negative deviations fom goup aveage cause inceasing discomfot, but thee is no discomfot if eveybody in the goup behaves identically and chooses the same λ. As a esult, outcomes will likely be homogenous within a social efeence goup, but may be adically diffeent between goups. 3.2 The geogaphy of social efeence goups We distinguish two goups that may constitute the social efeence goup and influence the behaviou of villages. We denote these as egional and national goups. The egional efeence goup is the community of people living in geogaphical poximity to the individual. We pagmatically identify this with the distict o aion, which has been the lowe level of the two-tieed administative system thoughout the fome Soviet Union. In most ual egions of Ukaine and Russia, each collective fam foms the economic and social cente of a village, and a aion contains a dozen o so collective fams. Sociological field wok in ual Russia has shown that thee has been some mobility within localities, fo example because villages wee abandoned by the govenment and the population foced to elocate to neaby places. Howeve, most ual people spend thei entie life in a cetain aea, whee they ae suounded by thei elatives (O Bien et al. 2000, 95). People living in a egion ae moe likely to meet in peson on a egula basis and hence fom a natual social efeence goup. 3 Social inteaction within this efeence goup, and infomation flow in paticula, is had to manipulate by local authoities. As a second potential efeence goup we posit a wide, potentially non-ual population that povides an altenative bluepint fo what is the ight thing to do and how to behave. Russian language has adopted the anglicist tem feme fo this goup, as used in public discouse and official statistics. It identifies the vision of an independent peasanty with a efom-oiented, uban majoity, and with family fams in Westen Euope o Noth Ameica, which ae pesented as a model fo agicultual estuctuing. With egad to de-collectivisation in agicultue, the mode of behaviou of this social efeence goup is codified in the national efom legislation, which gives an individual woke the ight to leave the collective and withdaw his/he assets. In counties with a stong tadition of individualised faming pio to collectivisation, the membes of fome geneations may constitute pat of the efeence goup. Infomation about this social efeence goup is pimaily tansmitted via the media, though television o news- 3 In addition, many egions ae homogenous in ethnic tems, some of them officially ecognised as autonomous aeas subject to the ethno-teitoial pinciple of both the fome Soviet and the cuent Russian constitutions (Stadelbaue 1996, 42-49).

7 6 pape, though tadition, but also via agents of change who ente a community, in village congegations, o by wod of mouth. 4 In contast to the naow efeence goup, inteaction with the wide efeence goup can be influenced by the local fam manage. He may o may not limit the hoizons of his wokes. As discussed in Section 2, manages may teat efom ideas (and those that live them) as stange, inappopiate and foeign. They may also withhold infomation concening pivatisation and othe civil ights, pevent political oganisation of fam wokes, exclude outsides fom the village, inhibit the ceation of suppot netwoks o businesses fo pivate fames, fighten defectos, and stess collective identity and local collective solutions to poblems. The degee to which the manage keeps hoizons limited detemines how stongly fam wokes identify themselves with the wide, efomminded efeence goup o whethe they identify thei egional compatiots as thei appopiate efeence goup. 3.3 Positive netwok extenalities in pivate faming In the context of fam estuctuing, positive netwok extenalities mean that as moe wokes of a given locale tun to pivate faming, the easie will be the access to technology suitable fo these new types of fams as well as to suppotive supplies and pocessos. These effects can be intepeted as eflecting economies of size in eoganisation. If a collective ceases to function and smalle fams emege, this involves substantial fixed disoganisation costs fo the shaeholdes: lage machiney can no longe be used, up- and downsteam maket channels ae disupted, and the oganisational knowledge to un a diffeent way of faming may not be thee. Howeve, as moe fams of a new type emege, the moe likely is it that altenative solutions which ae viable fo small fams will develop. Also public sevices, including appopiate extension, ae moe likely to adjust to the needs of pivate fames, and mutual leaning and coopeation among these entepeneus will be possible. In the model, disloyal wokes who opt to become fames eceive an uncetain net income fom pivate faming, π ( λ ), whee λ is the shae of loyal wokes in egion who have not taken up pivate faming and have loyally emained on the copoate fam. Individual income π is a andom vaiable with mean given by π ( λ ) and vaiance σ ( λ ). Reflecting netwok extenalities, we assume that π, σ 0. We futhe as- sume that pivate fames take λ as given and allocate othe vaiable inputs to maximize expected utility, but abstact fom explicitly including this decision into the model. 3.4 Wokes and manages choice A fam manage allocates the labou foce of his fam, N, and othe assets to poduce composite fam evenue, by using a given technology. Collective fam assets ae the 4 In Russia, this efeence goup has been epesented by the efom-oiented Association of Peasant Fams and Agicultual Coopeatives of Russia (AKKOR). Accoding to Wegen (1995, 28-29), AK- KOR had a netwok of banches in evey oblast by the mid 1990s. Since its foundation it has held annual congesses and publishes an own weekly newspape, The Russian Fame. It suppots pivate popety and feedom of land use. Although its pimay constituents ae pivate peasant fames, it appeas to have moe suppot in uban than in ual aeas. Simila movements exist in othe CIS counties, see section 5 on Moldova.

8 7 sum of all the individual asset shaes that have not been withdawn by disloyal wokes. Loyal wokes eceive an annual wage, w. Labou contacts can be enfoced costlessly by the copoate fam manage and thee is no othe employe in each fo fam wokes than the local copoate fam. 5 The manage is assumed to be the esidual claimant of fam pofit. 6 The manage is also the political authoity of a village and theefoe benefits fom the loyalty of fam wokes, who suppot him politically, pepetuate the existence of the collective fam and thus secue his income and local powe. Although not modelled fomally hee, benefits fo the manage may also be of a psychological natue. 7 In the following, we focus only on the loyalty decision of wokes and simply assume that all othe poduction factos available to the manage ae allocated in a standad pofit-maximising way, hence geneating a (cetain) goss pofit of Π fo the manage. In ode to maintain loyal wokes, manages must assue that the utility a woke can attain as a loyal collective fam membe, 2 ((1 ) ) V ( λ, θ, α λ = 1) = u( w) αv λ, (3) is no less than the expected utility that the woke could attain as a pivate fame, 2 [ ( π ( λ ))] α ((0 ) ) V ( λ, θ, α λ = 0) = E u v λ. (4) The equiement that (3) exceeds (4) can be used to define a minimum paticipation wage, w ~, defined as: { ( ) ( ) ( ) } 2 2 w% ( λ, θ, α) w u( w) = E u π( λ ) α v (0 λ) v (1 λ) (5) Note how the level of w ~ is influenced by both the cetainty equivalent of pivate fam income and the adjustment fo social utility. The latte is negative fo a pospective nonconfomist pivate fame and positive fo a non-confomist loyal woke. Thus, the loyalty peseving wage level can be lowe if all wokes in the social efeence goup emain loyal, while it must be highe if all individuals in the goup leave the collective. As expession (4) makes clea, the paticipation wage the manage must offe to attact wokes deceases if wokes social efeence goup can be esticted to a goup of lagely loyal, collective fam suppotes. Being awae of social efeence goup effects among his/he wokes, a manage can delibeately keep the hoizon of fam wokes limited by shelteing them fom the national efeence goup and by exeting explicit o implicit political and social pessue on them, as descibed above. Let θ = 1 if the man Migation as an altenative option is hence not consideed. In the mid 1990s, as living standads in uban centes often declined faste than in ual aeas, uban-ual migation in fact was a common tend in Russia (O Bien et al 2000). Little is known about the emuneation of the coe management of copoate fams. Koeste (1999, 216) epots that many of the fam chaimen appea to enjoy a espectable living standad despite the poo economic situation of agicultue. It is hence assumed that they ae the de facto esidual claimants of pofits. Psychological benefits may aise because manages enjoy being the head of a commune (Amelina 2000, 503). The manage may also have a pefeence fo seeing agicultual poduction oganized in copoate fams. This could be the case if he/she has pofessional concens that the fam s dissolution will be economic mischief and lead to a food cisis.

9 8 age actively keeps the hoizon of his loyal wokes limited and θ = 0 othewise, with θ [0..1]. Limiting the hoizon has an influence on which efeence goup fam wokes use to assess thei utility fom behaving loyally to the manage. If wokes hoizon is kept limited, they compae thei own behaviou with that of all othe wokes in the egion, λ = λ. If the hoizon is not kept limited, they compae thei behaviou with a n wide, national efeence goup, λ = λ. The national efeence goup is assumed to be n at least modeately efom-minded on aveage, so that λ < As a consequence, wokes who accept the wide social efeence goup always expeience less discomfot fom choosing disloyalty than fom emaining loyal to the manage. It is assumed that n λ is exogenous and can not be influenced by decisions of individual fam wokes. By choosing θ, the mange detemines the elative weight of the two possible efeence goups of wokes: λ ) n = θλ + ( 1 θ λ. (6) The cost function C θ = θ C ( θ, R) gives the costs associated with limiting hoizon. Costs may aise fom own political activity of the manage to tun down efom-minded influences fom outside the egion, o bibes to public authoities who might stand up fo civil ights of wokes. This function depends on a vecto of egional chaacteistics, R, that includes the existence of consevative vs. efom-oiented political netwoks in a given egion, stength of collective vs. individual taditions, distance to uban centes, climatic and technological dimensions of agicultual poduction, etc (such diffeences θ ae discussed, e.g., by Amelina 2000). It is assumed that > θ C 1 0 and C > 11 0, implying that it is maginally costlie to achieve highe levels of shelteing. The suvival of the copoate fam in a given village depends on the ability of its manage to assue loyalty of a sufficient numbe of wokes in that village, subject to a budget constaint. Assembling these pieces, the optimisation poblem fo a manage in village j in egion R is: Min C θ, w m θ = wn + C ( θ, R), (7) j subject to: θ Π C w (8) N j w w% ( λ, θα, ) (9) whee constaint (8) defines the copoate fam s budget constaint and constaint (9) defines the loyalty paticipation constaint (heeafte denoted as the LPC). Both will be just binding unde optimising behaviou. (7) to (9) thus detemine least-cost loyalty elicitation fo the manage. 4 Regional efom equilibium with citical mass effects The pio section has laid out the basic stuctue that detemines wokes and manages choices. Howeve, these choices ae not independent, and this section consides the distinctive types of egional equilibia that can emege.

10 9 The effects of modifying the thee paametes λ, θ and α can be studied in Figue 1 (a)-(c). It displays the wage wokes equie to emain loyal as a function of the egional shae of loyal wokes, λ. The solid line in each gaph hence denotes the paticipation wage w ~ accoding to (5). Figue 1 (a) assumes that thee is no social utility effect and no limiting of hoizons (i.e., α = 0, θ = 0 ). The solid line shows the pue positive extenality effect of each woke s loyalty decision, diven by the poductivity of pivate faming, π. It illustates how positive netwok effects fom establishing independent pivate fams lead to highe pay-offs if moe wokes choose to exit. Wage payments highe than π imply that wokes emain loyal, wheeas payments lowe than π lead wokes to withdaw. The tem w = { w u( w) = E[ u( ( λ 1) )]} π denotes the minimum wage the manage must offe if all wokes in the locale ae loyal. Note that this is the cicumstance when Π pivate faming is least poductive. The tem wmax = indicates the maximum wage N that the manage can affod to offe, given the poductivity of the collective fam. The intesection of w max with the solid line defines a egional lowe theshold of loyal wokes, λ. If λ falls below this lowe theshold, copoate faming will no longe be sus- l min = tainable in the egion and the only institutional equilibium fo the egion will be complete agicultual pivatisation. 8 We thus see the sense in which this model has a tipping point at λ l. At loyalty levels above λ l, copoate faming will be etained. But once a citical mass of efom-minded wokes is eached, the efom equilibium is self einfocing, as indicated by the solid aow. Fo this eason, even though independent faming yields highe pay-offs, extenal foces, fo example an infomation campaign, may be necessay to each this equilibium. How likely it is that pivate fams emege depends on the location of λ l. The futhe this is on the left, the highe the pobability that collective fams emain intact. 8 If the collective s budget constaint is soft, fo example because it is subsidised by the egional govenment, loyalty equilibia ae of couse moe likely to pesist.

11 10 Figue 1: The impact of citical mass effects on egional loyalty equilibium paticipation wage w ~ (a) netwok extenality only θ = 0, α = 0 w max w mi n all withdaw 0 λ l 1 λ all loyal (b) netwok extenality & confomity θ = 0, α > 0 wmin λ ' (c) netwok extenality & confomity & limiting of hoizons θ > 0, α > 0 θ* l w max * w* λ * l ~ λ Souce: authos figue.

12 11 Dawing the solid line staight is subject to the assumption that both u and π ae linea. Dopping this assumption may lead to a diffeent shape. Conside the following modifications. Fist, a concave utility function would imply isk avesion on the side of wokes. It leads to a downwad shift of the paticipation wage, as uncetain income fom pivate faming is educed by a isk pemium. This shift is stonge fo lowe levels of λ, as the maginal utility of additional pofit diminishes. Second, fixed set-up costs fo pivate fames, fo example due to the establishment of appopiate up- and downsteam industies, make pivate faming less pofitable if these costs have to be bone by a few efomes only. The moe wokes individualise, the lowe is the cost pe fame. As a esult, π becomes locally convex fo high levels of λ. Finally, if all wokes leave the fam, netwok extenalities may diminish due to cowding effects. Fo example, it may become hade to find buyes fo additional fam poducts. This makes π locally concave fo low levels of λ. We have dawn an altenative dotted line into Figue 1 (a) that takes these effects into account. If this was the tue shape of the paticipation wage line, λ l would shift to the left and pivatisation became less likely. Fo simplicity easons, we stick to the staight line in the following. Howeve, ou findings ae not affected by the shape of the cuve as long as it emains stictly deceasing in λ. We conside now what happens when we eintegate social utility in the model, α > 0. Initially, we assume that hoizons ae not limited such that social noms ae set by the national level efeence goup, i.e., λ = λ n < Unde this assumption, wokes ae influenced by po-efom goups and feel uneasy about behaving loyally. Loyalty is now moe expensive to elicit, the theshold point, λ, shifts to the ight, and w min shifts upwad (denoted with a pime). The ange ove which a egional loyalty equilibium obtains thus shinks, making it moe likely that the egion will shift to the full efom o pivatisation equilibium. While social effects thus ae a theat fo the existence of the collective fams (and hence the income base and powe of the manage), they also povide a mechanism that the manage can use to avet this fame theat (Van Atta 1993). In ou model, the manage can eithe make loyalty moe attactive by inceasing wages, o influence the social efeence goup of wokes by actively limiting thei hoizon. The latte effect is displayed in chat (c), whee θ > 0. The LPC in Figue 3 (c) is assumed to epesent θ at the optimal level fo some elatively high level of λ. Fo that optimal level θ*, the paticipation wage line will cut fom below the LPC fo the no social effects case, as loyal woke feel less iitated by the po-efom behaviou of the wide efeence goup. The given level of θ* defines a ~ λ as the citical value of λ such that λ = At this ~ λ, the individual woke s disutility fom choosing eithe loyalty o disloyalty is just equal, 2 2 and v ((0 λ ) ) v( (1 λ ) ) = 0. Fo λ > ~ λ, manages can elicit loyalty by offeing a wage lowe than that equied to elicit loyalty in the absence of social effects. Fo λ < ~ λ, a highe wage must be paid to elicit loyalty. Howeve, this wage will still be less than that equied to elicit loyalty when manages do not limit hoizons in the pesence of social effects. Note that w max and the budget available to pay wages diminishes when manages choose θ > 0. But despite this offsetting effect, the capacity of the man- age to limit hoizons will always shift the theshold loyalty level, λ l, e-expanding the ange ove which copoate faming can be sustained, as shown in Figue 3 (c). l

13 12 By changing the pay-offs, keeping the hoizons limited has anothe pofound effect on the egional equilibium which is vey much in the inteest of the manage: it establishes a second pola equilibium fo aveage egional loyalty levels to the ight of ~ λ. Once the majoity of wokes has decided to emain loyal, this pocess is self-einfocing, as indicated by the additional solid aow. By stessing the collective identity of wokes and keeping away any efom-minded attitudes fom villages, the manage can establish a stable equilibium that guaantees the existence of the collective fam. Because wokes feel comfotable with doing what is, in the community, the nomal thing to do, they have an incentive to choose loyalty if the majoity did so aleady. With egad to wokes pay-off, howeve, the pola loyalty equilibium as dawn in chat (c) is inefficient as compaed to the full de-collectivisation outcome. Even so, because limiting the hoizon shifts the theshold loyalty level λ l back to the left, it makes it even moe likely that a loyalty equilibium occus. It follows that within a pool of egions with compaable social and geogaphical chaacteistics, thee ae eithe egions which totally de-collectivise, so that thee ae no loyal fam wokes and a widespead establishment of pivate fams, o egions which keep collectives completely intact, so that thee ae only loyal fam wokes and a loyalty cultue pesists. Because only pola efom equilibia ae stable, thee will be no intemediate o mixed estuctuing outcomes, ceteis paibus. If manages keep hoizons limited, maginal impovements in the elative etuns to independent faming have no effect on wokes loyalty. Whee keeping hoizons becomes exceedingly costly because widespead access to infomation and unambiguous efom policies ease coodination on a de-collectivisation equilibium, collective fams will dismantle unless manages ae able to offe highe wages. 5 Exploing bipola land efom egimes in Moldova We now exploe what ou model contibutes to undestand fam estuctuing pattens in Moldova. In the following, we show that conventional factos of estuctuing, such as egional pice and poductivity levels, ae not sufficient to explain vaiation in decollectivisation outcomes. In paticula, bipola pivatisation egimes coexist and outcomes within these egimes diffe in distinct ways that ae consistent with ou model pedictions. Contay to most othe counties that emeged fom the fome Soviet Union, Moldova undewent a slow efom pocess that took seveal yeas, but nevetheless poduced a significant numbe of individual fams. 9 This gives the eseache a unique souce of data to study the egional evolution in elative efom pogess, showing consideable vaiation acoss egions. Detailed figues on the numbe of individuals entitled to land and the actual ecipients of land wee collected by the Moldovan Depatment of Statis- 9 Moldova adopted a elatively libeal land efom legislation in Howeve, consevative foces had gained political powe when it came to its implementation in the following yeas. It was only unde pessue fom intenational oganisations such as the Intenational Monetay Fund (IMF) that a moe compehensive land efom pogamme was olled out nationwide afte See Leman et al. (1998) and Goton and White (2003) fo details.

14 13 tics at the distict (aion) level. 10 The pivatisation ate, PR, is the shae of land ecipients in all individuals with a ight to land, in pecent pe distict. While most ual beneficiaies had obtained povisional land titles on pape by the end of 1996, they had to actively apply to local authoities to eceive a physical plot of land in exchange fo thei pape titles (Leman et al. 1998, 15). PR is the complement to λ in ou model, the shae of loyal wokes in a egion, in the sense that λ = 100 PR. Geogaphically, Moldova is pedominantly flat with about 75 pecent of land used by agicultue, mostly coveed by fetile black soils (Petick 2008). Population and infastuctue ae spead moe o less evenly acoss the 40 administative disticts of the county, each including a medium-sized town with the seat of the local govenment. Natual conditions ae elatively homogeneous and infastuctue is only polaised to the extent that thee is a clea administative and economic cente of the county, the capital Chisinau. In the light of ou theoetical model, the distance to the capital could be used to measue how costly it is fo manages to shelte thei villages fom po-efom influences. Simila to the situation in othe post-soviet counties, po-efom goups of politicians and fames have pimaily been active in the capital. In the teminology of ou model, they epesent the wide efeence goup. 11 Opposed to this have been the pedominantly ual disticts of the Moldovan hinteland, whee diectos of collective fams attempted to inhibit individualisation. Qualitative evidence on the stategies employed to pevent estuctuing of collectives stikingly esembles the epots fom Ukaine and Russia quoted above (Csaki et al. 1997, 10, 24). These authos also epot that local authoities esistance to decollectivisation was paticulaly ponounced in the easten and southen peipheal disticts (p. 11). Holding othe factos constant, fams futhe away fom the capital should theefoe exhibit highe loyalty ates, as manages ae moe likely to limit the hoizons of wokes. To exploe the latte hypothesis, Figue 4 displays the elationship between the aion loyalty shae (in pecent) in 1998 and the distance in kilometes (km) between aions and the Moldovan capital city of Chisinau. 12 To visually captue the tendency in the data, a kenel egession cuve is included. As can be seen, thee is a stong positive elationship in the data. Moeove, moe distant aions exhibit elatively little change in loyalty shaes The data was published by Moldovan Economic Tends (MET) (1998). No data is available fo the seceding disticts east of the ive Nistu. The disticts foming the autonomous egion of Gagauzia ae teated as one spatial unit. Refom attempts wee backed by the newly founded National Fames Fedeation, the Chistian Democatic Popula Font, which is a cente-ight political paty, as well as by the U.S. Agency fo Intenational Development and the Wold Bank (Goton and White 2003, 308). As disticts wee concentated into nine lage administative units afte 1998, the disaggegated data was no longe published

15 14 Figue 2: Loyalty in Moldovan disticts vesus distance to capital Collective Fam Loyalty, λ (%) Distance to Capital (km) Souce: Authos calculations based on MET (1998) fo loyalty ates; Intenational Foundation fo Electoal Systems (1994) fo distance between distict capital and Chisinau. These diffeent pattens of change could of couse be explained by othe factos, including diffeential incentives fo pivate faming. We theefoe use λ as a dependent vai- able in a egession model of fam estuctuing: λ = β d + β x + ε, (10) i d i x i i whee λ i is the shae of loyal wokes in aion i, d i is distance to capital, x i is a vecto of covaiates, the β s denote paametes to be estimated and ε i is an identically and independently distibuted eo tem.

16 15 Table 1: Regession estimates of collective fam loyalty in Moldovan disticts Explanatoy vaiable Basic model Cente egime (distance <= 88 km) Peiphey egime (distance > 88 km) Mean (total sample) Coefficient p-value Coefficient p-value Coefficient p-value Intecept *** *** *** Distance to capital (km) 0.18 *** * Pice index of pivate fam output 1997 (Chisinau=100) ** ** Av. potato yield (dt/ha) *** Av. collective fam pofitability 1997 (ths. lei) Av. industial wage 1997 (lei/month) ** < ** * Numbe of obsevations R² Confidence inteval fo theshold (at 95%) Joint R² of theshold model 0.91 Notes: Dependent vaiable: egional shae of loyal fam wokes in Januay 1998 (in pecentage points). * (**, ***): significant at 10% (5%, 1%) level. Theshold model esults based on Hansen s (2000) pocedue. Souce: Authos calculations.

17 16 Ou set of contols includes a egional pice index of agicultual goods poduced by pivate fames in 1997; 13 the aveage potato yield in , in dt/ha; 14 the aveage pofit of collective fams in the egion in 1997, in thousand lei; and the aveage industial wage in 1997, in lei/month. The fist two indicate the competitiveness of pivate fams and should have a negative effect on aveage egional loyalty. The thid measues the economic and technological viability of lage-scale faming in a given egion. Moe pofitable fams povide moe leeway to keep wokes loyal, this indicato should theefoe be positively coelated with loyalty. High non-fam wages may pevent wokes fom applying fo a physical land plot and look fo industial employment oppotunities instead. This implies a positive effect on loyalty. Mean values of the covaiates ae given in the most ight column of Table 1. The fist column of Table 1 pesents odinay least-squaes estimates of ou basic egession equation. Distance poves to be the only significant explanatoy vaiable, suppoting the gaphical analysis above. Howeve, the distance indicato is too ough to exclude othe intepetations, such that pices o maket access fo pivate fames wee themselves diven by distance to capital. While we cannot fully esolve this poblem due to a lack of bette infomation on the costs of limiting hoizons, we can exploe whethe thee ae othe pattens in the data that ae consistent with ou theoy and ae at the same time had to explain by conventional factos of estuctuing. We focus on two pedictions of ou model. Fist, it implies that estuctuing tends towads bipola equilibia. Second, it says that pivate fam pofitability and hence individualisation is diven by pice and poductivity diffeentials only if manages do not limit hoizons of wokes. Ou stategy is to use a theshold estimato due to Hansen (2000) to test whethe thee ae two diffeent estuctuing egimes and to what extent decollectivisation follows diffeent incentives within these egimes. Figue 2 suggests that both a low and a high loyalty level equilibium may exist in Moldova. Hansen s (2000) pocedue allows to estimate whethe the data do beak into sepaate egimes when examined along the distance dimension. Moe fomally, we genealise ou egession model to: + + < ~ c c β d di β x ' xi ε i, if di d, λ i = (11) p p β d di + β x ' xi + ε i, othewise, whee d % is the citical o theshold value of distance whee a aion shifts fom the high loyalty equilibium to the low loyalty equilibium. The supescipts on the paametes indicate that we will now pemit the paametes to vay between close-in, cente aions (supescipt c) and moe distant o peipheal aions (supescipt p). Hansen s method allows us to estimate d %, test its significance and also to see if the two paamete vectos diffe fom one anothe Pice notations wee taken fom EU-Tacis (1997, 16) and wee aggegated accoding to an aveage poduction basket of pivate fams epoted by Leman et al. (1998, 75-76). All othe data is fom Depatment of Statistics of the Republic Moldova (va. issues). Potatoes tuned out to become the majo cop gown by peasant fams and in household gadens. Due to its labou-intensive poduction, collective fams begun to specialise in othe cops. We use data fo to measue poductivity that is undistoted by the onset of estuctuing.

18 17 The second two sets of columns in Table 2 pesent the esults fom this theshold estimation, identifying a theshold at a distance of 88 km fom the capital. It is highly significant with a 95% confidence estimato not diffeent fom 88 km. We theefoe have evidence that egions follow two opposite paths of estuctuing, as suggested by ou theoetical model. In addition, the individual paamete vectos vay consideably between the two egimes and the base model which implicitly imposed the estiction of equality between the two paamete vectos. In the cente egime, the coe dives of pivatisation ae the egional pice level fo individual fam poducts and the egional potato poductivity as a key output of pivate fams, wheeas the existence of pofitable collective fams delays estuctuing. Futhemoe, highe industial wage levels educe pivatisation ates. Inteestingly, distance is no longe a significant explanatoy facto fo loyalty within the cente egime, though it explains the shift between egimes. All this is consistent with the view that, close to the capital, decollectivisation stongly eacts to economic incentives because limiting of hoizons is too costly. Within the peiphey egime, slight diffeences in decollectivisation ae still weakly coelated with distance. While thee is also a small and only weakly significant wage effect, pice and poductivity incentives fo individual faming no longe have any negative impact on loyalty. Loyalty is even stonge in egions with highe pices fo pivate fam poducts. This esult is again consistent with ou theoy which suggests that esponses to maginal changes in etuns to pivatization ae muted by social foces in the peipheal aea. 6 Conclusions The pesence of citical mass phenomena has implications fo the design of policies aiming at the establishment of independent fams. Given a loyalty equilibium, it is not sufficient to impove manageial esouces and elax facto maket constaints fo pospective individual fames, as agued e.g. by Rizov (2003). The effect of maginal impovements in individual fam pofitability on the loyalty equilibium in ou model will be nil. Cucial fo efom in the pesent model is the fomation of a sufficient numbe of wokes who ae willing to establish independent fams, such as in the Russian case potayed by Pallot and Nefedova (2007, ). This could possibly be achieved by suppot pogammes which make loyal fam wokes awae of the fotune of successful non-loyal wokes, which lead to the emegence of individual consciousness aises among the goup of loyal fam wokes (and not only to the spead of disembodied ideas), o which make it moe costly fo fam manages to keep the hoizons of wokes limited. In othe wods, a big push in efom attitudes among wokes is a pecondition fo eaching the full de-collectivisation equilibium, which may be induced by a sufficient numbe of positive examples of independent faming in a egion. Whethe moe than fifteen yeas of stagnation in the non-efoming counties have einfoced o eoded existing noms of collective poduction may be an inteesting issue fo empiical eseach. Such eseach may also help to futhe disentangle the two citical mass phenomena we wee suggesting. Fo the case of Moldova, we showed that pivatisation inceased with bette pice and poductivity incentives in egions close to the capital. A moe complete test of netwok extenalities would demonstate that individual fam pofitability in-

19 18 ceases if moe wokes exit the collective. Diect evidence on social pefeences is difficult to obtain and would equie additional qualitative and/o individual-based psychological methods, such as used in Allina-Pisano (2008) o Aonson (1992). Altenatively, detailed data on individuals, including thei isk pefeences, could be used to compae pivate fam income with wage levels in the collective to check whethe esidual diffeences can be explained by social effects. In this way, even a money metic confomity pemium could be quantified. Following Schaffne (1995), ou theoetical model epesents a subtle depatue fom the taditional assumption of exogenous pefeences. By keeping hoizons limited, a manage can shape the social efeence goup of wokes and theeby influence what they egad as the nomal thing to do. Moeove, the manage can, fo his o he own benefit, depive wokes of a moe poductive efom option. Wokes then evaluate individual faming by efeing to thei cuent efeence goup, although they would be exposed to a diffeent efeence goup if they left the collective fam. Wokes who fo some eason escape the limited hoizon find themselves bette off than they thought they would be, and bette off than they had been.

20 19 Refeences Allina-Pisano, Jessica (2008): The post-soviet Potemkin village. Politics and popety ights in the black eath. Cambidge: Cambidge Univesity Pess. Amelina, Maia (2000): Why Russian Peasants Remain in Collective Fams: A Household Pespective on Agicultual Restuctuing. In: Post-Soviet Geogaphy and Economics, 41, Aonson, Elliot (1992): The Social Animal. Sixth Edition. New Yok: W.H. Feeman. Bock, William A.; Dulauf, Steven N. (2001): Discete Choice with Social Inteactions. In: Review of Economic Studies, 48, Csaki, Csaba; Leman, Zvi; Nucifoa, Antonio (1997): Land Refom and Pivate Faming in Moldova. EC4NR Agicultual Policy Note No. 9. Wold Bank. Washington, D.C. Depatment of Statistics of the Republic Moldova (va. issues): Anuaul Statistic al Republicii Moldova = Statistical Yeabook of the Republic of Moldova (in Moldovan). Chişinau: Depatment of Statistics of the Republic Moldova. EU-Tacis (1997): Noutatile Pietei Agicole (Agicultual Pice Notations). Iunie 1997 No 4. Chisinau: EU-Tacis. Goton, Matthew; White, John (2003): The Politics of Agaian Collapse: Decollectivisation in Moldova. In: East Euopean Politics and Societies, 17, Hansen, Buce E. (2000): Sample Splitting and Theshold Estimation. In: Econometica, 68, Intenational Foundation fo Electoal Systems (IFES) (1994): Republic of Moldova Paliamentay Elections Febuay 27. Chisinau: IFES. Koeste, Ulich (1999): Die Bedeutung de Oganisationsstuktu landwitschaftliche Untenehmen fü die Entwicklung des ukainischen Agasektos. In: Camon-Taubadel, Stephan von; Stiewe, Ludwig (ed.): Die Tansfomation de Landwitschaft in de Ukaine. Kiel: Wissenschaftsvelag Vauk, Leonad; Caol S. (2000): Rational Resistance to Land Pivatization: The Response of Rual Poduces to Agaian Refoms in Pe- and Post-Soviet Russia. In: Post-Soviet Geogaphy and Economics, 41, Leman, Zvi; Csaki, Csaba; Fede, Geshon (2004): Agicultue in Tansition. Land Policies and Evolving Fam Stuctues in Post-Soviet Counties. Lanham, MD: Lexington Books. Leman, Zvi; Csáki, Csaba; Mooz, Victo (1998): Land efom and fam estuctuing in Moldova. Pogess and pospects. Washington, DC: Wold Bank. Moldovan Economic Tends (MET) (1998): Moldovan Economic Tends Quately. Euopean Expetise Sevice; Ministy of Economy and Refom. Bussels, Chisinau. O'Bien, David J.; Patsiokovski, Valei V.; Deshem, Lay D. (2000): Household Capital and the Agaian Poblem in Russia. Aldeshot: Ashgate. Pallot, Judith; Nefedova, Tatyana (2007): Russia's unknown agicultue. Household poduction in post-socialist Russia. Oxfod: Oxfod Univesity Pess. Paxson, Magaet L. (2002): The Cultual Dimension: Social Oganization and the Metaphysics of Exchange. In: O'Bien, David J.; Wegen, Stephen K. (ed.): Rual Refom in Post-Soviet Russia. Washington D.C.: Woodow Wilson Cente Pess, Petick, Matin (2008): Landwitschaft in Moldova. Halle (Saale): IAMO (Discussion Pape, 117).