Content uploaded by Subhrendu K Pattanayak
Author content
All content in this area was uploaded by Subhrendu K Pattanayak on Feb 25, 2015
Content may be subject to copyright.
Deforestation and malaria: Revisiting the human ecology perspective
Subhrendu K. Pattanayak and Junko Yasuoka1
In CJP Colfer (ed.), People, Health, and Forests: A Global Interdisciplinary Overview. Earthscan.
1. Introduction
The ecological basis for disease dates at least as far back as 400 B.C. to Hippocrates’s writing of On Airs,
Waters, and Place. As Wilson (1995) clarifies, our understanding and therefore control of diseases would
be inadequate without an “ecological” perspective on the life cycles of parasitic microorganisms and the
associated infectious diseases. As Smith et al. (1999; p 583) contend, “many of the critical health
problems in the world today cannot be solved without major improvement in environmental quality.” In
this chapter we focus on malaria because its transmission (and control) has clear links to ecosystem
changes that result from natural resource policies such as land tenure, road building, and agricultural
subsidies. The resulting ecosystem change has a tremendous influence on the pattern of diseases such as
malaria (Martens 1998; Molyneux 1998; Grillet 2000). This is partly because, of all the forest species that
transmit diseases to human beings, mosquitoes are among the most sensitive to ecosystem change: their
survival, density, and distribution have been altered by environmental changes caused by different land
transformations. While we agree that ‘ecological lenses’ can help improve our understanding of disease
prevention, we use this chapter to articulate a particular ecological perspective – a human ecology
viewpoint that puts human behavior change front and center.
In the last decade, we have seen a series of widely cited papers drawing out the connections between
ecosystem change and diseases, many of which are synthesized in the 2005 Millennium Ecosystem
Assessment (Corvalan et al., 2005a; Corvalan et al. 2005b; Campbell-Lendrum, 2005; Patz et al., 2005;
McMichael et al., 1998). This renewed interest in the more distal causes of disease reflects in part the
emergence of new fields such as ‘sustainability science’ (Kates et al., 2001) and ‘biocomplexity’ (Wilcox
and Colwell, 2003) that argue for “a more realistic view [requiring] a holistic perspective that
incorporates social as well as physical, chemical, and biological dimensions of our planet’s systems.” The
resurgence also reflects the growing importance of fields of social epidemiology (e.g., Berkman and
Kawachi, 2000; Oakes and Kaufman, 2006) that draw on Rose’s (1985) call to examine the cause of
cause and resolve the prevention paradox in developing a population strategy for health.
In joining this growing chorus, we focus on an older human ecology tradition (Wessen, 1972;
McCormack, 1984), which posits that (a) we humans modify our natural environment, sometimes
increasing disease risks, and (b) we ultimately adapt to the new disease risk environment. Two stylized,
yet complicating, facts emerge from this viewpoint (Pattanayak et al., 2006a). First, disease prevention
behaviors (including ecosystem changes that modify disease risks) respond to disease levels, suggesting a
dynamic feedback exposure and control. Second, individuals and households typically will not consider
how their private actions affect public health outcomes and therefore will make socially inappropriate and
1Address correspondence to Subhrendu K. Pattanayak, Fellow and Senior Economist in Public Health and
Environment, RTI International and Research Associate Professor at North Carolina State University,
subhrendu@rti.org or Tel (919) 541-7355. Junko Yasuoka is a Post-Doctoral Fellow Harvard School of Public
Health, Department of Population and International Health. Many ideas reflected in this paper are based on
discussions with Erin Sills, Keith Alger, Gene Brantley, Kelly Jones, Christine Poulos, and George Van Houtven.
We are grateful to Carol Colfer for her encouragement and, most of all, patience with the development of this
chapter. This chapter was completed while Pattanayak was a Visiting Scholar at the University of California,
Berkeley.
sub-optimal choices, unless convinced otherwise. Typically, some combination of government laws (e.g.,
regulation), community norms (information), and markets prices (compensation) help narrow this wedge
between private and ‘optimal’ social behaviors. This modification of domain to now more systematically
human behavior is consistent with complaints that the ecology-and-health approach takes a predominantly
biophysical approach that can easily overlook the social, cultural, and economic driving forces that are
crucial to understanding anthropogenic ecosystem disruptions and their human health impacts
(McMichael, 2001; Parkes et al. 2003).
In this chapter, we focus on malaria and deforestation, rather than a sweeping review of broad links
between infectious diseases and ecosystem change to keep things manageable and present somewhat in-
depth arguments. We restrict ourselves to malaria not only because its transmission is clearly linked to
ecological changes, but because it is a major (if not the major) health concern in the tropics (Hay et al.,
2004). We focus on deforestation because it is a major development policy concern and often heralds
many other ‘malaria-causing’ land use changes (Pattanayak et al., 2006c).
The remainder of the chapter is organized as follows. In Section II, we briefly review the literature on
ecology of infectious diseases. In Section III, we re-introduce the human ecology perspective for better
understanding the role of humans in land use change as well as in a variety of behaviors to prevent (e.g.,
sleep under nets, take prophylaxis) and treat (e.g., seek medical care, follow the drug regimen) malaria. In
Section IV, we draw out the empirical implications of such a strategy, using our own fieldwork and
secondary data sets. Finally, we conclude with a call for systematic environmental and health impact
assessments that rely on inter-disciplinary longitudinal studies.
2. A brief synthesis of mosquito ecology and malaria epidemiology
While the impacts of ecosystem change on health are diverse and longstanding, its rate and geographical
range have increased markedly over the last few decades. Different kinds of environmental changes have
resulted from a wide variety of human activities, including deforestation, agricultural activities,
plantations, logging, fuel wood collection, road construction, mining, hydropower development and
urbanization (Walsh et al. 1993, Patz et al. 2000, Patz et al. 2004). It is the process of clearing forests and
subsequent land transformation that alters every element of local ecosystems, including microclimate, soil
and aquatic conditions, and most significantly, the ecology of local fauna and flora. These in turn have
profound impact on the survival, density and distribution of human disease vectors and parasites (Martens
1998, Grillet 2000), including influences on breeding places, daily survival probability, density, human-
biting rates, and incubation period. Thus, the altered vector/parasite ecology modifies the transmission of
vector-borne diseases such as malaria, Japanese encephalitis and filariasis (Sharma and Kondrashin 1991,
Walsh et al. 1993).
Numerous country and area studies have described how the density and distribution of local vector
species have been altered due to ecosystem change, and some longitudinal studies have shown that the
change in vector ecology has altered local disease incidence and prevalence (Sharma and Kondrashin
1991, Patz 2000). However, the mechanism linking ecosystem change, vector ecology and vector-borne
diseases is still unclear. We draw on a paper by Yasuoka and Levins (forthcoming) that examines the
mechanisms linking deforestation, anopheline ecology, and malaria epidemiology by drawing together 60
examples of changes in anopheline ecology as a consequence of deforestation and agricultural
development in Latin America, Africa, and South and Southeast Asia.
Massive clearing of forests has enormous impacts on local ecosystems and human disease pattern. It
alters microclimates by reducing shade, altering rainfall patterns, augmenting air movement, and
changing the humidity regime (Reiter 2001). It also reduces biodiversity and increases surface water
availability through the loss of topsoil and vegetation root systems that absorb rain water (Chivian ed.
2002). For anopheline species that breed in shaded water bodies, deforestation can reduce their breeding
habitats, thus affecting their propagation. On the other hand, some environmental and climatic changes
due to deforestation can facilitate the survival of other anopheline species, resulting in prolonged seasonal
malaria transmission (Kondrashin et al. 1991).
As shown in Table 1 (drawn from Yasuoka and Levins), different land transformations have different
impacts on local ecosystem and disease pattern. For example, rubber plantation increased local major
malaria vectors in all four cases in Malaysia and Thailand. In Malaysian hilly areas, forest clearance for
rubber plantation, which started early in the 1900s, exposed the land and streams to the sun and created
breeding places for An. maculatus, which led to an increase in this species and a marked rise in the
incidence and severity of malaria (Cheong 1983). Cyclic malaria epidemics in Malaysia over 50 years are
correlated with rubber replanting in response to market fluctuations (Singh and Tham 1990). Another
example is in Chantaburi, Thailand, where the land was transformed to rubber plantation and other fruit
tree cultivations, such as rambutan, durian and mangosteen spurred by high markets between 1974 and
1984. The consequent ecological changes favored An. dirus, which demonstrated its greatest capability
for adaptation in circumstances of rubber and fruit tree cultivations. As a result, local malaria reemerged
and malaria transmission was established at high levels (Rosenberg et al. 1990).
All papers on the development of irrigation systems reported an increase in the density of major vectors
and following increase in malaria incidence. For example, irrigation schemes developed by the Mahi-
Kadan Project across the River Mahi in India in 1960 had typical management problems, including over-
irrigation, lack of proper drainage, weedy channels, leaking sluice gates, and water-logged fallow fields.
These created extended breeding habitats for An. culicifacies, which resulted in an increase of the vector
and malaria transmission.
In some cases, different anopheline species responded differently to the same land transformation. For
example, due to deforestation for rice cultivation and irrigation development in Sri Lanka, An. annularis,
An. barbirostris, An. culicifacies, and An. varuna decreased, while An. jamesii and An. subpictus
increased, and An. nigerrimus and An. vagus did not change substantially (Amerasinghe et al. 1991;
Konradsen et al. 2000). Not only species abundance, but also species involvement in malaria
transmission changed markedly during the land transformation. Anopheles annularis, An. culicifacies and
An. vagus were the main vectors during the construction phase and the first irrigation year. Anopheles
subpictus was playing a major role in the second and third years, when rice fields were fully irrigated.
Throughout the process, An. culicifacies demonstrated continuous involvement in malaria transmission.
Other cases demonstrated species replacement. Land use such as cassava and sugarcane cultivations,
which need little water and provide little shade, often create unfavorable environment for anophelines,
especially those which require shade. In Thailand, the transformation from forest to cassava or sugarcane
cultivations eliminated shady breeding habitats for the primary vector species, An. dirus, but created
widespread breeding grounds for An. minimus, which have greater sun preference and was the
predominant species throughout the year. Consequently, malaria transmission among resettled cultivators
became high (Prothero 1999).
We also see that same kind of land transformation could result in totally different malaria situations,
depending on locality and ecological characteristics of local vector species. For example, deforestation
followed by development of coffee plantations in southeast Thailand favored the breeding of An. minimus
and made the previously malaria-free region to hyperendemic (Suvannadabba 1991). On the contrary, in
Karnataka, India, large-scale deforestation for coffee plantations reduced seepages, which were the
principal breeding sites for An. fliviatilis, a vector responsible for hyper-endemic malaria in the region.
As a result, this vector population completely collapsed, and malaria disappeared from the area (Karla
1991).
Deforestation for mine development is one of the examples that not only create breeding sites, but also
significantly increase human contacts with vectors. Where settlement and mining activities took place in
the Amazon, An. darlingi increased because of the increase in breeding sites, including borrow pits after
road or settlement constructions, drains, and opencast mine workings. As a result, malaria, which was
present in the Amazon’s indigenous population, was spread to immigrants and miners (Conn et al. 2002).
In summary, the changes in anopheline density and malaria incidence are both varied and complex,
depending on the kind of land transformation, ecological characteristics of local mosquitoes, and altered
human behavior (to be discussed further). Some key findings include:
Some anopheline species were directly affected by deforestation and/or subsequent land use, some
favored or could adapt to the different environmental conditions that were created, and some invaded
and/or replaced other species in the process of development and cultivation.
Malaria incidence fluctuated according to different stages of development, changes in vector density,
and altered human contact patterns with vectors.
More mosquitoes (vector density or variety) were neither a necessary nor a sufficient condition for
increases in malaria incidence. In fact, inverse relationships between the vector abundance and
disease incidence have been reported from different regions (Ijumba and Lindsay 2001, Amerasinghe
2003), presumably because of human adaptations (see next).
In general, a complex set of macroeconomic (changes in terms of trade), demographic (e.g.,
migration), policy (e.g., colonization of forest frontiers) and behavioral (e.g., ‘malaria literacy and
knowledge’) factors underlie the ecosystem changes and land transformations that influence mosquito
ecology and malaria epidemiology (Sharma and Kondrashin 1991; Molyneux 1998). We turn to these
considerations in some detail next.
3. Revisiting the human ecology perspective
If ecosystem changes impact mosquito density and activity, and possibly malaria incidence, then
environmental management (e.g., vegetation management, modification of river boundaries, drainage of
swamps, reduction of standing water, oil application etc.) could reverse these trends. Even though
insecticide-treated bed nets and indoor residual spraying of insecticides are the predominant vector
control tools, there is growing support for the management of vegetation and water bodies in light of
increasing resistance to insecticides and antimalarials (Lindsay and Birley, 2004). Keiser et al.’s (2005)
review of 24 environmental management studies suggests that environmental management can reduce
malaria risk ratio by 88% (compared to 79.5% for human habitation modifications, for example).
Furthermore, if these are indeed modifiable behavioral causes, it should be possible to induce these
behaviors. Yasuoka et al. (2006a) conducted a 20-week pilot education program to improve community
knowledge and mosquito control with participatory and non-chemical approaches in Sri Lanka. They
evaluated their program effectiveness using pre-educational and post-educational surveys in two
intervention and two comparison villages. Their controlled intervention shows that participatory
education program led to improved knowledge of mosquito ecology and disease epidemiology, changes in
agricultural practices, and an increase in environmentally sound measures for mosquito control and
disease prevention. The success of the intervention was attributed to four ‘human ecology’ characteristics:
a community-based education that enhanced residents’ understanding of the mosquito-borne disease
problems in their own community, a participatory approach that allowed participants to gain hands-on
experiences with actions to be taken, using non-chemical measures that decreased environmental and
health risks in residential areas and paddy fields, and an approach that required no cost or extensive
instruments. Furthermore, this community-based approach suppressed the density of adult Anopheles in
the southwest monsoon season, though little impact was detected on Culex and Aedes densities (Yasuoka
et al., 2006b).
Vegetation and water management, however, are just one class of human behaviors that impact the
transmission and control of malaria. The links between ecosystem change, vector ecology and disease
epidemiology all depend critically on human density, gender ratio, immigration of non-immune people,
and knowledge, attitudes and practices primarily because they alter the pattern and frequency of human
contacts with vectors. Furthermore, a recent special colloquium of the International Society of Ecosystem
Health (Patz et al., 2004) suggests that malaria can be exacerbated by a broad array of land use drivers
and underlying human behavioral factors beyond changes to the biophysical environment. These include
movement of populations, pathogens, and trade; agriculture; and urbanization. Deforestation features
prominently in this review and is closely linked to many of these mechanisms.
Pattanayak et al. (2006b) underscore this behavioral aspect of malaria control and present four reasons
why it is important to understand the role of deforestation from a policy and planning perspective. These
include:
1. Deforestation is not merely the exogenous (remote control) removal of forest cover. It is the
beginning of an entire chain of activities, including forest clearing, farming, irrigation, livestock, and
non-timber forest product collection, that have ecological (vector habitat) as well as behavioral
(exposure and transmission) consequences for malaria.
2. Deforestation is an integral part of life and the landscape in many parts of the world with high malaria
rates (Donohue, 2003; Wilson, 2001). Consequently, sustainable forest management has become an
important policy goal, as donor agencies and local policy makers take a more integrated view of
people in the natural landscape. The resulting changes in land cover, as well as changes in how
people interact with the forest, have implications for malaria. Thus, conservation policies aimed at
slowing deforestation will impact malaria (Taylor, 1997; and Walsh et al., 1993).
3. Millions of rural households depend directly on a wide variety of forest products and services (Byron
and Arnold, 1999). By lowering local people’s natural wealth, deforestation can reduce household
capacity to invest in health care and pay for malaria prevention and treatment. At the same time,
deforestation may increase the wealth of other households, who will then be better able to avoid and
cure malaria.
4. Deforestation and malaria are central elements of the vicious cycle of poverty in rural areas of
developing countries. In simplistic terms, malaria could be considered to “cause” deforestation,
because malaria can make people poorer and poverty has been found to “cause” deforestation under
some conditions. In reality, the linkages are more complex and site-specific.
These ideas lead us to a human ecology framework for understanding the links between deforestation and
malaria. Human ecology involves the study of human–environment interactions and extends notions of
ecology and health by explicitly traversing boundaries between “nature and culture” and “environment
and society” (Parkes et al., 2003). Others have labeled these the ‘environmental health’ or the ‘ecology
and health’ (Aron and Patz, 2001) perspectives. As Parkes et al. (2003) clarify, ultimately all these fields
converge on three themes:
(a) integrated approaches to research and policy,
(b) methodological acknowledgment of the synergies between the social and biophysical environments,
(c) incorporation of core ecosystem principles into research and practice
Specific to malaria, we need a shift in the view of humans as passive or constant factors in malaria
epidemiology to a view in which people are very active factors (actors) in causing significant changes in
epidemiological patterns (Wessen, 1972; MacCormack, 1987). The centrality of human behavior is
confirmed by the number of instances in which human behaviors show up in Figure 1 in this chapter and
in the Patz et al. (2004) review.
4. Empirics of human ecology: Approach and evidence
In this section we present an initial attempt to examine the importance of human behaviors in malaria
transmission and control, and recognize the “active” (dynamic) aspects of human behavioral response.
Omitting behavioral responses from any analysis of malaria and ecosystem change would result in a
classic case of confounding. Human behavior in this case has all attributes of a confounding factor
because it is (a) correlated with the outcome and the risk factor, (b) not necessarily in the causal chain,
and (c) very likely to be unbalanced across the different levels of risks. As such behavioral confounders
can mimic the risk factor and mask the ecological relationship we are attempting to discover.
What does this mean in practical terms? If we are, for example, using cross-sectional or time series
variation in data on deforestation and malaria only, we will face what is labeled an “omitted variable”
problem in statistics/econometrics. This problem leads to biased inferences and inconsistent estimates of
policy parameters because the real cause is an omitted variable, e.g., the in-migration of susceptible sub-
populations. A second related and possibly more pernicious issue is that of endogeneity or reverse
causality (or simultaneity). Consider an example from Sawyer (1993) to better understand this
‘endogeneity’ bias. High rates of malaria can encourage forms of land use in which men work as day
laborers (in logging or ranching), allowing their wives and children to live in towns with relatively lower
threat of malaria, rather than establishing family farms. It in such a situation is often difficult to
disentangle the causal role of deforestation in malaria transmission.
To further investigate the empirical implications of these ‘behavioral’ or ‘human ecology’ models, we
offer two simple tests that are conducted at three different scales. First, we compare a simple regression
model of malaria and deforestation (‘naïve model’) to model including linear behavioral controls (‘linear
controls model’). Second, we compare the same naïve model to one where the behavioral factors are used
as determinants of deforestation or the ‘endogenous’ risk exposure. Behavior in this case is an instrument
for the deforestation risk (the instrumental variable [IV] model). Economic theory provides one basis for
identifying variables that can explain deforestation and thus serve as instruments (Sills and Pattanayak,
2006).
Arguably the naïve model is a bit of straw man, but it allows us to investigate the importance of a human
ecology strategy. We conduct these evaluations at three scales: a micro analysis of child malaria and
community deforestation (case from Indonesia), a meso analysis of regional malaria and regional
deforestation (case from Brazil), and a macro analysis of national malaria and deforestation. Data
limitations preclude the use of accurate behavioral indicators and force us to use proxy variables.2 Thus,
our analysis should be considered as preliminary, and therefore illustrative of the overarching human
ecology approach proposed here.
4a. Macro analysis using global data from 120 countries
In this case study, we examine the macro level correlation of malaria and forest using a global data set.
Pattanayak et al. (2006) describe the combination of data from 5 sources to produce a global malaria
dataset and use it to examine how disease prevention behaviors respond to disease levels. The World
Health Organization’s Global Health Atlas provides data on a range of malaria variables, including the
number of cases, for up to 195 countries from 1990 to 2004. The World Development Reports provide
data on forest cover in 1990 and the annual rate of increase from 1990 to 2000. We obtain behavioral
2 If the measurement error (because of the use of proxy variables) is of the classical variety – i.e., uncorrelated with
the regression error – then we would face an attenuation bias (make the correlation seem smaller than it is). Data
weakness is not the main problem here. Instead, we would argue that the paucity of good data is ultimately because
of inadequate attention to the human ecology perspective in empirical analysis – both statistical estimation and
numerical simulation.
proxies from three other sources. First, data from the 2001 Human Development Report (HDR) provides
measures of economic conditions (per capita GDP) and social conditions (adult literacy rates, educational
enrollment rates, and life expectancy). Second, Kaufmann et al. (2003) provide data on political stability,
voice and accountability, and control of corruption. Additionally, we also include a malaria ecology index
to capture vector ecology and climatic factors (Kiszewski et al., 2004). This index combines climatic
factors (e.g., rainfall and precipitation), the presence of different mosquito vectors, and the human biting
rates of these vectors to proxy for mosquito transmission. This index captures the ecological conditions
with the strongest influence on the intensity of malaria prevalence and can therefore predict the actual and
potential stability of transmission. Descriptive statistics and other details of the data compilation and
synthesis are included in Pattanayak et al. (2006).
Our key variable is the number of malaria cases in a country in the 1996-2000 period. Various variables
(malaria cases, malaria ecology index, and GDP index) are converted into natural logarithms to reduce
scale differences, improve linearity and pull in outliers. Median regression methods are used. Results of
the three models – naïve, linear controls, and IV are presented in columns 2, 3 and 4 of Table 2 (Panel 1).
We report the coefficient on the deforestation variable, the probability value (p.value) associated with this
coefficient and the overall significance of the model. The regression coefficient reflects the size and sign
of the correlation with malaria incidence. The p.value reflects the statistical significance of the correlation
(i.e., less than 0.1 is suggestive of a relationship).
The naïve model is statistically significant and explains about 41% of the variation. We also find
confirmation of our key hypothesis: annual rate of forest cover increase (during the 1990 to 2000 period)
is negatively correlated with malaria incidence in the 1996-2000 period: more deforestation is positively
correlated with higher levels of malaria.
The linear-controls model (where we account for potential confounding due to GDP, school enrollment,
voice and accountability, and stability of the governmental institutions) is also statistically significant and
explains about 52% of the variation in the malaria cases. We also find that deforestation is positively
correlated with malaria, except now the size of this correlation is twice as big.
Finally, the IV model is also significant and explains about 54% of the variation. In this model, first
behavioral variables are used to predict deforestation, and then the predicted deforestation is used to
explain malaria. Again we see that the deforestation variable is positive correlated with malaria, but now
the size of this coefficient is almost 4 times as big as the naïve model – providing a statistically significant
evidence of a much stronger correlation between the disease and exposure change due to deforestation.
4b. Meso (regional) analysis using the case of 480 Brazilian micro-regions
In this case study, we examine the hypothesis regarding the regional level correlation of malaria and
forest cover. We use a cross-sectional data set of approximately 490 Brazilian micro-regions, which in the
Brazilian context is anything between one to twelve counties. The malaria data comes from DATASUS
(website). It is reported in terms of 1000 inhabitants, and represents hospital morbidity over the 1992-
2000 period. Climate is represented by long run temperature and rainfall (averaged over several years) in
the 490 micro-regions, based on weather stations that are located approximately one per micro-region.
Census data (website address) on housing, population, education levels, income, medical care (proxied by
number of doctors and hospital beds) and infrastructure (percent of the households connected to water,
sanitation, and all-weather roads) is for 1991. Forest cover and vegetation data of the same vintage are
from IPEA and protected area data is from INPE, both Brazilian data agencies. Pattanayak et al. (2006b)
present additional detail on the compilation and use of this data in analysis.
Instead of dwelling on the details on the analysis, we focus on the key results using the structure from the
previous case study. The naïve model (including some ecological controls for weather) is statistically
significant and explains 46% of the variation. First, we see that micro-regions with higher forest cover
have lower rates of malaria, all things considered. Second we find that micro-regions with higher
deforestation (in the 1985 to 1995 time period) have greater rates of malaria.
The linear controls model (where we account for potential confounding due to demographics, income,
infrastructure, and institutions such as protected areas) is statistically significant and explains 56% of the
variation. First, we find that micro-regions with higher deforestation have greater rates of malaria – with
the correlation that is significantly larger than the naïve model coefficients (almost twice as large).
Second, micro-regions in the Amazon with conservation units have lower malaria rates for a given level
of deforestation.
Finally, the IV model uses a variety of regional factors – presence of protected area, distance to highway
and to state capital, population, size and location of the micro-region – as instruments for deforestation in
the micro-regions. The overall model is significant. Now the size of the deforestation coefficient is almost
3 times as big as the linear-controls model. The results are consistent across the three models (i.e.,
deforestation is correlated with more malaria), but the sizes of the estimated coefficient are much larger
(3-6 fold) in the models that include proxies for human behavior.
4c. Micro analysis using data on 340 children from Flores, Indonesia
Malaria is highly contextual, with incidence and transmission depending on local conditions,
perturbations, and catastrophes. Thus, household or community-level multi-factor research is perhaps
best suited to incorporate the diversity and heterogeneity of the ecological, epidemiological, and
economic phenomena surrounding malaria. This case study examines the evidence on whether
deforestation causes child malaria in the setting of Ruteng Park on Flores Islands in eastern Indonesia.
The data for this analysis are drawn from a household survey in the Manggarai district of Flores,
Indonesia in 1996 around a protected area (Ruteng Park), established to conserve biodiversity. The survey
and accompanying secondary data collection generated household data on wealth, housing quality, and
number of adults, as well as individual data on age, gender, occupation, education and disease history
during the twelve months prior to the survey. GIS is used to combine environmental statistics, including
the amount and extent of primary and secondary forest cover at the village level, with the survey data and
secondary data on public infrastructure such as sub-regional health care facilities. The sample includes
approximately 340 kids under the age of 5. Given the binary nature of the data on malaria in kids under
the age of 5, we estimate and report probit models of child malaria. Pattanayak et al. (2005) include
details.
Starting with the naïve model, we find that the overall model is significant and, this being micro data,
explains only about 6% of the variation. We find that the extent of protected (primary) forest cover is not
statistically related to malaria, whereas the extent of disturbed (secondary) forest is positively correlated
with malaria rates.
The linear-controls model accounts for potential confounding due to various individual, household and
village characteristics. The overall model is significant, now explaining about 15% of the variability in
the malaria data. As in the naïve model, the extent of disturbed forests is positively correlated with
malaria (although now the coefficient is twice as big as before). Most interesting, we now confirm our
key hypothesis that the extent of protected forests is indeed negatively correlated with malaria incidence.
Finally, the IV model uses a variety of community level factors – distance to highway, population, village
size, elevation, and rainfall – as instruments for protected and disturbed forest cover around the villages.
The overall model is highly significant. Most crucially, now the sizes of the coefficients are almost 3
times as big as the linear-controls model. Malaria in little children is highly positively correlated with the
extent of disturbed forests and negatively correlated with the extent of protected forests.
5. Concluding thoughts
Vector-borne diseases such a malaria wreak havoc on the lives of many millions of people in poor,
tropical countries, partly because these regions are exposed to environmental conditions such
deforestation, livestock rearing, irrigated farming, road construction, and dam-building that encourage
vector abundance and disease transmission. We argue that it is critical to focus on the deforestation
linkage because it is the beginning of an entire chain of activities that affect malaria risks; can trigger
behavioral changes due to accompanying increases or decreases in wealth; can lock communities into a
vicious cycle of poverty, illness and environmental degradation; and is an integral part of the landscape
and therefore of donor agencies and policy maker focus. Recognizing that deforestation often precedes
many other relevant land use changes (particularly conversion to agriculture), taking deforestation as a
starting point allows us to look at the impact of other elements in the “matrix of transformations.” As such
it serves as a broad indicator of change in the ecology of infectious disease paradigm. This lead us to
recommend a human ecology that focuses on the role of humans in land use change as well as in a variety
of behaviors to prevent (e.g., sleep under nets, take prophylaxis) and treat (e.g., seek medical care, follow
the drug regimen) malaria. We then review the implications of this framework change for empirical
research and application – both in data collection and analysis and inference.
The empirical case studies draw attention to the role of socio-economic determinants of malaria and
importance of including behavioral variables in empirical models of malaria incidence and prevalence.
They illustrate how omitting behavioral factors from the analysis can lead to erroneous and biased
interpretations regarding the nature of ecosystem changes and disease transmission – the size, sign, and
statistical significance of regression coefficients can be wrong. In general, they are intended to highlight
different elements of human-induced ecosystem change, disease outcomes, and economic causes and
consequences.
What we have not discussed is the inherent dynamics of coupled natural and social systems. In a recent
paper, for example, Pattanayak et al. (2006a) analyze global and micro data to show that malaria
prevention behaviors depend on malaria prevalence. They find that households and countries engage in
greater degree of prevention if they face high rates of malaria and fewer prevention behaviors if they
confront low rates of malaria. That is, the causal arrow can also flow in the other direction (such an arrow
is shown as a dotted arrow in Figure 1, typically missing from most assessments). This logical feedback
and dynamic between prevention and prevalence suggests that it is insufficient and inappropriate to model
and consider socio-economic behaviors as something outside the malaria infection and transmission
process. Behavior and its determinants are part and parcel of the ecology and epidemiology and must be
built into the analysis and planning.
In fact, it is safe to say that many of these findings hold for a general class of vector-borne infectious
diseases such as dengue, leishmaniasis, hantavirus pulmonary syndrome, schistosomiasis, filariasis, lyme
disease, onchocerciasis and loiasis. Space limitations preclude a comprehensive discussion of these
diseases (for additional details, see Wilson [2001] and tables 4 and 5 in Colfer et al. [2006], for example).
As suggested in Figure 1, ecosystem changes influence the emergence and proliferation of these diseases
by altering the ecological balance and context within which disease hosts or vectors and parasites breed,
develop and transmit diseases (Patz et al 2000). For example, deforestation is often followed by water
resources development and livestock management, which open up numerous possibilities for disease
risks.
Moreover, the simultaneity between prevalence and prevention discussed previously Pattanayak et al.
(2006a) only points to the proverbial tip of the dynamic that is inherent in coupled natural and social
systems. As Hammer (1993) suggests, in the case of malaria, very little is known about the inter-related
dynamics of ecosystem changes, vector density and infectivity, development of immunity and resistance
(to pesticides and drugs) and human response. Wiemer’s (1987) case of schistosomiasis in China and
Gersovitz and Hammer’s (2005) model of malaria prevention and treatment are early attempts to examine
these dynamics through mathematical simulations. Much more conceptual work is needed before
ecosystem change dynamics can be incorporated into such models. Empirical research must test
hypotheses about the nature and magnitude of these relationships and generate statistical parameters that
can then be used for policy scenario analysis.
In the interim, however, the human ecology approach to public health can take root and thrive through the
conduct of systematic economic and health impact assessments of forest policies. Such evaluations need
to be inter-disciplinary longitudinal studies, with at least the following features:
1. It is impossible to design and implement a rigorous study and make credible inferences without a
clear understanding of the policy scenario. Specificity of the policy scenario – be it a project at a site,
a program that includes a collection of projects, or a nation/region-wide policy – allows the analyst to
understand the mechanism of disease transmission and economic impacts in terms of ‘modifiable
causes’.
2. With a clear scenario, it is then possible to design rigorous evaluations to infer ‘causal policy
impacts’. These are typically through randomized assignment of the program or a quasi-experimental
design that includes data collection in program and control (non-program) sites during various stages
of program implementation, including baseline (pre-program) and endline (post-program) data.
3. The credibility of the resulting evaluation will ultimately ride on the quality of the data and the rigor
and care in data analysis. For a study of this type, outcomes variables include indicators of health,
wealth, and the environment. Extent of forest cover and forest condition are among the key
explanatory variables. Other explanatory variables include socio-economic, demographic,
environmental, health, and public health policy indicators. The challenge in empirical work is to
identify robust measures of these variables and separate independent and dependent variables. The
multiple channels for feedback between malaria, deforestation and poverty suggest that these
variables would be dependent variables in some specifications, and independent variables in other
specifications and data sets.
4. Although researchers can employ an array of sophisticated techniques to remedy defects in available
data, clearly “prevention” in the form of careful data collection is superior to “cure” in the form of ad
hoc statistical fixes. Longitudinal data sets – and particularly panel data sets – are key to addressing
at least three critical issues in the types of research proposed here: heterogeneity, endogeneity, and
dynamics or mobility (Ezzati et al., 2005). Ideally, data should be collected at several scales, ranging
from individual level health and demographic data, to household level economic information, to
community and regional level environmental statistics and policy factors.
The human ecology approach proposed in this chapter that is built on these conceptual and empirical roots
can be used for at least two practical purposes (Pattanayak et al. 2006c). First, it can help organize the
conceptual links between coupled natural and socio-economic systems and serve as a platform for
generating testable hypothesis and policy parameters. Such efforts are critical for understanding the
ecological, entomological, epidemiological and economic aspects of deforestation, malaria, and their
behavioral underpinnings. Second, it will be vital for building decision analysis and scenario simulation
tools (Kramer et al., 2006), which rely on estimated parameters, for formulating integrated strategies that
cut across health, environment and economic sectors to address the broad idea of ecosystem change and
disease control. Scenario simulation can for example inform the design of surveillance and monitoring
framework necessary to detect changes in the environment, vector density, human migration and
behavior, and incidence of diseases in order to both contain vector-borne diseases and prevent epidemics.
6. References3
Amerasinghe FP, Amerasinghe PH, Peiris JSM, Wirtz R. 1991. Anopheline ecology and malaria infection
during the irrigation development of an area of the Mahaweli project, Sri Lanka. Am J Trop Med
Hyg 45:226-235.
Amerasinghe FPA. 2003. Irrigation and mosquito-borne diseases. J Parasitol. Special Edition. Selected
Papers of the 10th International Congress of Parasitology.
Berkman, LF, and I Kawachi (eds.) 2000. Social Epidemiology. New York, Oxford University Press
2000.
Byron, N. and M. Arnold. 1999. “What Futures for the People of the Tropical Forests?” World
Development 27(5): 789-805.
Campbell-Lendrum D, Molyneux D et al (2005) Ecosystems and Vector-borne Disease Control. Chapter
12 in Ecosystems and Human Well-being: Policy Responses, Volume 3. Millennium Ecosystem
Assessment. Pages 353-372.
Cheong WH. 1983. Vectors of filariasis in Malaysia. In: Filariasis (Mak JW, eds). Kuala Lumpur:
Institute for Medical Research. Bulletin No.19, 37-44.
Chivian E, eds. 2002. Biodiversity: Its Importance to Human Health, Interim Executive Summary.
Boston, MA: Harvard Medical School.
Colfer, C. J. P., D. Sheil, and M. Kishi. 2006. Forests and human health: assessing the evidence. CIFOR
Occasional Paper. 111. Bogor, Indonesia, Center for International Forestry Research.
Conn JE, Wilkerson RC, Segura MN, de Souza RT, Schlichting CD, Wirtz, RA, et al. 2002. Emergence
of a new neotropical malaria vector facilitated by human migration and changes in land use. Am J
Trop Med Hyg 66:18–22.
Corvalan C, Hales S, McMichael A et al (2005a) Ecosystems and Human Well-being: Human Health
Synthesis. Millennium Ecosystem Assessment. 63 pages.
Corvalan C, Hales S, Woodward A et al (2005b) Consequences and Options for Human Health. Chapter
16 in Ecosystems and Human Well-being: Policy Responses, Volume 3. Millennium Ecosystem
Assessment. Pages 467-486.
Donohue, M. 2003. “Causes and health consequences of environmental degradation and social injustice.”
Social Science and Medicine 56: 573-587.
Ezzati, M., J. Utzinger, S. Cairncross, A.J. Cohen and B.H. Singer. 2005. “Environmental risks in the
developing world: exposure indicators for evaluating interventions, programmes, and policies”.
Journal of Epidemiology and Community Health. 59: 15-22.
Gersovitz, M., and J. Hammer. 2005. Tax/subsidy policies toward vector-borne infectious diseases.
Journal of Public Economics 89 (4): 647–674
Grillet ME. 2000. Factors associated with distribution of Anopheles aquasalis and Anopheles oswaldoi
(Diptera: Culicidae) in a malarious area, northeastern Venezuela. J Med Entomol 37:231-238.
Hammer, J., 1993. Economics “The economics of malaria control”. The World Bank Research Observer.
8 (1): 1 – 22.
3 Citations in Table 1 are presented in Yasuoka and Levins (forthcoming).
Hay, S.I., Guerra, C.A., Tatem, A.J., Noor, A.M., Snow, R.W. 2004 the Global Distribution and
Population at Risk of Malaria: Past, Present and Future. The Lancet—Infectious Diseases 4: 327–
336.
Ijumba JN, Lindsay SW. 2001. Impact of irrigation on malaria in Africa: paddies paradox. Med Vet
Entomol 15:1-11.
Karla NL. 1991. Forest Malaria Vectors in India: Ecological Characteristics and Epidemiological
Implications. In: Forest Malaria in Southeast Asia (Sharma VP, Kondrashin AV, eds). New Delhi:
WHO/MRC, 93-114.
Kates, R.W., W.C. Clark, R. Corell, J.M. Hall, C.C. Jaeger, I. Lowe, J.J. McCarthy, H.J. Schellnhuber, B.
Bolin, N.M. Dickson, S. Faucheux, G.G. Gallopin, A. Grubler, B. Huntley, J. Jager, N.S. Jodha, R.E.
Kasperson, A. Mabogunje, P. Matson, and H. Mooney. 2001. Sustainability science. Science
292(5517): 641-642.
Kaufmann, D., Kraay, A., and Mastruzzi, H. 2003. Governance Matters III: Governance indicators for
1996-2002. Washington, DC: World Bank.
Keiser, J., B.H. Singer, and J. Utzinger. 2005. Reducing the burden of malaria in different eco-
epidemiological settings with environmental management: a systematic review. Lancet
Infectious Diseases 5: 695-708.
Kiszewski, A., Mellinger, A., Spielman, A., Malaney, P., Sachs, S.E. and Sachs, J. 2004. A Global Index
of the Stability of Malaria Transmission. American Journal of Tropical Medicine and Hygiene
70(5): 486-498.
Kondrashin AV, Jung RK, Akiyama J. 1991. Ecological aspects of forest malaria in Southeast Asia. In:
Forest Malaria in Southeast Asia (Sharma VP, Kondrashin AV, eds). New Delhi: WHO/MRC, 1-28.
Konradsen F, Amerasinghe FP, van der Hoek W, Amerasinghe PH, eds. 2000. Malaria in Sri Lanka,
Current Knowledge on Transmission and Control. International Water Management Institute.
Kramer, RA, KL Dickinson, VG Fowler, ML Miranda, CM Mutero, KA Saterson, and JB Weiner. 2006.
Decision Analysis as an Integrative Tool for Improved Malaria Control Policy Making. Duke
University Working Paper.
Lindsay, S.W., and M. Birley. 2004. Rural development and malaria control in Sub-Saharan Africa.
EcoHealth 1: 129-137.
Martens P. 1998. Health & Climate Change: Modeling the Impacts of Global Warming and Ozone
Depletion. London: Earthscan.
MacCormack, CP., 1984. “Human Ecology and Behavior in Malaria Control in Tropical Africa”. Bulletin
of the World Health Organization. 62: 81-87 Suppl. S.
McMichael, AJ. 2001. Human Frontiers, Environments and Disease: Past Patterns, Uncertain Futures.
Cambridge, UK:Cambridge University Press.
McMichael, A., Patz, J. and S. Krovats, 1998. “Impacts of Global Environmental Change on Future
Health and Health Care in Tropical Countries”. British Medical Bulletin 54(2): 475-488.
Molyneux DH. 1998. Vector-borne parasitic diseases – an overview of recent changes. Int J Parasitol
28:927-934.
Oakes, JM and JS Kaufman (eds). 2006. Methods in Social Epidemiology. John Wiley & Sons, Inc., San
Francisco, California. 504 pages.
Parkes, M., R. Panelli, and P. Weinstein. 2003. “Converging Paradigms for Environmental Health
Theory and Practice.” Environmental Health Perspectives 111 (5): 669-675.
Pattanayak, S.K, C. Poulos, K. Jones, J-C. Yang and G. Van Houtven. 2006a. “Economics of
Environmental Epidemiology”. RTI Working Paper. Research Triangle Park, North Carolina.
Pattanayak, S.K., M. Ross, C. Timmins, B. Depro, K. Jones, and K. Alger. 2006b. Climate Change,
Human Health, and Biodiversity Conservation. Presented at U.S.E.P.A conference on
Multidisciplinary Approach to Examining the Links Between Biodiversity and Human Health.
Washington D.C., September.
Pattanayak, S.K., K. Dickinson, C. Corey, E.O. Sills, B.C. Murray, and R. Kramer. 2006c.
“Deforestation, Malaria, and Poverty: A Call for Transdisciplinary Research to Design Cross-
Sectoral Policies”. Sustainability: Science, Practice and Policy. 2(2): 1-12
Pattanayak, S.K., C.G. Corey, Y.F. Lau, and R. Kramer. 2005. “Conservation and Health: A
microeconomic study of forest protection and child malaria in Flores, Indonesia.” RTI Working
Paper. Research Triangle Institute, North Carolina.
Patz J, Confalonieri U et al (2005) Human Health: Ecosystem Regulation of Infectious Diseases. Chapter
14 in Ecosystems and Human Well-being: Current State and Trends, Volume 1. Millennium
Ecosystem Assessment. Pages 391-415.
Patz JA, Daszak P, Tabor GM, Aguirre AA, Pearl M, Epstein J, et al. 2004. Unhealthy landscapes: Policy
recommendations on land use change and infectious disease emergence. Environ Health Perspect
112:1092-8.
Patz JA, Graczyk TK, Gellera N, Vittor AY. 2000. Effects of environmental change on emerging parasitic
diseases. Int J Parasitol 30:1395-405.
Prothero RM. 1999. Malaria, Forests and People in Southeast Asia. Singap J Trop Geogr 20:76-85.
Reiter P. 2001. Climate change and mosquito-borne disease. Environ Health Perspect 109(S1):141-161.
Rose G. 1985. Sick individuals and sick populations. International Journal of Epidemiology 14 (1):32–38.
Rosenberg R, Andre RG, Somchit L. 1990. Highly efficient dry season transmission in malaria in
Thailand. Trans R Soc Trop Med Hyg 84:22-28.
Rosenberg R, Maheswary NP. 1982. Forest Malaria in Bangladesh II: Transmission by An. dirus. Am J
Trop Med Hyg 31:13-191.
Sawyer, D. 1993. Economic and social consequences of malaria in new colonization projects in Brazil.
Social Science and Medicine 37 (9): 1131–1136.
Sharma VP, Kondrashin AV, eds. 1991. Forest Malaria in Southeast Asia. New Delhi: WHO/MRC.
Sills, E. O., and S. K. Pattanayak. 2006. “Tropical Tradeoffs: An Economics Perspective on Tropical
Deforestation”. Chapter 5 n S. Spray and M. Moran (eds.), Tropical Deforestation. Rowman and
Littlefield Publishers Inc. Pages 104-128
Singh YP, Tham A. 1990. Case history of malaria control through the application of environmental
management in Malaysia. WHO/WBC/88.960.
Smith, KR, Corvalán, CF and T. Kjellstrom, 1999. “How Much Global Ill Health is Attributable to
Environmental Factors?” Epidemiology 10: 573-584.
Suvannadabba S. 1991. Deforestation for Agriculture and its Impact on Malaria in Southern Thailand. In:
Forest Malaria in Southeast Asia (Sharma VP, Kondrashin AV, eds). New Delhi: WHO/MRC, 221-
226.
Taylor D. 1997. Seeing the forests for more than the trees. Environ Health Perspect 105:1186-1191.
Walsh JF, Molyneux DH, Birley MH. 1993. Deforestation: effects on vector-borne disease. Parasitology
106(suppl):55-75.
Wiemer, Calla. 1987. "Optimal Disease Control through Combined Use of Preventive and Curative
Measures." Journal of Development Economics 25: 301-19.
Wessen, A. F. 1972. “Human Ecology and Malaria”. American Journal of Tropical Medicine and
Hygiene. 21(1): 658-662.
Wilcox, B.A., and R.R. Colwell. 2003. Emerging and reemerging infectious diseases: biocomplexity as
an interdisciplinary paradigm. EcoHealth 2: 244-257.
Wilson, ME. 1995. “Infectious Diseases: an Ecological Perspective.” British Medical Journal, 311(7021):
1681-1684.
Wilson, ML. 2001. “Ecology and Infectious Disease.” In Ecosystem Change and Public Health J. Aron,
and J.A. Patz. (eds.) The Johns Hopkins University Press: Baltimore, MD: 285-291.
Yasuoka J, Levins R.(forthcoming). Impact of Deforestation and Agricultural Development on
Anopheline Ecology and Malaria Epidemiology. American Journal of Tropical Medicine and
Hygiene.
Yasuoka, J., T. W. Mangione, A. Spielman, and R. Levins. 2006a. "Impact of education on knowledge,
agricultural practices, and community actions for mosquito control and mosquito-borne disease
prevention in rice ecosystems in Sri Lanka." Am J Trop Med Hyg 74(6): 1034-42.
Yasuoka, J., R. Levins, TW Mangione and A. Spielman. 2006b. Community-based rice ecosystem
management for suppressing vector anophelines in Sri Lanka. Transactions of the Royal Society of
Tropical Medicine and Hygiene 100: 995—1006
Table 1. Ecosystem change and malaria
Density decrease Density increase Increased human contacts
Deforestation/
Agricultural
development
Country/
Region Species Malaria Species Malaria Species Malaria
References
Thailand An. dirus - Taylor 1997
Nepal An. minimus An. fluviatilis Sharma VP. 2002
India An. fluviatilis Kalra 1991
An. barbirostris An. annularis +
An. jamesii +
An. nigerrimus +
An. subpictus +
Sri Lanka
A peditaeniatus ?
Amerasinghe and Ariyasena 1990,
Konradsen et al. 2000
Deforestation
Sahel, Africa An. funestus Mouchet et al. 1996
An. labranchiae -
An. sacharovi -
Land exploitation/
pollution Mediterranea
n
An. superpictus -
Coluzzi 1992
Cacao plantation Trinidad An. bellator Downs and Pittendrigh 1946, Ault 1989
Thailand An. dirus - An. minimus + Bunnag et al. 1978, Sornmani 1987,
Prothero 1999
Cassava
Thailand An. dirus - Rosenberg et al. 1990
Sugarcane Thailand An. dirus An. minimus + Sornmani 1972, Sornmani 1974, Bunnag et
al.1978
Coffee plantation
+ irrigation dams India An. fluviatilis - Kalra 1991
+ tree crops Thailand An. minimus + Suvannadabba 1991
Density decrease Density increase Increased human contacts
Deforestation/
Agricultural
development
Country/
Region Species Malaria Species Malaria Species Malaria
References
Tea plantation Sri Lanka A. culicifacies Jones 1951
Rubber Malaysia
An. maculatus +
Cheong 1983, Singh and Tham 1990,
Walsh et al. 1993
Thailand An. dirus +
Rosenberg and Maheswary 1982,
Prasittisuk et al. 1989, Rosenberg et al.
1990
+ Fruits
Thailand An. dirus Prasittisuk et al. 1989
+ Orchards Thailand An. dirus + Taylor 1997
China An. sinensis
Baolin 1988, Service 1989, van der Hoek
et al. 2001
Malaysia An. umbrosus - A. campestris + Ooi 1959, Sandosham 1970
Indonesia An. aconitus + Marwoto and Arbani 1991
Southeast
Asia An. dirus Kondrashin et al. 1991
Nepal An. fluviatilis A. culicifacies +
Walsh et al. 1993, Sharma et al. 1984,
Subedi et al. 2000, Reuben 1989
An. annularis An. jamesii
An. barbirostris An. subpictus
An. culicifacies
Sri Lanka
An. varuna
Amerasinghe et al. 1991, Konradsen et al.
2000
An. funestus
Rice
Africa An. gambiae Mouchet et al. 1996, Reiter 2001
Rice + maize Thailand An. dirus An. minimus Prasittisuk et al. 1989, Konradshin et al.
1991
Density decrease Density increase Increased human contacts
Deforestation/
Agricultural
development
Country/
Region Species Malaria Species Malaria Species Malaria
References
India An. culicifacies + Ault 1989, Amerasinghe et al. 1991
Afghanistan An. superpictus An.
pulcherrimus + Service 1989, Amerasinghe et al. 1991
An. arabiensis +
Africa An. gambiae + Service 1989, Amerasinghe et al. 1991
Sahara An. gambiae + Coluzzi 1984, Coluzzi 1992
Irrigation system
Guyana An. darlingi An. aquasalis + Ault 1989, Amerasinghe et al. 1991
Hydropower dam Sri Lanka An. culicifacies + Wijesundera 1988, Konradsen et al. 2000
Malaysia An. sundaicus + Ooi 1959, Walsh et al. 1993
Indonesia An. sundaicus + Marwoto and Arbani 1991
Clearing of
mangroves/
swamps for fish
pond or mining Indonesia An. sundaicus Ooi 1959, Sandosham 1970
Mining Thailand An. dirus Kondrashin et al. 1991
+ settlement Amazon An. darlingi + Marques 1987, Conn et al. 2002
Amazon An. darlingi Tadei et al. 1998, Tadei and Thatcher
2000, Conn et al. 2002
Indonesia A. balabacensis + Marwoto and Arbani 1991
A. balabacensis
Indonesia A. leucosphyrus
Marwoto and Arbani 1991
Settlements +
urbanization or
highway
construction
India An. stephensi + Kalra 1991
Source: Yasuoka, J., Levins, R. 2007. Impact of deforestation and agricultural development on anopheline ecology and malaria epidemiology. Am.
J. Trop. Med. Hyg.
Table 2. Empirics of ‘human ecology’ modeling of malaria and deforestation links
4a. MACRO naïve linear controls IV
annual forest increase -0.049 -0.089 -0.168
p.value (0.065) (0.013) (0.038)
ecology controls yes yes yes
behavioral controls no yes as IV
Pseudo RSq. 0.407 0.519 0.540
4b. MESO naïve linear controls IV
deforestation 7.89e-07 2.25e-06 6.99e-06
p.value (0.047) (0.000) (0.004)
ecology controls yes yes yes
behavioral controls no yes as IV
Pseudo RSq. 0.461 0.555 0.235
4c. MICRO naïve linear controls IV
log (primary forests) -0.062 -0.163 -0.382
p.value (0.497) (0.106) (0.046)
log (secondary forests) 0.234 0.401 0.609
p.value (0.006) (0.000) (0.008)
ecology controls yes yes yes
behavioral controls no yes as IV
Pseudo RSq. 0.055 0.153 0.153
Figure 1. Ecology of vector-borne diseases - impact of human activities