Text-only Preview



Institute for Development Policy and Management, University of Manchester,
Harold Hankins Building, Oxford Road, Manchester, M13 9QH, UK;
(tel: +44161 2752827, fax: +44161 2738829; email: [email protected])

Abstract: The paper investigates the complex system of causes affecting tropical deforestation at a
worldwide level. There is no generally accepted theory in the deforestation literature to indicate
which variables should be included in a model of deforestation at an aggregate global level. The
paper begins, therefore, by presenting an analytical structure based on formal farm household
economic modelling literature. The empirical findings derived from a global regression model tend
to confirm the profit maximising market approach to deforestation, i.e. policy and structural
variables at the macro-level that stimulate agricultural production provide farmers with incentives to
deforest and expand their arable land areas. However, subsequent statistical tests suggest that the
causes of tropical deforestation are difficult to identify and quantify at a global level, and that these
should be analysed at a more disaggregated level.
Keywords: global tropical deforestation, farm household models
JEL classification: C23, Q23

1 The paper represents a revised and shortened format of the author’s Master thesis undertaken during his MA
studies in Development Economics, 2000-2001, at the University of Sussex, Department of Economics. I am
grateful for valuable comments made by Colin Kirkpatrick and Armando Barrientos on this revised version of
the paper. The thesis can be found in the University of Sussex Main Library Catalogue


1. Introduction

Deforestation refers to the removal of trees from a forested site and the conversion of land to another
use, most often agriculture (van Kooten, 2000). Deforestation is primarily confined to developing
countries, and mainly in the tropics (Myers 1994). There is growing concern over shrinking areas of
tropical forests (Barraclough and Ghimire 2000). The livelihoods of over two hundred million forest
dwellers and poor settlers depend directly on food, fibre, fodder, fuel and other resources taken from
the forest or produced on recently cleared forest soils. Furthermore, tropical deforestation has
become an issue of global environmental concern, in particular because of the value of tropical
forests in biodiversity conservation and in limiting the greenhouse effect (Angelsen et al 1999). This
has led economists to increase their efforts to model the process of deforestation and conversion of
forests to other land uses.
The rationale of this paper is to investigate the set of driving forces that might induce
tropical forest depletion. The focus is on the socio-economic macro-level factors, for which a global
regression model has been constructed in order to analyse empirically and test the existence of
macroeconomic explanations of deforestation and the channels through which these might work.
The global regression model is constructed using panel data on macro-level variables across 50
tropical countries over an 18-year period.1
The main findings confirm the validity of a market profit maximising approach to tropical
deforestation. In other words, factors that stimulate production and profits such as increased
agricultural output prices, decreased input prices and increased flow of technology into agriculture,
have a negative effect on the preservation of forestland, and contribute to its conversion to
agricultural uses. However, after a thorough statistical analysis, the model correctness and the
validity of the initial findings are critically evaluated and the significance of estimated coefficients is
questioned. The validity of generalised macroeconomic explanations of deforestation is questioned
therefore, emphasising the importance of microeconomic and case study work.


The paper is structured as follows. Section 2 considers a conceptual analysis of the driving
forces of deforestation. Section 3 discusses the theoretical underpinning. Section 4 constructs a
global regression model and presents the main results obtained. Finally, section 5 concludes.

2. The driving forces of deforestation - a conceptual framework.

Pearce and Brown (1994) identify two main forces affecting deforestation:
• Competition between humans and other species for the remaining ecological niches on land and
in coastal regions. This factor is substantially demonstrated by the conversion of forest land to
other uses such as agriculture, infrastructure, urban development, industry and others.
• Failures in the workings of the economic systems to reflect the true value of the environment.
Basically, many of the functions of tropical forests are not marketed and, as such, are ignored in
decision-making. Additionally, decisions to convert tropical forests are themselves encouraged
by fiscal and other incentives.
The second of these fundamental forces nurturing deforestation has been intensively analysed by the
environmental economics literature. Poor farm households or commercial loggers have little
incentive to care about the environmental effects of their actions. Such unaccounted costs give rise
to so-called economic failures, which could be classified into local market failures, policy failures
and global appropriation failures (Panayotou 1990). Market failures are present because an
unregulated market economy will fail to produce an optimal outcome where prices generated by
such markets do not reflect the true social costs and benefits from resource use and convey
misleading information about resource scarcity, providing inadequate incentives for management,
efficient utilisation and enhancement of natural resources. Policy failures or market distortions are
present where misguided intervention or unsuccessful attempts to mitigate failures result in worse
outcomes (Panayotou 1990). For example, lack of respect for traditional land rights make property
rights to forestland uncertain, and could encourage short-term exploitation of forests rather than
long-term sustainable use. Finally, global appropriation failures are present because, in the case of


tropical forests, the benefits of preservation, of biodiversity, and the value of the genetic pool in
developing new medicines, crops, and pest control agents, are poorly reflected in market allocations.
These provide services that extend far beyond the borders of the host country, reducing incentives to
implement globally efficient policies (von Amsberg 1998).
Thus, the existing institutional and legal framework leaves forests outside the domain of
markets, unowned, unpriced and unaccounted for (von Amsberg 1998), fostering their excessive use
and destruction, despite their growing true economic values.
The absence of first-best policies that could effectively internalise the externalities arising
from the economic failures previously described reinforces the factors shaping the decisions of
agents to deforest. Thus, fundamental forces combine with specific factors to influence the decisions
made by the deforestation agents. These interconnections are illustrated in figure 1.

Figure 1: The driving forces of deforestation

Human-other species



Fundamental forces


Specific factors

Source: Author’s diagram.

Most of the existing literature typically distinguishes between two levels of specific factors:
direct and indirect causes of deforestation. The direct causes of deforestation are also typically
referred to as sources of deforestation (Caviglia 1999), first-level or proximate causes (Panayotou
1992; Barbier et al. 1994). These, as Shafik (1994) observed, are fairly obvious: forests are cleared
either to harvest timber or other forest products or they are cleared to use the land for agriculture or
livestock.2 However, as a practical matter, the interaction between different types of agents


frequently makes it difficult to separate their impacts and determine their relative importance. Often,
ranchers and loggers facilitate small farmers’ entrance into forested areas, farmers engage in logging
to finance agricultural expansion, and ranchers follow small farmers into agricultural frontier areas
(Angelsen and Culas 1996). The indirect causes of deforestation are more complex and
controversial. These include both factors that immediately affect the decisions of agents to deforest
(such as output and input prices) and those that have a delayed impact on agents’ decision-making
(such as underlying terms of trade and technological progress). The major drawback of the direct-
indirect classification is that it merges the immediate and underlying causes under the label of
indirect or second-level. Since the underlying causes determine the decision parameters, mixing
these two levels flaws the cause-effect relationship and creates serious problems in empirical
investigations and regression models, such as high levels of multicollinearity (Angelsen et al 1999).
In order to avoid this, the specific factors can be more appropriately classified into three distinct
groups: sources of deforestation, local-level causes, and macro-level causes of forest depletion.
A first step is to identify the agents of deforestation, i.e. small farmers, loggers, ranchers,
and their relative importance in forest clearing. Their actions are considered to be the sources of
deforestation.3 In the second step, the agents make decisions about certain choice variables based on
their own characteristics and given decision parameters (Kaimowitz and Angelsen 1998). The
authors give examples of endogenous choice variables such as land allocation, labour allocation and
migration, capital allocation, consumption and other technological and management decisions. The
characteristics of deforestation agents are mainly described by their objectives and preferences,
initial resource endowments, knowledge and cultural attributes. The agents’ decision parameters are
external to individual agents and consist of output, labour and other factor input prices, accessibility,
available technology and information, risk, property regimes, government restrictions and physical
environmental factors (Kaimowitz and Angelsen, 1998). Thus, the characteristic and decision
parameters determine the set of permissible choices and constitute the local level causes of
deforestation.4 Finally, broader policy variables and trend and structural variables such as
demographic and technological forces indirectly influence the agents’ characteristics and decision


parameters. These are grouped under the label of macro-level causes of deforestation.5 Hence, the
conceptual framework as depicted in figure 2 provides a distinction between local-level and macro-
level causes that makes it a useful and necessary basis from a methodological point of view.
Figure 2: Conceptual framework of the causes of deforestation


Agents of deforestation:
Sources of
Choice variables

Decision parameters and
Local level causes
agent characteristics
of deforestation

Institutions Infrastructure Markets Technology

Macro-level causes of
Policy variables and trend / structural factors

Source: Adapted from Kaimowitz, D. and A. Angelsen (1998) Economic models of tropical deforestation. A
review, Centre for International Forestry Research.

3. Theoretical underpinnings

There is no consensus around a theory of deforestation indicating which explanatory variables at the
macro-level should be included in an empirical model (Andersen 1996). The derivation of such a
model is not the intention of this paper. However, the paper will rely on the deforestation literature
providing economic theories at the farm household level related to the links between deforestation,
proxied by agricultural land expansion, and changes in agents’ decision parameters (i.e. local-level
causes of deforestation). This delivers a theoretically underpinning for the empirical analysis and
systematic guidance as to which macro-variables might be inserted in the regressions. The section


outlines two different and extreme models of agricultural land expansion: the subsistence
(population or full-belly) approach and the market (open economy or profit maximising) approach.6


The subsistence approach assumes an extreme case, i.e. that no markets exist. This theoretical
approach begins from the assumption that a person’s objective is to satisfy his subsistence
requirement by producing agricultural commodities (Angelsen et al., 1999). The economic problem
is to minimise the labour efforts given a subsistence target, implying that consumption beyond that
level has no value. This is labelled by Dvorak (1992) as the full belly version of clearing fields.
Production is determined by:
X = Af (L, H , F)
where X is production in physical units, A represents the technological level, L is (on the field)
labour input, H is total land area (land assumed to be of homogenous quality), and F is fertiliser
input. The production function is assumed to be concave, with positive but decreasing marginal
productivity of all inputs. All inputs are normal and any pair of inputs is complementary.
Because no market for land is assumed, uncultivated land (forest) can be brought into cultivation on
a “first come first served” basis (Angelsen et al 1999). There are, however, costs related to the
clearing of new land, and also costs from having a large area to cultivate, for example, in terms of
walking, transport of inputs and output. These additional costs are represented by a convex function
h(H). Hence, the optimisation problem is to minimise L + h(H) subject to the
constraint sN = pX ? qF , where the subsistence target is given by subsistence consumption (equal
to income) per capita (s), multiplied by the total population (N), and p and q are output, and,
respectively fertiliser (input) prices.
The Lagrangian (denoted by G) of this minimisation problem is:
G = L + h(H ) ? ?[ pAf (L,.H , F) ? qF ? sN ] , where ? is the Lagrangian parameter.


Setting the first-order conditions (i.e. the derivatives of the Lagrangian G with respect to its
arguments, L, F, H and ?) to zero, the following are obtained:

pA =

FOC2: pAf (L, H , F ) ? qF = sN
The term (1/?) in the first-order conditions (FOC1) can be interpreted as the shadow wage of labour
(social opportunity cost of labour), which is endogenous in the respective model. Thus, at the
optimum the marginal costs per output unit of the three inputs equal the price of output (p),
multiplied by the technological level (A) (Angelsen et al., 1999).
The effects of exogenous changes on the land area under cultivation are relatively obvious.
An output price increase or technological progress makes it attractive for farmers to meet the
subsistence target by producing from a smaller land area. Lower fertiliser prices will induce farmers
to use more fertilisers and less land and labour inputs, and thereby reduce the pressure on forests.
Improved accessibility, in terms of lower costs of bringing new land into cultivation, has the
opposite effect, namely, an increase in forest depletion. Finally, population growth increases the
overall consumption requirement, and therefore leads to increased area of cultivation and
A major limitation of this model is the key assumption that households only seek to meet a
pre-established consumption target and lose all interest in working once they have reached that goal
(Kaimowitz and Angelsen 1998). Nevertheless, the model could be empirically relevant in those
situations where producers are virtually isolated from markets or where norms require any
production beyond subsistence levels to be shared, which greatly reduces the incentives to produce



The key change in the underlying model assumptions is the introduction of a perfect labour market
where labour can be sold or hired at a fixed exogenous wage rate (w), which determines the
opportunity cost of labour used in agriculture. This, in turn, implies that the level of population is
endogenous (whereas in the subsistence model it was exogenous), since labour must be allowed to
move freely between the farm and off-farm sectors to ensure labour supply and demand converge at
the predetermined wage rate. The land expansion decisions can thus be studied as a profit
maximising problem, where the household maximises total profits or land rent.
Maximize: pAf (L, H , F ) ? qF ? [
w L + h(H )]
Setting the first-order conditions (i.e. the first derivatives of the profit function with respect to its
arguments, L, H, F) equal to zero the following is obtained:
pA =

Although the first-order conditions look similar in both approaches, the interpretation of the impact
of exogenous changes on agricultural land area expansion in the market approach differs greatly
from that in the subsistence approach. This is because in the subsistence model the population
variable was assumed to be exogenous (and the shadow wage endogenous), whilst in the market
model the wage rate is exogenous (and population endogenous). In other words, agricultural
production and land use within the market approach are determined by the relative profitability of
agriculture and not by any subsistence requirement.
Therefore, higher output prices or technological progress will increase the relative
profitability of agriculture, which puts pressure on forests through an increase in land cultivation.
Increased fertiliser prices will, assuming complementarity between fertiliser and land area, reduce
the area of cultivation. Better access to the forest margin will, as in the subsistence case, lead to an
area expansion. A key variable for the determination of the extent of deforestation is the wage rate -
higher opportunity costs of labour will make cultivation on the forest margin unprofitable. Finally,
population does not enter the market approach model directly. However, by extending the approach


to include general equilibrium effects, a population increase will have indirect negative effects on
forested areas through lower wages and higher food prices (Angelsen et al. 1999).
The main criticism of this model is associated with the strong perfect labour market
assumption. In many contexts this is unlikely to hold, especially in the short run when there is no
migration and given the hypothesis that family labour is completely interchangeable with hired
labour is violated.
The comparative static results of the two theoretical approaches are summarised in table I.

Table I: Hypotheses derived from the subsistence and market approaches
…on land expansion and deforestation
Effect on def. of an increase in…
Subsistence approach
Market approach
Output price (p)
Input /fertiliser price (q)
Wage /alternative employment (w)
not applicable
Agricultural productivity / technology (A)
Population (N)
Costs of clearing and access (h(H))
Source: Angelsen, A., E. Shitindi and J. Aarrestad (1999) “Why do farmers expand their land into forests?
Theories and evidence from Tanzania”, Environment and Development Economics 4 (03): 313-31.

It is important to note that the predicted effects of changes in the technological level and in
output and fertiliser prices are different across the two models. Furthermore, the subsistence
approach highlights the effect of population growth, whereas the market approach highlights the role
of alternative employment, as expressed through the wage rate. In other words, while the subsistence
approach focuses exclusively on the agricultural sector, the market approach draws attention to
linkages with the rest of the economy.


Document Outline
  • 1. Introduction
  • 2. The driving forces of deforestation - a conceptual framework.
                  • Figure 1: The driving forces of deforestation
  • 3. Theoretical underpinnings
  • 4. Empirical analysis
                  • Figure 3: Proxies to be used in the initial empirical model
                  • Table III: Results obtained in the initial static model
  • 5. Conclusions
  • Notes
  • References
                  • Appendix
    • Data description