The Effects of Economic Development on Neotropical Deforestation : Household and Village Evidence from Amerindians in Bolivia

The Effects of Economic Development on Neotropical Deforestation:
Household and Village Evidence from Amerindians in Bolivia
Ricardo Godoy (HIID), Jeffrey R. Franks (IMF),
David Wilkie (Associates in Forest Research and Development),
Mario Alvarado (Universidad Mayor de San Andrés),
George Gray-Molina (HIID), Raul Roca (UDAPSO),
Jairo Escóbar (UDAPE), and Marina Cárdenas (UDAPSO)
Abstract
A survey of 200 lowland Amerindian households (Mojeño=88;
Yuracaré=50; Chimane=62) in the department of Beni, Bolivia, was
done to test the effects of income on the clearance of primary
rain forest. The link between income and forest clearance
resembles an inverted U, with clearance peaking among middle-
income households. Deforestation drops after households reach a
threshold of income because at higher income households seem to
enjoy: a) more secure property rights, b) higher educational
attainment and better educational facilities in their villages,
c) greater use of modern farm technologies (eg, chemical
herbicides and pesticides), which allows them to practice more
intensive agriculture, and d) higher income from off-farm
employment. After controlling for income, household variables
which lower pressure include radios, female education, and
credit; village variables which lower pressure include schools
and cattle ranchers. The presence of loggers, distance to market
town, and the number of stores in the village seem to be
associated with greater clearance.
Key words: deforestation, Bolivia, Beni, human capital, rain
forest, Mojeño, Yuracaré, Chimane, markets, Tobit, lowland
Amerindians.

---

.....Forests are fewer and fewer, rivers dry up, game
becomes extinct, the climate is ruined, and every day
the earth gets poorer and uglier. Here you are with
that mocking look in your eyes, and all I say seems to
you not very serious and...., indeed it may be
foolishness; but, when I pass the peasants' woods that
I have saved from being chopped down, or when I hear
the sound of my young wood rustling, the stand I
planted with my own hands, I realize that climate too
is a little in my power, and that a thousand years from
now if man should be happy, why, then I'll be a small
part of that too.
-- Uncle Vanya, Anton Chekhov, 1899 --

---

Introduction. Since the early 1970s researchers have been
studying the causes of neotropical deforestation (Gómez-Pompa et
al. 1972; Denevan 1973) but so far they have paid scant attention
to the role Amerindian households play at present in the
clearance of primary rain forest. This article contains a model,
a set of hypotheses from the model, and evidence from households
and villages from three lowland Amerindian groups in Bolivia to
test the hypotheses with multivariate techniques. In the model
and in the hypotheses we stress the role of income or integration
to the market on the clearance of primary rain forest.
We study rural households because they decide how much
forest to clear even after governments put in place policies to
lower deforestation (Bilsborrow 1992; Hecht 1993; Moran 1993a).
We study Amerindians because they have received less attention
than ranchers, loggers, or than smallholders and because they are
deciding more and more how land, minerals, and timber should be
used within their reserves (Conklin and Graham 1995; Boremanse
1983; Greenbaum 1989; Dodds 1993; CEDI 1993; Vidal 1989; Turner
1995). We focus on the clearance of primary rain forest because
it contains more biological diversity than secondary forest
(Frumhoff 1995; Saldarriaga et al. 1985; Lawrence et al. 1995).
And we study income or integration to the market because it
simultaneously affects how a household produces, consumes, saves,
and invests, all of which affects the use of the rain forest
(Wilkie et al. 1996).
We hypothesize that the link between income and neotropical
deforestation resembles an inverted U, with clearance peaking
among middle-income households. Poor households clear small
plots of forest because they have simple technology (Carneiro
1983; Denevan 1992; Johnson 1993) and because they save in trees
to enrich the fallow and to reduce the costs of foraging (Irvine
1989; Posey 1984). The expansion of markets into isolated
villages changes incentives, which induces households to clear
forests to plant annual crops (Libecap and Alston 1992), to amass
rents, assets, and status symbols (Lipton 1982; Uhl et al. 1991),
to meet demand for more industrial goods, to increase herd size
(Hecht 1993; Partridge 1984, 1989; Browder 1994a; Schmink and
Wood 1987; MacDonald 1981), and to invest in activities with a
fast pay off (Cancian 1979; Pinedo-Vásquez et al. 1992; Hiraoka
1986; Schneider 1993; Schneider et al. 1993; Foweraker 1981).
Neotropical forests may therefore face most pressure during the
transition stage of economic development or among middle-income
households. In richer households and villages tightly linked to
markets, deforestation drops because incentives change. Rules of
land tenure become clearer (Roosevelt 1990; Kishor and
Constantino 1993), causing a rise in investments to improve
yields from crops and from livestock (Toniolo and Uhl 1994;
Browder 1994; Anderson and Hill 1990), and people have greater
opportunities to work outside the farm and so depend less on the
forest (Painter 1995).

---

To test this idea we did field work among Chimane, Mojeño,
and Yuracaré Amerindians in the lowlands of the department of
Beni, Bolivia. Below we present the model, hypotheses, methods,
and the results of the analysis.
Model. We use theories from household, trade, and development
economics to develop a model of why and how Amerindian households
deforest as they become part of market economies. We combine the
approaches because they yield a parsimonious, generalizable, and
testable model of deforestation by rural populations across and
within different levels of integration to the market or levels of
income.
The context of household decision making. As Amerindian
households become part of market economies changes in the demand
for land, labor, and for cash and changes in private time
preferences cause the clearance of rain forest to rise and then
to fall. Below we explain what lies behind the parabolic shape
of land clearance. In the next section, entitled, "Household and
village-based paths of forest clearance," we explain what drives
forest clearance within any one level of income or point along
the parabola.
Land. As indigenous households become part of market
economies the demand for their crops rise in line with increases
in income and in population (Netting 1982, 1993). Traditional
households clear small plots of forest to farm because they have
simple technology and few people, but as markets envelope
isolated villages and as households adopt new technologies for
clearing forests, exports of crops to the outside world grow.
All else held constant, the increase in the demand for crops by
the outside world and rising labor productivity should cause
forest clearance to rise (Behrens 1992; Ehui et al. 1990).

Labor. As Amerindian households become part of open
economies, demand for their labor also grows. Remote Amerindian
societies export few workers (Stier 1982, 1983) and earn a small
share of their income from jobs outside the farm (Werner et al.
1979). When villages become part of market economies, the demand
for workers to extract forest goods or to work in plantations or
in ranches grows, thereby raising the opportunity cost of labor
and lowering the incentives to remain in the forest or to do
traditional agriculture (Homma 1992; Burkhalter and Murphy 1989;
Behrens 1992a).
Cash. In traditional economies households do not need
to rely much on cash because they can smooth consumption through
reciprocity and gifts from kin and neighbors. As households move
toward a market economy, reciprocal obligations weaken or break
down in part because growing income affects people's perceptions
of ownership (Smith and Taput 1995). As reciprocity weakens,
2

---

households need to sell more crops and goods from the forest to
pay for inputs or to weather shocks; land clearance then rises.
In the advanced stages of integration to the market dependence on
the forest to finance household activities declines because
households can rely on formal financial institutions.
Time preferences. Economic theory and experimental
studies suggest that time preferences fall as incomes and wealth
rise (Hausman 1979; Gately 1980; Benzion et al. 1989; Pender
1993). As time preferences drop, households increase savings in
physical assets, such as cattle, in part because villages lack
banks. But as formal financial markets expand the need for
households to save in physical assets to smooth consumption falls
and so does the need to clear forest to put in cattle.
The simultaneous rise in the demand for off-farm labor and
for crops offset each other and cause land clearance to resemble
an inverted U (Figure 1). The rise and the decline in the use of
the rain forest as a source of cash and the rise and the decline
in the use of cattle as savings reinforce the inverted U-shaped
parabola.

-- INSERT FIGURE 1 ABOUT HERE --
Household and village-based paths of forest clearance.
Holding constant income or the level of integration to the market
of a household, forest clearance will be driven by attributes of
the person, of the household, and of the village. The weight of
personal, household, and village attributes will vary over the
continuum of integration to the market, with village attributes
probably overshadowing household and personal attributes in the
early stages of an idealized autarky-to-market continuum. Norms
of reciprocity in traditional villages should allow households to
overcome shortcomings in their personal endowments (eg, illness)
and help them to do things which they may otherwise find
difficult to do. For example, families can rely on extra-family
pooling of labor to clear forests. In the more advanced stages
of integration to the market and at higher levels of income, as
households split apart from one another and as village
institutions and networks of reciprocity break down (Smith and
Tapuy 1995; García-Barrios and García-Barrios 1990), households
have to fend more and more on their own; we would then expect
personal and household attributes to eclipse village attributes
in driving forest clearance.
Village variables matter for at least three reasons. First,
they help to control for endogenous variation in the main
explanatory variable, income or integration to the market.
Second, they may correlate with household and with personal
variables. Leaving out village variables may produce misleading
parameter estimates for household and for personal variables
3

---

(Duraisamy 1992; Rosenzweig and Wolpin 1988; Strauss 1990; Tansel
1993). Lastly, village variables interact with household and
with personal variables to raise or to lower the private costs
and benefits of clearing forest to the household.
Village variables such as the number of encroachers or
circumscription (Gross et al. 1979; Schneider 1993), the amount
of usable primary forest per person or the ratio of primary to
secondary forest (López 1993), presence of outside institutions
(Behrens et al. 1994), or proximity to roads and to market towns
(Plattner 1989) can push or can lure households to become part of
market economies and earn higher income. Other village
variables, such as the average education, income, or wealth of
villagers or the presence of modern physical infrastructure can
lower the costs and raise the benefits of getting information
about modern farm technologies, health, and markets for capital,
labor, and for crops and allow households to enhance farm
productivity and clear less forest.
Hypotheses. A + or a - sign next to the sub-headings below shows
the expected marginal effect of explanatory variables on the
clearance of primary rain forest holding constant all other
explanatory variables. Explanatory variables take on a positive
sign if an increase in the variable raises the area of primary
forest a household will clear because it: i) lowers the marginal
cost of clearing primary forest or ii) raises the marginal costs
of alternatives to clearing primary forest.
Household and personal variables:
Income (inverted U). Although many case studies of
nations or regions within nations and many international
comparisons show that the link between income and deforestation
resembles an inverted U (Lugo 1992; Vincent and Othman 1995;
Vincent and Binkley 1992; Singh 1994; Hyde et al. 1993; Patel et
al. 1994; Ruitenbeek 1988, 1989), some researchers have found
weak links between income and deforestation or have even found
evidence of a U-shaped parabola (Allen and Barnes 1985; Rudel
1989, 1993; Cropper and Griffiths 1994; World Bank 1992; Shafik
and Bandyopadhyay 1992).
For reasons discussed in the model above we hypothesize that
the link between income and deforestation should resemble an
inverted U. We think past researchers have found weak evidence
for the hypothesis because: i) they relied on secondary or
aggregate information which was not necesarily designed to test
the hypothesis, ii) the information may not have contained enough
variance in the chief explanatory variable, income or integration
to the market, and iii) researchers lacked an explicit model and
intermediate-level explanatory variables to link deforestation
with income.
4

---

Number of adults (+). We expect the number of men and
of women over the age of 10 to be positively associated with
forest clearance for two reasons. Only adult men cut primary
forest but adult women help to weed, harvest, and maintain
fields. The number of both adult men and women in the household
should predict clearance because their tasks in the farm
complement one another. Second, the number of people over a
survival threshold of about 10 years of age should bear a
positive link to clearance if parents leave land as inheritance
for their children (Piland 1991; Godoy et al. 1996a). The number
of males should play a more dominant role than the number of
females if there is a growing agnatic bias in land inheritance
from land shortage.

Number of children (+). The number of children should
bear a positive link to the clearance of forest because it should
increase the demand for food and for other commodities
(Bilsborrow 1992).
Both demographic variables, number of adults and number of
children, should have a stronger effect on forest clearance among
households with weak links to the market because of the closer
fit that seems to exist between consumption, income, and
production in traditional villages with imperfect markets for
labor and for capital (Benjamin 1992). With better functioning
markets, households can borrow and hire workers at will to
overcome demographic constraints on their family labor supply.
Or, as Benjamin (1992:288) puts it, "the number of workers in
Baron Rothschild's vineyards should not depend on the number of
daughters he has."
Formal education (-). Formal education should lower
pressure on the forest (Godoy, Groff, O'Neill 1996; Moran 1989,
1989a; Tongpan et al. 1990) because it should ease out-migration
and the adoption of modern farm technologies that raise the
productivity of land and of labor (Phillips 1994; Schultz 1988;
Lockheed et al. 1980; Jamison and Lau 1982; Jamison and Moock
1984). Research in Asia shows that farmers with more than four
years of education found it easier to adopt new farm technologies
(Phillips 1994). It remains unclear whether the education of
women will have a greater or a lesser effect on deforestation
than the education of men.

Cattle (+). The stock and flow of cattle should
encourage forest clearance. When the number of cattle exceeds
the carrying capacity of the village grazing commons people have
to clear the forest for pasture. A reduction in the stock (or an
increase in the flow) of cattle should also encourage clearance
because it provides the household with meat or with cash to
recruit laborers for agriculture.
5

---

Credit (-). Households facing credit constraints may
be forced to cut trees for sale as lumber, charcoal, or canoes,
whereas households that can borrow from the outside world may
have wider options for earning and smoothing income and
consumption (Morduch 1995) and so may not need to depend on the
forest as much.

Information (-). Households with better access to
information from the outside world about labor markets and new
farm technologies should clear less forest because they will be
more likely to work outside the farm or to adopt innovations in
farming (Southgate and Whitaker 1992; Moran 1993; Ledec and
Goodland 1988; Southgate 1991).
Village explanatory variables:
Distance from village to market town (+). Although
researchers in Brazil and in Ecuador (Southgate et al. 1991;
Pérez-Garcia 1991; Pfaff 1995) have found a close fit between
roads and forest clearance, proximity to market towns could also
lower forest clearance because it could make it easier for
households to find work outside the farm and to adopt new farm
technologies, thus increasing the productivity of land already
cleared (Mann 1989; Brush et al. 1992).
Size of village population (unclear). Some researchers
find a strong positive link between population growth or
population size and deforestation (Palo 1994; Allen and Barnes
1985; Rudel 1989; Cleaver and Schreiber 1991; Cruz 1992), but
other researchers find only weak or lagged effects (Pérez-García
1991; Deacon 1994; Bilsborrow 1992). Pfaff (1995) used
instrumental variables to control for endogeneity when using
population as an explanatory variable for deforestation in Brazil
and found that population did not matter after controlling for
other variables.
Age of village (-). The age of the village should
lower forest clearance because it proxies: 1) for the
availability of secondary forest, which is easier to cut than
primary forest (Bedoya 1995), and 2) for more secure rights of
property to land.
Tenure (unclear). Results of international studies
show that political (Deacon 1994) and macroeconomic (Almeida and
Campari 1994) instability undermines property rights to land and
increase deforestation. But the evidence from household surveys
is less clear. López (1993, 1993a) found that communal tenure in
Africa caused clearance that exceeded the social optimum. In
Ecuador, titling reduced clearance (Pichón et al. 1994). In
Peru, Bedoya (1987) found that squatters deforested about twice
as much as legal owners. But elsewhere Bedoya (1991) and others
6

---

show that tenure security only changes the way people use the
forest and not the total area of forest they clear. In the
Ecuadorian and in the Peruvian Amazon title induced people to
clear forests to plant trees rather than to plan annuals but did
not change the size of the area they cleared (Bedoya 1991;
Southgate 1990). The effect of titling shrinks once researchers
control for the shorter residence duration of squatters (Bedoya
1995). Household surveys in Brazil show that titling was
insufficient to lower itinerancy and encroachment (Almeida 1992).
Modern extractive technologies (+). Modern extractive
technologies (eg, chainsaws) should raise forest clearance
because they lower the marginal costs of cutting.
Methods. To test these hypotheses we did field research among
the Yuracaré of the river Sécure in the Parque Nacional Isiboro-
Sécure, Mojeños in the Territorio Indígena Multiétnico, and
Chimane in the Territorio Indígena Chimán, all in the department
of Beni, Bolivia. We chose these groups because they had large
variance in explanatory variables, particularly in their degree
of integration to the market or income.
Field work took place during ten weeks, from July to
September, 1995. During that time we interviewed 200 households
(88 Mojeño, 50 Yuracaré, and 62 Chimane) in 20 villages. The
indigenous government of each ethnic group helped us pick three
villages in their group along a continuum of integration to the
market. For each village they helped us select one or two
replicates with the same degree of integration to the market.
Thus, each ethnic group contained a total of at least six
research sites straddling different amounts of involvement in the
market. To avoid missing the tail of the acculturation
distribution -- for instance, households that had moved to towns
-- and so biasing the sample toward poorer households with weak
links to the market, we also interviewed households that lived
and worked in nearby towns but that kept a toehold in village
life.
In each village we interviewed the village headmen, school
teachers, and heads of households. From them we collected
statistics on village variables, such as wages, encroachers,
number of stores, number of villagers studying outside the
village, and presence of public institutions and non-government
organizations.
From (mainly) male heads of households we collected
information about each household. Before interviewing heads of
households the village headman called a meeting to explain our
study to villagers. Interviewing only those who attended the
meeting would have skewed the sample toward the more civic-minded
villagers, toward the bilingual speakers, toward the ones with
7

---

more free time, or toward the non-migrants. To avoid the slant,
we also interviewed as many households as we could, including
households headed by women, households which did not attend the
meeting, or households which did not speak Spanish. We used
bilingual speakers to interview people who only spoke an Indian
language. In the villages visited we interviewed the following
percentage of households: 45% (Chimane), 56% (Mojeño), and 65%
(Yuracaré).
Chimane, Mojeño, and Yuracaré compared. The three groups
resemble each other at many levels. The villages we studied lie
along several parallel rivers, which drain the Eva-Eva Mosetene
basin before discharging their waters into the river Mamoré.
Except for a few villages which abut or include savanna, most of
the villages lie in primary rain forest. Many settlements of the
Mojeño and of the Chimane face encroachment from loggers and from
cattle ranchers (Thiele et al. 1995). Although the three groups
forage, they rely on swidden farming along river banks for the
bulk of their subsistence and use wooden dibbles and metal tools
to prepare their land (Alvarado 1996). The chief cash crops
include rice, cassava, and fruits. Some villagers also sell
firewood, logs, and thatch. During the past five years, but
especially since the floods of 1992 and 1993, ranchers,
missionaries, and non-government organizations have brought
chemicals, chainsaws, mechanical dibbles, and new varieties of
rice, maize, and pasture grasses (Lehm et al. 1994). Innovators
are experimenting with the cultivation of beans, wheat,
vegetables, and soybeans. The three groups buy or barter with
outsiders for kerosene, clothing, salt, school supplies,
medicines, lard, and metal tools. Besides selling crops, they
earn cash from wage labor during the dry season (May to November)
in logging camps, in cattle ranches, and in towns.
The three groups also resemble each other in their
horticultural customs (Alvarado 1996). Households cut forest to
put in crops for their own use or for sale, not so much to put in
livestock, as do ranchers or highland migrants. Some do it to
leave inheritance for their children, to buy status frills, or to
pay for family festivities. Tillagers use fields for one to four
years and rotate crops over the life cycle of the plot (Piland
1991; Lehm et al. 1994). The first year they put a new field to
use households sow rice but they also plant scattered patches of
cassava, bananas, maize, or tobacco. Maize, bananas, sugar cane,
cassava, and fruit trees begin to overshadow rice in area and in
effort by the second year. By the third and fourth years --
toward the end of the cropping cycle of a plot -- perennial crops
begin to replace grains and tubers (Huanca 1995). Border crops,
such as bananas, sugar cane, cotton, citrus, and tree crops
signal permanent rights of use over fallow forests in villages
with little land. But in villages with ample land, as among
remote Chimane or Yuracaré settlements, people re-use forests in
8

---