Brazilian land tenure and conflicts: the landless peasants movement.
Barros, Carlos Pestana ; de Araujo, Ari Francisco, Jr. ; Faria, Joao Bicardo 等
This article analyzes conflicts in Brazil involving landless
peasants and the violence that frequently results from their invasion
and occupation of privately owned rural land for the period 2000--08.
Land ownership in Brazil is overwhelmingly and historically
characterized by large, family-owned estates (Pichon 1997). The unequal
and inequitable .allocation of land together with weak institutions,
weak markets, and low asset endowment may make land reform a low
priority (Binswanger and McIntire 1987, Sjaastad and Bromley 1997). In
the absence of effective land reforms, these factors may lead to the
occupation of land by the landless poor peasants by violent means
(Assunrcao 2008). In such an environment, land-related conflicts are
common and have been previously analyzed in several studies, with a
particular focus on Africa (Andre and Platteau 1998, Deininger and
Castagnini 2004) and Latin America (Alston, Libecap, and Mueller 2005).
Credit rationing is part of the problem, since without credit
individuals may not be able to undertake indivisible investments, such
as purchasing land, which have a long period of maturation. As a result,
due to the lack of access to credit markets, the poor peasant may fail
to escape from poverty by not being able to own land (Fenske 2011).
The landowners have responded to the threat of land invasion and
occupation with large-scale evictions, adopting extensive livestock
production and highly mechanized cultivation methods, which reduce the
need for peasant labor on the farms, thereby creating obstacles to land
reform (Binswanger, Deininger, and Feder 1995). In the light of these
events, Latin American land reform has been described as a lost cause
(De Janvry and Sadoulet 1989).
The political orientation of the leadership of organized peasants
is an important determinant of land related violence. The political
leadership of the main group of landless peasants in Brazil, known as
Movimento dos Trabalhadores Sem Terra (MST) or the Landless Peasants
Movement, is a Marxist organization with ties to the Partido dos
Trabalhadores (PT) or Workers Party--the political party that has been
in power since 2003, first under President Luis Ignacio Lula da Silva
(2003--10) and now under President Dilma Rousseff. The MST homepage
(www.mst.org.br) makes it clear that among its political objectives is
the destruction of commercial agriculture in Brazil, one of the engines
of Brazilian growth. Land reform through violence is an essential part
of MST strategy to gain power, and is based on the Chinese and Cuban
ideology and revolutionary experiences.
In this article, we extend the research on land conflict in Brazil
by focusing on land occupation by farmers without land endowments. We
analyze poverty, political effects, population density (Andre and
Platteau 1998), and land endowment (Binswanger and McIntire 1987,
Sjaastad and Bromley 1997) to explain these land occupations. Earlier
studies have examined violence and land reform in Brazil (Alston,
Libecap, and Schneider 1995; Alston, Libecap, and Mueller 1997, 1999,
2005). However, our study is the first to be undertaken at a national
level using a contemporary data span. An additional innovation is the
use of a count data model that allows for heterogeneity, endogeneity,
and dynamics. Unobserved heterogeneity has been the subject of concern
and analysis in many previous studies (e.g., Chesher 1984, Chesher and
Santos-Silva 2002, McFadden and Train 2000). This type of model is used
frequently for data concerning events, and its omission is likely to
lead to inconsistent parameter estimates or, more importantly,
inconsistent fitted parameters. Endogeneity also yields estimation
problems causing biased results and may arise when a covariate is
simultaneously determined with the endogenous variable or when a
covariate is not inserted in the regression (Greene 2007). A dynamic
Poisson model is also presented with lags of endogenous variables and
leads of exogenous variables, enabling a more accurate analysis of the
problem (Cameron and Trivedi 1998).
We begin with a brief review of the literature on land conflicts.
We then describe the contextual background of land conflicts in Brazil
and present a stylized model that predicts how political, institutional,
and socioeconomic variables affect violent land occupation. Next, we
present our data, discuss methodological issues, and provide our
empirical findings. The policy prescriptions follow along with our
conclusions.
Literature Review
A major study on the violent invasion and occupation of land by
landless peasants in Brazil was conducted by Alston, Libecap, and
Mueller (2005). They describe how land invasions led by the Landless
Peasants Movement generated negative publicity for politicians,
stimulated broad sympathy of urban voters toward the landless peasants,
and led to further invasions. In another important study, Alston,
Libecap, and Mueller (2010) develop a multi-principal, multitask model
of interest group behavior to examine how groups with limited resources,
such as the Landless Peasants Movement, influence government by
manipulating media information to voters. They examine how the Landless
Peasants Movement in Brazil molds information, and study the reaction of
politicians in changing the timing and nature of policy. Meanwhile,
Alston and Mueller (2010) find that land conflicts reduce the likelihood
of tenancy, which results in a reduction in agricultural efficiency, a
welfare loss to potential renters, and an expansion of the agricultural
frontier through deforestation. Finally, a recent study by Oliveira
(2008) examines land conflicts and deforestation in the Amazon region
due to distorted agrarian, forest, and environmental policies, laws, and
regulations (see also Pacheco (2009), Ludewigs et al. (2009), and
Simmons et al. 2010).
Other factors such as population growth combined with limited
economic opportunities may lead to an increase in land invasion and
occupation, since they increase nonagricultural demand for land and
intensify competition for a limited or decreasing amount of land
available. This could also result in conflicts between groups,
particularly in environments where risk is high and land is a key asset
and source of livelihood (Andre and Platteau 1998). In contrast,
property rights and institutional frameworks that safeguard these rights
decrease land occupation (see, e.g., Alston, Libecap, and Schneider
1995, and Mueller 1997).
Land occupation has sometimes been considered a strategic policy in
less-developed countries to thwart far left-wing insurgency. In Brazil,
there is involvement of the current governing party, the Workers Party
or PT, and elements of the Catholic Church in land occupation (Simmons
et al. 2010). According to Ludewigs et al. (2009), land reform in Brazil
is a powerful tool in the struggle to reduce rural poverty mid may
attenuate environmental destruction, chiefly in the state of Amazonas
(Simmons et al. 2010, Pacheco 2009).
Aspects that characterize the violence related to land reform in
general and land occupation in particular are contextual variables such
as poverty (Waeterloos and Rutherford 2004, Rigg 2006), population
density (Simmons et al. 2010), and land productivity (Minten and Barrett
2008, Place 2009). Moreover, political forces, like the Brazilian
left-wing Workers Party, also shape the land reform process (Alston,
Libecap, and Mueller 1997, 1999, 2005). Institutional support for land
occupations is manifested in measures such as the expropriation of land
from the landowners by decree (Binswanger and Deininger 1993, Pacheco
2009). Other events, such as conflicts over water (Bakker et al. 2008),
actions of resistance to land occupation, demonstrations supporting land
occupation, generic conflicts related to land occupation, attempted
murders, and death threats are all part of the backdrop to land reform
and clearly may affect it.
Brazilian Land Reform and Violent Land Occupations
Brazil land tenure is characterized by large, family-owned
properties. A federal land reform agency, INCRA (Instituto National de
Colonizacao e Reforma Agraria/the National Agency for Land Reform and
Settlement) was established in 1969. The government allocated funds to
buy land and to redistribute it among poor families. Since its
inception, INCRA engaged in lengthy, bureaucratic processes of land
expropriation, with an average cost per beneficiary of $58,000. A
Federal Ministry of Agrarian Reform was created in 1996. Land
expropriation was expedited through the a priori selection of the land
by community groups, establishing an agreement on a
willing-seller/willing-buyer basis, paying the landowners in cash and
funding the endeavors of the new peasant-landholders. The expropriation
price per beneficiary decreased to $19,600. Grant financing is provided
for complementary and community infrastructures. The Central Institute
for Agrarian Studies was established to encourage discussions and
research on rural reforms. A recent survey of Brazilian land reform can
be seen in Simmons et al. (2010).
Land reform and rural conflict scenarios in Brazil have changed
significantly since the creation and growth of landless peasants and
rural workers interest groups, most notably file Landless Peasants
Movement (MST), which was officially founded ill 1984, on the departure
from power of the Brazilian military dictatorship. Alston, Libecap, and
Mueller (2005), using a principal-agent model, describe this transition
scenario.
Prior to the creation of the MST, organizations formed to promote
and defend the interests of landless peasants and rural workers were too
weak to be able to influence governmental land reform policies. At the
same time, the powerful landowners, in contrast, could afford to spend
both time and money on efforts to avoid expropriation, or to pursue
claims for generous compensation from the government in the event of
land reforms actually being enacted.
Since the emergence of the MST, combined with the spread of urban
voters sympathy for the ease of the landless rural workers, political
pressure for land reform has increased. Consequently, so have
land-related conflicts. In practice, this pressure is applied through
social insurgency (e.g., farm occupations, (1) marches, invasions of
government offices including INCRA, and roadblocks), combined with
accusations that the government is failing pledges or is dragging its
feet with regard to land reform implementation, always with the
objective of influencing public opinion by way of media coverage. The
MST's political influence through the media channels, as described
by Alston, Libecap, and Mueller (2005), is highlighted in Table 1.
Today a myriad of social movements of the landless peasants exists
in Brazil besides the MST, which is the largest and most important.
There are various splinter movements, such as the CPT (Pastoral Land
Commission), the MLT (the Straggle for Land Movement), the MLTS
(Syndicate of Small Rural Producers), the OTC (the Rural Workers
Organization), and the STR (local Rural Workers Unions), all adhering to
left-wing revolutionary ideology and coupled with regional
organizations, such as the Movement of Corumbiara Peasants (Simmons et
al. 2010).
The aggressive occupation of private and public lands is justified
on the basis of the moral authority and constitutional right of the
landless peasants (Stedile 1997, Wolford 2004). There is evidence that
while some original settlers remain on the acquired land and develop
villages (Ludewigs et al. 2009), others, after fragmenting and selling
off their plots initially received from INCRA, migrate to new
agricultural frontiers in order to start the process once again (Alston,
Libecap, and Muller 1999), or to urban centers (Bowder and Godfrey
1997).
A distinctive aspect of the Brazilian land occupation political
movement is the role played by the Workers Party, as well as by the left
wing of the Catholic Church.
Table 2 presents the characteristics of the Brazilian states in
relation to the problem of land occupation. It shows that land
occupation varies throughout the country, with northern rural states
more prone to land occupation and related activities than southern
states.
A Model of Land Invasion and Occupation
In this section, we present a highly stylized model. The objective
is to determine the role, if may, of the factors identified by the
empirical literature on land conflicts in increasing or reducing land
invasion. Our model provides a platform to hypothesis testing; it is not
an end in itself. There are useful models available in the literature
one could build on to model the same issue, such as the conflict models
of Grossman (1991) and Blomberg et al. (2004) in which peace is
disturbed by disgruntled groups seeking to increase their voice, power,
and control, or the game-theoretic models developed by Addison, Le
Billon, and Mushed (2002) in which collective action is highlighted.
Our intertemporal dynamic model with a landless peasant as a
representative agent allows one to study the dynamic path of land
invasion and landless peasant consumption. We assume the peasant seeks
to maximize his welfare over time subject to the dynamics of land
invasion and occupation, which are related to the collective action of
the Landless Peasants Movement. This is why the MST is exogenous from
the viewpoint of an individual peasant who may or may not choose to join
the movement. The framework of the model is in line with Levy and
Faria's (2007) conflict model.
The solution of the model clearly shows a rationale for the peasant
to join the group that organizes land invasion and occupation, since in
the optimal path, given the initial conditions, the peasant may have an
extraordinary increase in welfare by joining the group.
The dynamics of land invasion and occupation depend on the
differences between forces favorable to invasion (given by function F)
and against invasion (given by function G). As arguments of function F
we have: (1) the expected return of occupied land, (2) political support
for land occupation, (3) conflicts over common resources, (4)
agricultural credit funded by a public agency, and (5) peasant poverty
(m).
Regarding the expected return of occupied land, if the peasant
thinks that occupied land (L) yields an output f(L) that is enough for
his current welfare, as captured by a desired consumption level
(c)--that is, if f(L) - c > O--then this can lead him to join the
movement and invade and occupy privately owned lands. If f(L) = c, then
the argument vanishes from function F and the expected return of
occupied land has no role on the forces for land invasion and
occupation.
In the same vein, every type of political support for land reform,
represented by p, such as governments ruled by left-wing political
parties, active land reform carried out by the government, organized
political movements, and organized political violence for land reform,
increase function F, since the peasant feels more confident in invading
because the risk of punishment is smaller.
Conflicts over the management of common resources, represented by
a, such as water allocation, can spill over and become a political
issue, increasing calls for land reform and making the peasant more
likely to invade.
Agricultural credit funded by a public agency, A, stimulates land
invasion because the provision of credit is attached to the land
obtained either by legal land distribution or by land obtained through
organized and politically motivated land invasion. Peasant poverty, m,
is usually associated with more land invasions because it lowers file
opportunity cost of joining a rebellion.
Given the above discussion, function F has the following
characteristics:
F(f(L) -- c,p,a,A, m), [F.sub.1] > 0, [F.sub.2] > O,
[F.sub.3] > O, [F.sub.4] > O, [F.sub.5] > O, and [F.sub.11] =
0, [F.sub.12] > 0, [F.sub.13] > 0, [F.sub.14] > 0, [F.sub.15]
> 0; that is, political participation, conflicts over common
resources, agricultural credit funded by a public agency, and peasant
poverty increase the marginal impact of the expected return of occupied
land on F.
The forces against land invasion and occupation are captured by
function G. The arguments of function G are: (1) agricultural
production, wL, where w is the unit value of production per unit of
land, L; (2) peasant poverty, m; (3) probability of successful land
occupation, v; and (4) rural population density, n.
Agricultural production reduces land invasion and occupation since
it uses land and employs labor, hence increasing the opportunity cost of
land invasion for a landless peasant. Belated to agricultural production
is the issue of land eviction of peasants by landowners. In order to
decrease the risk of land conflicts landowners can reduce the demand for
labor by mechanizing agriculture and/or adopting extensive large-scale
livestock production. Of course, this is a burdensome growth process for
the peasant population in which more evictions lead to more poverty.
According to this line of reasoning, productive privately owned land is
more difficult to be invaded and occupied, so at least locally, poverty
of the peasant population may be associated with less land invasion and
occupation. An additional explanation is to assume that poor peasants
may prefer to look for federal assistance programs, including food
assistance, rather than engage in conflicts in order to occupy privately
owned land.
Uncertainty of land occupation is an important factor influencing
the decision to invade land. If the probability of successful land
invasion and occupation, v, is high, the peasant decides to join a
peasants movement and invade land; otherwise he will not join and invade
land. For instance, if the government reacts to land invasion enforcing
the rule of law and preserving the landowners property rights, then the
landless peasant has less incentive to invade. In the same vein, if
rural landowners fight invasions back, by reinforcing the protection of
their properties, this may increase the uncertainty and reduce land
invasions.
Given the above discussion, the function G has the following
characteristics:
G(wL, m, v, n,), [G.sub.1] > 0, [G.sub.2] > 0, [G.sub.3] >
0, [G.sub.4] < 0, and [G.sub.11] > 0, [G.sub.12] > 0,
[G.sub.13] > 0, [G.sub.14] < 0; that is, function G is convex in
agricultural production, rural population density reduce the marginal
impact of the agricultural production on G, and poverty and probability
of land occupation increase it.
It is important to stress that poverty (m) is an argument in both
functions F and G. Its role is highly complex because there are reasons
to think that it has a positive impact on land invasion and occupation
and, at the same time, a negative impact. Although its final impact
cannot be a priori theoretically determined, being clearly an empirical
issue, we hypothesize that the negative impact prevails.
The landless peasant problem is:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
(1) s.t. [??] = F(f(L) -- c, p, a, A, m) -- G(wL, m, v, n),
where the landless peasant preferences for consumption over time
are represented by the utility integral, the instantaneous utility
function U(c) is nonnegative and a concave increasing function of the
consumption, c, [theta] is the landless peasant rate of time preference,
and f(L) is a production function, where labor is supplied
inelastically. The current value Hamiltonian associated with the problem
is:
H = U(c) + [lambda] {F(f(L) -- c, p, a, A) -- G(wL, m, n, v)},
where [lambda] is the costate variable, the shadow price of land
invasion and occupation, L, for the landless peasant. The first order
conditions are:
(2) [U.sub.c] (c) - [lambda] [f.sub.1](f(L) - c, p, a, A) = 0
(3)) [??] - [theta] [lambda] = -[lambda][[F.sub.1] (f(L) - c, p, a,
A) [f.sub.L] (L) - w[G.sub.1] (wL, m, v, n)] plus the transversality
condition:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
Differentiating equation (2) with respect to time and using
equations (2) and (3) yields a differential equation describing the
evolution over time of consumption:
(4) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
Equation (4) presents the Keynes-Ramsey rule for the landless
peasant model. In a typical Ramsey model without population growth the
marginal rate of transformation from production corresponds to the
difference between the rate of time preference and the marginal product
of capital (Blanchard and Fischer 1989). In this model there is no
capital, and the marginal rate of transformation from production is
given by the difference between the rate of time preference [theta], mad
the marginal expected return of invaded land (the term [F.sub.1]
[f.sub.L]) plus the forgone income if land is invaded (the term wG1).
Given that [U.sub.c] > 0, [U.sub.cc] < 0, it follows from equation
(4) that consumption grows with the marginal expected return of invaded
land, and decreases with the rate of time preference and the forgone
income if land is invaded.
The determinant of the Jacobian (det J) of the dynamic system
formed by the differential equations (1) and (4) is equal to:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.]
And the modal displays saddle point stability if det J < 0,
which holds if [F.sub.1] [f.sub.LL] < [w.sup.2] [G.sub.11].
In the steady state: [??] = 0 = Lin equations (1) and (4), we have:
(5) F(f(L) - c, p, a, A, m) = G(wL, m, v, n)
(6) [F.sub.1] (f(L) - c, p, a, A, m) [f.sub.L] (L) = [theta] +
w[G.sub.1] (wL, m, v, n).
Equations (5) and (6) determine simultaneously the steady state
equilibrium values of consumption, [c.sup.[??]], and land invasion and
occupation, [L.sup.[??]]. With [c.sup.[??]] and [L.sup.[??]], then
equation (3) determines [lambda], the shadow price of land invasion and
occupation. By using explicit functions for F, G and U, we can obtain
explicit expressions for [c.sup.[??]] and [c.sup.[??]].
The dynamic system formed by equations (1) mad (4) yields the
optimal path of landless peasant consumption and land invasion
converging towards [c.sup.[??]] mad [L.sup.[??]]. If the initial level
of land invaded, [L.sub.0] is below [L.sup.[??]], land invasion grows
towards [L.sup.[??]], and for a given level of initial consumption
[c.sub.o] associated with [L.sub.o] we have an increase in the landless
peasant consumption. So this optimal path shows that the landless
peasant has a lot to gain by joining the movement of land invasion and
occupation, since his welfare increases.
Our focus, however, is to investigate the impact of poverty (m),
agricultural productivity (w), population density (n), political support
for land occupation (p), probability of successful land invasion and
occupation (v), water conflict (a), and agricultural credit (A) on
equilibrium land invasion and occupation ([L.sup.[??]]). The comparative
statics analysis of the system (5) and (6) yields the following
multipliers:
(7) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
(8) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
(9) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
(10) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
(11) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
(12) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
(13) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
According to the comparative statics analysis [(7)--(13)],
equilibrium land invasion and occupation ([L.sup.[??]) increases with
population density (n), political support for land occupation (p), water
conflict (a), and agricultural credit funded by a public agency (A).
Land invasion and occupations decreases with agricultural productivity
(w) and uncertainty (v). The impact of poverty (m) on land invasion and
occupation is, a priori, ambiguous: it may increase or decrease it. In
the following empirical part of this article, we test these predictions
of the model.
Methodological Framework and Hypotheses
The following variables are considered in the empirical estimations
of our model: percentage of poverty in the state, agricultural
productivity, population density, state governed by the Workers Party,
lands allocated by decree, agricultural credit, water conflicts,
resistance events, demonstrations, murder attempts, number of land
conflicts, number of murders related to land reform, and the number of
death threats related to land reform. The panel data by Brazilian state
is for the period 2000--08. Table 3 presents the characteristics of the
data used.
The number of land occupation actions is of paramount importance
for its strategic management. Our analysis includes a count data model
to identify statistical significant covariates in the sample (Greene
2005). The research utilizes data from the Comissao Pastoral da Terra, a
Catholic organization that supports poor peasants and landless farmers
(www.cptnac.com.br/?system=news&eid=6), supplemented with additional
variables from other sources (www.ipeadata.gov.br).
According to file theoretical model, land occupation depends on the
following factors:
H1 (PT): States governed by the left-wing Partido dos Trabalhadores
(PT), the Workers Party, increase land occupation. This effect is based
on ideological preferences by left-wing parties for equity over
efficiency, inducing the occupation of the lands owned by absentee
landlords (Alston, Libecap, and Mueller 1997, 1999, 2005; Araujo Jr,
Shikida, and Alvarenga 2008).
H2 (INCBA): Lands allocated for reform by decree by INCRA, tend to
validate a posteriori land invasion and occupation, fueling further land
occupations. This problem has been analyzed by Binswanger and Deininger
(1993) and Pacheco (2009).
H3 (PRONAF): Agricultural credit funded by Programa Nacional de
Fortalecimento da Agricultura Familiar/National Program to Strengthen
Family Agriculture (PRONAF), a program of the Ministry of Agricultural
Development that finances individual projects or groups of farmers and
agrarian reform settlers, tends to support land occupation, being with
hypothesis 4 and 5 part of the political process that the left-wing
government has established to promote land occupation.
H4 (Poverty): Poverty may increase or decrease land occupation,
this variable is often used to study land conflicts (Waeterloos and
Rutherford 2004, Finan et al. 2005, Rigg 2006).
H5 (Agricultural Productivity): Agricultural productivity generates
more wealth and food supply in the state and therefore it decreases land
invasion and occupation. It is often used in land reform models
(Caviglia-Harris 2003, Minten and Barrett 2008, Place 2009).
H6 (Population Density): Rural population density increases land
occupation. It is frequently use in land reform models (Simmons et al.
2010).
H7 (Water Conflicts): Conflicts over water management, such as
those arising from severe shortages in semi-arid states in northeastern
Brazil, increase land occupation (Bakker et al. 2008).
H8 (Resistance Actions and Demonstrations): Resistance actions and
demonstrations are indicators of political determination to invade and
occupy land, which help reduce uncertainty over the success of land
invasion. Although these variables have not been used in previous
research of land reform, they are intrinsically part of the process of
land occupation.
H9 (Murder Attempt and Death Threats): The number of individual
land conflicts, such as murder attempts and death threats, are generic
aspects of individual aspects of land occupation, usually occurring
prior to the land occupation and derived from land evaluation by the
occupiers (Peters 2009), and serve to decrease land occupations since
they increase the uncertainty of successful land invasion.
In order to test the foregoing hypotheses, we first estimate an OLS
model that serves as a reference for the Poisson model. The motivation
to use the Poisson model is derived from the fact that the dependent
variable is the number of times the Landless Peasants Movement invades
privately owned land (Cameron and Trivedi 1998). This variable is a
counting variable that is characterized as being nonnegative, which
should be modeled as a Poisson or a negative binomial model (Greene
2005).
The Poisson model is based on the hypothesis that the endogenous
variable [y.sub.i] (counts of the number of land invasions), given the
covariates x, is independent with the Poisson distribution and
probability density function.
It is currently assumed that the basic Poisson model is too
restrictive with regard to the features of the observed data (Cameron
and Trivedi 1998). Common deviations from the basic Poisson model are
endogeneity, dynamic nature of data, and overdispersion. Endogeneity
occurs when there is a correlation between the exogenous variable and
the error term. It can also arise as a result of measurement error,
simultaneity, omitted variables, and sample selection errors. Another
cause of endogeneity lies on the dynamic nature of the data and with
autocorrelated errors, justifying the dynamic analysis of the data
(Greene 2005).
Overdispersion is the failure to satisfy the conditional mean and
conditional variance restriction. If the conditional variance of the
data exceeds the conditional mean, overdispersion is present. The most
commonly given explanation for overdispersion is file unobserved
heterogeneity in the data--that is, there are omitted variables in the
mean function. Other explanations are measurement errors in explanatory
variables and the stochastic character of the structural parameters. A
common approach to overcoming this problem is to estimate Poisson models
allowing for heterogeneity in the mean (Greene 2005), or to estimate a
random Poisson model.
Our empirical strategy follows these procedures. We first estimate
the OLS model, followed by estimates of a standard and fixed effects
Poisson model (Hilbe 2008). Next, we estimate a Poisson model with
heterogeneity and a Poisson model with endogeneity (Mullahy 1997).
Finally, we estimate a dynamic Poisson model to investigate lags of
endogenous variables and leads of exogenous variables.
Results and Discussion
Table 4 presents the results. Stata software was used to estimate
the model. The first model is the OLS presented as a reference. The
second column presents a standard Poisson model. The third model is the
Fixed effects Poisson model. The fourth model is the random Poisson
model allowing for heterogeneity. The fifth model is the endogenous
Poisson model allowing for endogeneity in the political variables,
namely PT, INCRA, PRONAF, poverty, and productivity. The Generalized
Method of Moments (GMM) estimator of Poisson regression is adopted
allowing for endogenous variables to be instrumented by excluded
instruments. Standard errors are estimated by bootstrapping. Prior to
estimating the endogenous mode, we performed a Hausman test that reached
the value of 28.16 with a small p-value, signifying that there was
endogeneity in the variables. We then ran the model without each
possible endogenous variable such as possible political induced land
invasion (PT, INCRA, P1RONAF) and contextual induced land invasion
(poverty, agricultural productivity, and population density) and
concluded that the endogenous variables are PT, INCRA, and PRONAF. The
instrumental variables were adopted using the percentage of state PT
municipal heads in the state total, total agricultural bank credit in
the state, and each variable lag.
The immediate thing we observe in the results is that almost all
signs of variables are maintained throughout the various models. Based
on the loglikelihood, the dynamic Poisson is chosen. In the dynamic
Poisson model, it is verified that lags in the endogenous variable are
statistically significant, meaning that this is a process with
persistence. The dummy variable PT, which indicates the left-wing party
controls the state, is positive and significant, which means that it
contributes to land occupation. (2) INCBA is also positive, which
signifies that number of lands taken from landowners by decree induce
land occupation. Agricultural credit also increases land occupation, but
this variable is not endogenous. The fact that agricultural credit is
not endogenous means that it contributes to the process through market
dynamics, after the property rights are allocated to the occupant.
The first conclusion is that land occupation in Brazil is explained
by political variables such as political support by the party in power,
institutional support by public entities through INCRA that expropriate
the occupied land from the landowners by decree, and agricultural credit
by a public agency. Note that INCRA and PRONAF have statistically
significant leads, which add to the statistically significant lag of the
endogenous variable--meaning this is an ongoing process with
persistence. These results validate Hypotheses 1, 2, and 3.
The second result is that poverty is negative and statistically
significant in several models. However, there are two models in which
poverty is positive but statistically insignificant. Poverty is positive
but statistically insignificant in the fixed effects model. The fixed
effects model displays a weak fit, meaning there are no Fixed effects in
the data. A possible explanation for this is that invasions are
concentrated in a small number of Brazilian states. Poverty is also
positive and statistically insignificant in the OLS model. In all other
models, poverty is negative and statistically significant with the
exception of the dynamic Poisson model. Although the literature
considers poverty an important covariate of land invasion, the results
of all our models convinced us that poverty is a not an important
covariate, and based on the chosen model it is not statistically
significant. Thus, it is not addressed in our policy prescriptions.
The third result is that productivity in agriculture decreases land
occupation, validating Hypothesis 5, that is, high agricultural
productivity increases peasants income and therefore decreases the need
for land ownership through land occupation. Population density increases
land occupation, validating Hypothesis 6 and suggesting that it is among
the main causes of land reforms in developing countries (Hidalgo et al.
2010).
Conflicts over water management are not statistically significant
and decrease land occupation, which does not confirm Hypothesis 7.
Resistance and demonstrations increase land occupation, validating
Hypothesis 8. Finally, murder attempts and death threats have a mixed
effect on land occupation, not validating Hypothesis 9, and implying
that fear and uncertainty about the success of land invasion has a mixed
effect on land conflicts.
Policy Prescription
The policy implication of this research is that file government
should use its institutional framework to minimize conflicts and conduct
land reform where it is necessary under file strict rule of the law. The
government has to reduce the power of peasants political movements that
feed on violence by not granting property rights to invaders of
privately owned land. The land reform has to reinforce property rights
rather than oppose them. For example, Brazil's Land Statute
(Federal Law, Nr. 4.504, November 1964) prevents the expropriation by
the government of lands that have suffered invasion. The role of law is
of paramount importance. If the government, for political reasons, aims
at maintaining political support from these peasants movements with a
clear revolutionary agenda, like the MST, and, at the same time, does
not reinforce property rights and the rule of law, it sends a mixed
message that fuels violence, (3)
According to our results, the Landless Peasants Movement focuses on
regions with less poverty and greater population density to invade and
occupy land. This of course may disrupt production in the main
agricultural regions of Brazil. The government has to preserve the
highly productive areas from disruption caused by land invasions. The
protection of highly productive areas is explained by its economic
importance. Commercial agriculture in Brazil evolved over the past three
decades from being extensive to intensive--that is, production per
hectare has increased significantly (IBGE 2011).
As a result Brazil has become not only one of the world's
largest producers of many important agricultural commodities but also a
major exporter. For instance, Brazil is the largest producer and
exporter of sugar, coffee, and orange juice; it is the second largest
producer and largest exporter of ethanol, beef, tobacco, and chicken. It
is one of the five largest exporters of soybeans, leather and fur,
soybean oil, corn, pork meat, and cotton (MDIC 2011). Therefore, any
changes in the land tenure of these highly productive areas will have
enormous costs in terms of forgone income and employment of capital and
labor, disrupting how this modern agricultural sector relates with the
rest of the economy.
There are several ways to minimize violence related to the
equity-efficiency trade-off in this case. The government can assist the
landless peasants without violating the farmers property rights by
providing land located in appropriate areas--that is, low productivity
areas or unproductive and publicly owned land. This would minimize the
disruption of the commercially productive agricultural sector and, at
the same time, satisfy the demand for land ownership from landless
peasants. If the MST leadership refuses this alternative and exerts
pressure to invade productive lands, it shows its political character
aiming at imposing serious economic damage on the modern agricultural
sector without properly addressing the needs of landless peasants. The
government should aim at satisfying individuals that want to have access
to land, not the MST's political agenda.
How does the present research compare with previous research on
Brazilian land reform? It provides a specific and unique insight into
the land reform movement, based on the actions of the Landless Peasants
Movement, focusing on contextual and conflict variables, giving a clear
view of the factors that influence violence. Compared with earlier
research on Brazilian land reform, this article's use of data for
the entire country and estimation methods that take account of
heterogeneity, endogeneity, and dynamics enables a more accurate view of
the land reform movement.
Conclusion
This article analyzes land occupation and land reform in Brazil for
the period 2000--08. It is the first study to be undertaken at a
national level, with a contemporary data span, using a count data model
that 'allows for heterogeneity, endogeneity, and dynamics. It
studies contextual variables that affect land occupation, such as
political, institutional, conflict, and socioeconomic variables. It
shows that political and institutional variables have a positive effect
on land occupation. However, the socio-economic variables have a mixed
effect, with rural population density increasing land occupations, while
poverty and land productivity reduce land occupations. Conflict
variables also have mixed effects, with resistance movements,
demonstrations, and generic conflicts increasing land invasion and
occupation, while water-related conflicts, murder attempts, and death
threats cause their decrease.
In order to minimize conflict, we recommend that the government
minimize land invasion and occupation and follow the role of law. We
also suggest that the government should focus on the needs of
individuals that want to have access to land, not on the MST's
political agenda. In this regard, the government should provide landless
peasants with land located in low productivity areas or unproductive and
publicly owned land. Brazil's land reform should reinforce private
property rights, not destroy them.
References
Addison, T.; Le Billon, P.; and Murshed, S. M. (2002)
"Conflict in Africa: The Cost of Peaceful Behaviour." Journal
of African Economies 11 (3): 36,5-86.
Agencia Estado (2011) "Dilma Reduz Verba de Reforma Agraria,
Dizem PT e MST." Available at
www.em.com.br/app/noticia/politica/2011/12/02/interna_politica,265288/dilma-reduz-verbade-reforma-agraria-dizem-pt-e-mst.shtml.
Alston, L. J.; Libecap, G. D.; and Mueller, B. (1997)
"Violence and the Development of Property Bights to Land in the
Brazilian Amazon." In J. V. C. Nye and J. N. Drobak (eds.)
Frontiers on the New Institutional Economics'. San Diego: Academic
Press.
--(1999) "A Model of Rural Conflict: Violence and Land Reform
Policy in Brazil." Environment and Development Economics 4:
13,5-60.
--(2005) "How Interest Groups with Limited Resources Can
Influence Political Outcomes: Information Control and Landless Peasants
Movement in Brazil." Institute of Behavioral Science Working Paper
No. 2005-0005.
--(2010) "Interest Groups, Information Manipulation in the
Media, and Public Policy: The Case of the Landless Peasants Movement in
Brazil." NBER Working Paper No. 15865. Alston, L. J.; Libecap, G.
D.; and Schneider, R. (1995) "Property
Bights and the Precondition for Markets: The Case of the Amazon
Frontier." Journal of Institutional and Theoretical Economies 151
(1): 89-107.
Andre, C., and Platteau, J. P. (1998) "Land Relations under
Unbearable Stress: Rwanda Caught in the Malthusian Trap." Journal
of Economic Behaviour and Organisation 34 (1): 1-47.
Aranjo Jr., A.; Shikida, C. D.; and Alvarenga, P. S. (2008)
"Economia Politica da Disputa por Terras em Minas Gerais."
Bevista de Economia e Sociologia Rural 46 (3): 803-30.
Assuncao, J. J. (2008) "Rural Organization and Land Reform in
Brazil: The Role of Nonagricultural Benefits of Landholding."
Economic Development and Cultural Change 56 (4): 851-70.
Bakker, K; Kooy, M.; Shofiani, N. E.; and Martijn, E. J. (2008)
"Governance Failure: Rethinking the Institutional Dimensions of
Urban Water Supply to Poor Households." World Development 36 (10):
1891-1915.
Binswanger, H. P., and Deininger, K. (1993). "South African
Land Policy: The Legacy of History and Current Options." World
Development 21 (9): 1451-75.
Binswanger, H. P.; Deininger, K.; and Feder, G. (1995) "Power,
Distortions, Revolt and Reform in Agricultural Relations." In J.
Behrman and T. N. Srinivasan (eds.) Handbook of Development Economics.
Vol. 3. Amsterdam: Elsevier.
Binswanger, H. P., and McIntire, J. (1987) "Behavioral and
Material Determinants of Production Relations in Land-abundant Tropical
Agriculture." Economic Development and Cultural Change 36 (1):
73-79.
Blanchard, O., and Fischer, S. (1989) Lectures on Macroeconomics.
Cambridge, Mass: MIT Press.
Blomberg, S. B.; Hess, G. D.; and Weerapana, A. (2004) "An
Economic Model of Terrorism." Conflict Management and Peace Science
21: 17-28.
Bowder, J. o., and Godfrey, B. J. (1997) Bainforest Cities:
Urbanization Development and Globalization of the Brazilian Amazon. New
York: Columbia University Press.
Bruck, T., and Schindler, K. ('2009) "Smallholder Land
Access in Postwar Northern Mozambique." Worm Development 37 (8):
1379-89.
Cameron, A. C., and Trivedi, P. K. (1998) Regression Analysis of
Count Data. Cambridge: Cambridge University Press.
Caviglia-Harris, J. L. (2003) "Sustainable Agricultural
Practices in Rondonia, Brazil: Do Local Farmer Organizations Affect
Adoption Rates?" Economic Development and Cultural Change 52 (1):
23-49.
Chesher, A. D. (1984) "Testing for Neglected
Heterogeneity." Econometrica 52: 865-72.
Chesher, A. D., and Santos Silva, J. M. C. (2002) "Taste
Variation in Discrete Choice Models." Review of Economic Studies
69: 147-68.
Deininger, K., and Castagnini, R. (2004) "Incidence and Impact
of Land Conflict in Uganda." World Bank Policy Research Working
Paper No. 3248.
De Janvry, A., and Sadoulet, E. (1989) "A Study in Resistance
to Institutional Change: The Lost Ganae of Latin American Land
Reform." World Development 17: 1397-1407.
Estado de Sao Paulo (2009) "MST Destroi 7.000 Pes de Laranja
da Cutrale." Available at
www.estadao.com.br/estadaodehoje/20091007/not_imp447136,0.php (10 July).
Fenske, J. (2011) "Land Tenure and Investment Incentives:
Evidence from West Africa." Journal of Development Economics 9 (2):
137-56.
Finan, F.; Sadoulet, E.; and de Janvry, A. (2005) "Measuring
the Poverty Reduction Potential of Land in Rural Mexico." Journal
of Development Economics 77 (1): 27-51.
Greene, W. (2005) Econometric Analysis. Upper Saddle River, N.J.:
Prentice Hall.
--(2007) "Correlation in the Bivariate Poisson Regression
Model." Working Paper, Department of Economies, Stem School of
Business, New York University.
Grossman, H. I. (1991) "A General Equilibrium Model of
Insurrections." American Economic Review 81: 912-21.
Hidalgo, F. D.; Naidu, S.; Nichter, S.; and Richardson, N. (2010).
"Economic Determinants of Land Invasion." Review of Economics
and Statistics 92 (3): 505-23.
Hilbe, J. M. (2008) Negative Binomial Model. Cambridge: Cambridge
University Press.
IBGE (2011) "IndicadoresAgropeeucarios." Available at
www.ibge.gov.br/home/estatistica/indicadores/agropecuaria/lspa/lspa_201112comentarios.pdf.
Levy, A., and Faria, J. R. (2007) "Ramsey in Dual-Population
Lands: Internal Conflict and Utility-maximizing Consumption."
Defence and Peace Economics 18: 339-52.
Ludewigs, T.; D'Antona, A. O.; Brondizio, E. S.; and Hetrick,
S. (2009) "Agrarian Structure and Land-Cover Changes along the
Lifespan of Three Colonization Areas in the Brazil Amazon." World
Development 37 (8): 1348--,59.
McFadden, D., and Train, K. (2000) "'Mixed MNL Models of
Discrete Response."Journal of Applied Econometrics 15: 447-70.
MDIC (2011) "Knowing Brazil in Numbers." Available at
www.mdic.gov.br/arquivos/dwnl_1312203713.pdf.
Minten, B., and Barrett, C. B. (2008) "Agricultural
Technology, Productivity, and Poverty in Madagascar." World
Development 36 (5): 797-822.
Mueller, B. (1997) "Property Rights and the Evolution of a
Frontier." Land Economics 73: 42-,57.
Mullahy, J. (1997)"Instrumental-Variable Estimation of Count
Data Models: Applications to Models of Cigarette Smoking Behavior."
Review of Economics and Statistics 79 (4): 586-93.
Oliveira, J. A. P. (2008) "Property Rights, Land Conflicts and
Deforestation in the Eastern Amazon." Forest Policy and Economics
10 (5): 303-15.
Pacheco, P. (2009) "Agrarian Reform in the Brazilian Amazon:
Its Implication for Land Distribution and Deforestation." World
Development 37 (8): 1337-47.
Peters, P. E. (2009) "'Challenges in Land Tenure and Land
Reform in Africa: Anthropological Contributions.'" World
Development 37 (8): 1317-25.
Pichon, F. J. (1997) "Colonist Land-Allocation Decisions, Land
Use, and Deforestation in the Ecuadorian Amazon Frontier." Economic
Development and Cultural Change 45 (4): 707-44.
Place, F. (2009) "Land Tenure and Agricultural Productivity in
Africa: A Comparative Analysis of the Economics Literature and Recent
Policy Strategies and Reforms." World Development 37 (8): 1326-36.
Rigg, J. (2006) "Land, Farming, Livelihoods, and Poverty:
Rethinking the Links in the Rural South." World Development 34 (1):
180-202.
Simmons, C.; Walter; R.; Perz, S.; Aldrich, S.; Caldas, M.;
Pereira, R.; Leite, F.; Fernandes, L. C.; and Arima, E. (2010)
"Doing It for Themselves: Direct Action Land Reform in the
Brazilian Amazon." World Development 38 (3): 429-44.
Sjaastad, E., and Bromley, D. W. (1997)"Indigenous Land Rights
in Sub-Saharan Africa: Approriation, Security and Investment
Demand." World Development 25 (4): 549-62.
Stedile, J. P. (1997) A Reforma Agraria e a Luta do MST.
Petropolis, RJ: Vozes.
Waeterloos, E., and Rutherford, B. (9.004) "Land Reform in
Zimbabwe: Challenges and Opportunities for Poverty Reduction among
Commercial Farm Workers." World Development 32 (3): 537-53.
Wolford, W. (2004) "This Land Is Ours: Spatial Imageries and
the Struggle for Land in Brazil." Annals of the Association of
American Geographers 94 (2): 409-24.
(1) See, for example, Estado de Sao Paulo (2009). Members of the
MST destroyed 7,000 orange trees on a private estate.
(2) In contrast with President Roussef's policies, the
agrarian policy committee of PT demands increasing expenditures with
land reform. PT internal politics demands the strengthening of social
movements like MST, which tends to promote more land conflicts (Agencia
Estado 2011). However, it is important to note that our research covers
the period 2000-08.
(3) Our policy prescription does not consider a political
equilibrium. In Alston, Libecap, and Mueller (2010), the political
equilibrium is that the government will concede to the MST given its
ability, to mold information available to urban voters so as to generate
broad urban support for its land reform agenda.
Carlos Pestana Barros is Associate Professor at ISEG and UECE,
Technical University of Lisbon, Portugal. Ari Francisco de Araujo Jr. is
Assistant Professor of Economies at IBMEC Minas Gerais, Brazil, mad joao
Bicardo Faria is Professor of Economies at the University of Texas at El
Paso.
TABLE 1
THE LANDLESS PEASANTS MOVEMENT (MST)
AND THE MEDIA
Characteristics Summary
Low marginal cost to Regardless of voters sympathy, the MST has
influence the news the "technological" means to obtain exten
sive media coverage for its activities. One
of the factors that favorably affect this is
that the invasions and occupations are
very labor-intensive, which increases their
visibility. Another factor is the low oppor
tunity cost of landless peasants ("peasant
army reserve"), compared with the oppor
tunity cost of farmers.
Productive efforts to In practice, the MST receives more benign
influence the news visibility than farmers. Farmers have not
on land reform achieved the same influence in the
issues political arena.
Extreme configuration Voters are favorable to land reform, and this
of voters preferences is, empirically speaking, a variable that
increases the popularity of the president.
SOURCE: Adapted from Araujo Jr., Shildda, and Alvarenga (2008).
TABLE 2
CHARACTERIZATION OF BRAZILIAN LANDLESS
MOVEMENT IN 2008
% Number of
population lands taken
Number below from
of lands poverty landlords
State occupied line by decree Population
Acre 0 0.224 6 680,073
Alagoas 20 0.212 5 3,127,557
Amapd 0 0.109 0 613,164
Amazonas 1 0.165 0 3,341,096
Bahia 28 0.151 22 14,502,575
Ceard 4 0.186 12 8,450,527
Distrito Federal 0 0.000 16 2,557,158
Espfrito Santo 2 0.034 2 3,453,648
Goids 10 0.034 21 5,844,996
Maranhao 1 0.228 11 6,305,539
Mato Grosso 1 0.042 6 2,957,732
Mato Grosso 6 0.028 2 2,336,058
do Sul
Minas Gerais 10 0.032 15 19,850,072
Pard 17 0.104 12 7,321,493
Parafba 10 0.165 25 3,742,606
Parand 14 0.036 1 10,590,169
Pernambuco 52 0.172 15 8,734,194
Piaui 0 0.204 27 3,119,697
Rio de Janeiro 2 0.040 2 15,872,362
Rio Grande 1 0.142 5 3,106,430
do Norte
Rio Grande 11 0.042 13 10,855,214
do Sul
Rond6nia 3 0.096 0 1,493,566
Roraima 1 0.117 0 412,783
Santa Catarina 5 0.013 4 6,052,587
Sao Paulo 49 0.029 4 41,011,635
Sergipe 3 0.142 10 1,999,374
Tocantins 1 0.120 8 1,280,509
Mean 9 0.106 9 7,022,696.82
Std. deviation 13 0.073 8 8,422,434.61
TABLE 3
DATA DESCRIPTIVE ANALYSIS
Variables Description Mean
Invasion and Number of Occupations 12.736
Occupation
Political and Institutional Context
PT Dummy equal to one 0.127
if PT governs
the state
INCRA Number of lands 10.13
taken from
landowners by decree
PRONAF Agricultural credit 320,939,801
real value, 2008
Economic and Social Context
Poverty % of population 0.147
below poverty line
Agricultural Agricultural 1.093
production by
hectare
Productivity
Population Density Population density 63.166
by state
Conflicts
Water Conflicts Number of conflicts 1.456
Resistance Actions Number of resistance 0.378
actions per state
related to land
reforms
Demonstrations Number of 25
demonstrations per
state related to
land reforms
Individual Land Conflict
Murder Attempt Number of murder 2.617
attempts per state
related to land
reforms
Death Threats Number of menaces 8.530
per state related to
land reform
Variables Std. Dev. Min. Max.
Invasion and 19.885 0 165
Occupation
Political and Institutional Context
PT 0.334 0 1
INCRA 12.16 0 78
PRONAF 1,284,035,714 4,576 19,427,486,432
Economic and Social Context
Poverty 0.005 0 0.409
Agricultural 0.526 0.175 3.534
Productivity
Population Density 94.914 0.038 440.744
Water Conflicts 2.478 0 20
Resistance Actions 1.050 0 7
Demonstrations 20 0 118
Murder Attempt 5.422 0 32
Death Threats 18.725 0 118
TABLE 4
EMPIRICAL RESULTS
DEPENDENT VARIABLE: NUMBER OF LAND INVASIONS AND OCCUPATIONS
Standard Poisson with
Variable OLS Poisson Model Fixed Effects
Constant -38.661 -2.221 --
(-2.59) (13.82)
[Occupation.sub.T-1] -- -- --
[Occupation.sub.t-2] -- -- --
[Occupation.sub.t-3] -- -- --
PT 4.654 0.230 0.418
(2.16) (3.63) (2.03)
INCRA 0.388 0.016 0.0046
(2.69) (13.80) (2.02)
[INCRA.sub.t-1] -- -- --
LogPRONAF 1.416 0.731 0.523
(2.54) (10.25) (3.218)
[LogPRONAF.sub.t-1] -- -- --
Poverty 1.510 -1.505 0.193
(0.88) (-3.81) (0.10)
Agricultural 6.515 -0.110 -0.126
Productivity (3.43) (-6.71) (-0.29)
LogPopulation 1.470 0.318 0.017
Density (1.43) (4.16) (0.94)
LogPopulation -- -- --
[Density.sub.t-1]
Water Conflicts 0.395 -0.010 -0.032
(1.38) (-1.49) (-1.99)
Resistance Action 3.915 0.023 0.057
(3.03) (3.21) (1.56)
Demonstrations 0.453 0.021 0.003
(5.58) (23.39) (4.31)
Murder Attempts 0.144 -0.003 0.0084
(0.78) (-0.89) (2.16)
Death Threats 0.040 0.006 0.015
(0.75) (2.86) (1.57)
Alpha -- -- --
Nobs 234 234 234
Loglikelihood -1,221.16 -1,431.45 -879.990
Chi-Squared 100.66
Prob [chi sqd > (0.000)
value]
Poisson with Poisson with Dynamic
Variable Heterogeneity Endogeneity Poisson
Constant -19.598 -2.334 -8.624
(-35.39) (21.32) (-3.91)
[Occupation.sub.T-1] -- -- 0.009
(3.15)
[Occupation.sub.t-2] -- -- -0.005
(-1.32)
[Occupation.sub.t-3] -- - 0.013
(3.02)
PT 0.145 0.833 0.293
(2.16) (3.82) (3.23)
INCRA -0.001 0.068 0.011
(-1.46) (3.07) (2.47)
[INCRA.sub.t-1] -- -- -0.017
(-2.67)
LogPRONAF 0.536 0.134 0.127
(12.77) (0.88) (3.218)
[LogPRONAF.sub.t-1] -- -- 0.032
(3.127)
Poverty -2.448 -2.702 -1.539
(-8.70) (-2.53) (-1.04)
Agricultural -0.461 -0.539 -0.113
Productivity (-5.84) (-0.91) (-0.49)
LogPopulation 0.219 0.126 0.472
Density (5.95) (1.44) (3.25)
LogPopulation -- -- -0.002
[Density.sub.t-1] (-2.47)
Water Conflicts -0.024 -0.071 -0.006
(-2.95) (-1.82) (-0.36)
Resistance Action 0.080 0.207 0.121
(4.60) (2.34) (2.81)
Demonstrations 0.046 0.037 0.015
(39.08) (5.23) (3.89)
Murder Attempts -0.023 0.024 0.009
(-6.22) (1.21) (0.85)
Death Threats -0.007 -0.021 0.004
(-3.44) (-2.14) (0.69)
Alpha 0.163 -- --
Nobs 234 234 234
Loglikelihood -2,065.49 -1,932.45 -2,172.35
Chi-Squared 392.56
Prob [chi sqd > (0.000)
value]
NOTE: The parameters in bold are statistically significant at 1
percent or 5 percent.
