OPTIMAL RESETTLEMENT STRATEGIES TO COPE WITH
LANDSLIDE RISK ON MOUNT ELGON, EAST UGANDA
By
P. Vlaemincka, F. Vanderhoydonksa, M. Isabiryeb and L. Vrankena
a
Department of Earth and Environmental Sciences, KU Leuven, Belgium
b
Department of Natural Resources Economics, Busitema University, Uganda
In light of the VLIR-UOS South Initiative Project:
“An Integrated Assessment of Landslides at Mount Elgon, Uganda: Hazards,
Consequences and Opportunities”
1
Optimal Resettlement Strategies to Cope with Landslide Risk on Mount
Elgon, East Uganda
ABSTRACT
Landslides have been recognized as a widespread phenomenon in the East African highlands,
having a great social, economic and geomorphological impact. Landslides significantly affect
livelihoods and income generation, reduce agricultural productivity and increase soil
degradation in Bududa district, East Uganda. In addition, the livelihoods of the poorest are
most likely to be adversely affected by environmental change. Traditionally, social and
economic losses due to landslides are reduced by means of effective planning and
management such as physical measures and development of warning systems. However, due
to the local situation of high population density and land shortage it becomes clear that these
traditional measures are insufficient to offer a long-term solution for the problem. Therefore,
our paper investigates whether a well-managed long-term resettlement strategy could be a
feasible coping strategy for the affected population. We use a discrete choice experiment to
examine whether a resettlement is a feasible coping strategy to mitigate landslide risks on
Mount Elgon, and if so under which conditions and compensatory schemes. Besides we
investigate whether the willingness to resettle depends on the probability of a landslide
occurring. Our study provides the first analysis of resettlement related preferences of people
that are affected by environmental change. Our results enable us to give valuable policy
advice for different resettlement policies which are not yet implemented in a very-cost
effective manner.
Key words: Choice experiment, community supported resettlement, landslide risk mitigation,
rural development, agricultural productivity
2
Optimal Resettlement Strategies to Cope with Geological Mobility on
Mount Elgon, East Uganda
1. INTRODUCTION
Landslides have been recognized as a widespread phenomenon in the East African
highlands, having a great social, economic and geomorphological impact (Claessens et al.,
2007; Knapen et al., 2006). The entire East African Highlands have been categorized as an
inherently susceptible region (Glade and Crozier, 2004). Both Claessens et al. (2007) and
Knapen et al. (2006) categorize the high annual rainfall, the steep slopes, deforestation, and
high weathering rates and slope material with a low shear strength or high clay content as the
main preparatory causal factors for mass movements. These are caused by the underlying
driver of an increasing population pressure leading to slope disturbance, inconsiderate
irrigation and deforestation.
Landslides have the potential to significantly affect livelihoods and income generation,
reduce agricultural productivity and increase soil degradation (Shiferaw, 2002). In addition,
the livelihoods of the poorest are most likely to be adversely affected by environmental
change. Many authors show that the unprecedented increase in the number and ferocity of
disasters is exacerbated by climate change coupled with an increase in the population
occupying high risk areas (Rukundo et al., 2014; Black et al., 2011; Guterres, 2009). Since
poor people tend to live on marginal lands that are most subject to environmental hazards,
they are most likely to be affected by small changes in climate variability (UNDP, 2004;
FAO, 2000). Migration as a strategy to cope with these increased hazards requires resources
that the poor and vulnerable in particular often do not possess (Lübken, 2012). Therefore in
this paper we look at resettlement as a key element for managing environmental and other
risks in Bududa district, East Uganda, and not as a problematic outcome of global
environmental change (Black et al. 2011)
Traditionally, social and economic losses due to landslides are reduced by means of
effective planning and management such as physical measures and development of warning
systems (Dai et al., 2002). Besides, the instability of the slopes in Bududa district could
partly be reduced by tempering the human impact and reforestation with deep-rooted trees
(Claessens et al. 2007). However, due to the local situation of high population density and
land shortage it becomes clear that these traditional measures are insufficient to offer a long3
term solution for the problem (Claessens et al., 2007; Knapen et al., 2006). Therefore, our
paper investigates whether a well-managed long-term resettlement strategy1 could be a
feasible coping strategy for the affected population. In this study, we use a discrete choice
experiment to examine whether a resettlement is a feasible coping strategy to mitigate
landslide risks on Mount Elgon, and if so under which conditions and compensatory
schemes. Besides we investigate whether the willingness to resettle depends on the
probability of a landslide occurring.
Our study provides the first analysis of resettlement related preferences of people that
are affected by environmental change. Our results enable us to give valuable policy advice
for different resettlement policies which are not yet implemented in a very-cost effective
manner. Through a well-designed resettlement scheme, government can limit the elements at
risk thereby eliminating the expected loss of a landslide in the future. Finally, since hazard
resettlement has the potential to marginalize environmental migrants, we stress the
importance that the government should not only look to how a resettlement may reduce the
landslide risk but should pay equal attention to eliminating any social vulnerability in
destination areas such that the solving of one problem does not create a bigger one (Stal and
Warner, 2009).
2. LOCAL SITUATION
2.1 Description of area and landslides
The 274 km² research area of Bududa district is located in eastern Uganda on the
southwestern foot slopes of the extinct Mt. Elgon volcano, 20 kilometers east of Mbale, a
large trade hub (Figure 1). Bududa district was created in 2006 when it was separated from
Manafwa district. The altitude ranges from 1300 to 2850m a.s.l. and the district has a wet
tropical climate. The high average rainfall is 1800 mm with two separated rain seasons, one
from March to June and the second from August to November. The average annual
temperature is 23°C (BDPU, 2012) and is more or less constant the whole year long.
Topographic (steep slopes), climatic (high rainfall) and soil conditions together with
human presence (cultivation, deforestation and excavation for housing, irrigation, and foot
paths) make of Bududa district a landslide prone area (Knapen et al., 2006). Most landslides
occur on the east and north orientated slopes (dominant rainfall direction) with a rather small
1
Claessens et al., 2007 and Knapen et al., 2006 also encourage family planning as a mitigation strategy in order
to make the fast population growth take a turn and the provision of policies that provide alternative livelihoods.
4
critical slope of 14°. Based on soil characteristics and past landslides experiences, the district
can be divided in three zones (Kitutu et al., 2009) (Figure 1). The Central Bukigai zone is
characterized by a carbonatite dome underneath the soils which is highly stable due to high
cohesion of cementing minerals such as calcium carbonate (Kitutu et al., 2009). Landslides
hardly occur in this zone. The second zone is the Western Bududa-Bushika zone. Different
soil types are identified in this zone: Cambisols, Nitisols, Acrisols and Lixisols (Deckers et
al, 1998).
Landslides occur, but are relatively rare. However they contribute significantly to the
landside problem due to their large dimensions and the high population density of the area
(Knapen et al., 2006). In the Eastern Bukalesi zone, soils have higher clay contents which lead
to higher saturation rates of water. Landslides occur rather frequently and are shallower than
in the Western zone because the parent material is nearer to the surface.
2.2 Socio-economic impact of landslides
Landslides have a disastrous effect on the livelihoods of the farmers in Bududa district
since the beginning of the twentieth century (Knapen et al., 2006; Jenkins, et al. 2011;
Rukundo et al., 2014). Landslides are the main factor in the loss of income and land for local
farmers in Bududa district for example the loss of valuable coffee plantation land
(Gorokhovich et al., 2013). In addition, the debris of landslides imposes the government with
direct economic costs related to the reconstruction of bridges, roads, dams and the obligatory
funding of the disaster relief aid and displacement. On top of that, the indirect costs such as
decreased water quality and reduced land fertility can easily surpass these direct costs in
magnitude (Knapen et al., 2006).
Generally, the catastrophic landslides are triggered following extreme rainfall events
attributed to global weather patterns associated with the El Niño Southern Oscillation. Knapen
et al. (2006) describe the type of chain reaction that landslides can have on the livelihoods of
Bududa people: in 1997, landslides killed at least 48 people, erased the crops and dwellings of
885 families, made 5600 people homeless, reduced the amount of arable land causing landscarcity and property conflicts, polluted water supplies with a consecutively epidemic and hit
Manjiya County with a food-shortage. More recently, in 2010, a major landslide triggered by
heavy rains struck the village of Nametsi, killing over 300 people and affecting a population
of about 10,000 people which needed to be evacuated to a UN funded temporary camp in
Bulucheke and eventually triggered the first permanent landslide induced resettlement to
Kiryandongo (Rukundo et al., 2014; Gorokhovich et al. 2013; Jenkins et al., 2011).
5
2.3 Agriculture and population pressure
Bududa has an estimated population of 182 867 people (BDPU, 2012), living in 15
sub-counties and 1 town council. Agriculture is the most important economic activity for over
86% of the households living in the 16 different sub-counties (BDPU, 2012). The existing
farming system is mixed crop-livestock subsistence farming. The main crops grown are
banana, coffee, beans, cocoyam, cassava, sugarcane, onions and sweet potato. An average
population density of 952 persons/km2, raising up to more than 1300 persons/km2 in the
densely populated parishes in the west, makes available land per household rather small (35
m2). A population growth rate of 5.6 % since 1991 predicts even smaller land amounts per
household and more cultivation of unstable, steeper slopes (often steeper than 80%).
Agricultural pressure will also cause increased deforestation and excavations leading to a
further reduction of the stability of the slopes in the future. Besides, due to climate change, the
seasonal weather patterns are expected to become more extreme which will affect farmers’
income and food security strongly and the way their lands can be used (Oxfam, 2008). The
combination of these factors prevents people from abandoning the most landslide prone areas
and as a consequence the risk on damage by slope failure is projected to increase (Knapen et
al., 2006).
2.4 Resettlement as landslide risk mitigation strategy
Both society and environment affect each other’s resilience and vulnerability as
landslides present a threat to life and livelihood. The landslide risk can be expressed in the
following generic hazard-risk equation (Crozier and Glade, 2006):
Risk = hazard x vulnerability x elements at risk
This equation identifies the principal factors contributing to risk, where risk itself is defined as
the expected loss in a unit of time. First, hazard consists of the probability that a damaging
landslide of a given magnitude occurs. Second, vulnerability is expressed as the ratio of the
value of damage to the total value of the element or the amount of damage expected from the
specified landslide magnitude. Last, the elements at risk represent the valued attributes at risk
such as population, buildings, economic activities, public services utilities and infrastructure in the
area. The fact that risk is calculated by multiplication of the factors means that when one of
the independent factors is zero, the landslide risk becomes zero. In other words, if a landslide
occurs in an unpopulated area, the risk is zero (Crozier and Glade, 2006).
6
3. METHODOLOGY
3.1 Resettlement Assessment through Discrete Choice Experiments
To assess individual’s preferences for resettlement strategies we rely on a discrete
choice experiment (DCE) introduced by Louviere and Hensher (1982). This is a survey-based
stated preference elicitation method that allows modelling preferences for hypothetical
resettlement strategies and thereby revealing which strategies have a higher likelihood to be
community supported. In a DCE respondents are presented with several choice sets that
include alternative varieties of a good or service – in this case a resettlement strategy differentiated by their attributes and attributes levels, and asked to select their most preferred
alternative. A baseline alternative, corresponding to the status quo or ‘do nothing’ situation is
included in each choice set in order to be able to interpret the results in standard welfare
economic terms. At least one attribute of the alternative is systematically varied across
respondents so that preference parameters of an indirect utility function can be inferred
(Carson and Louviere 2011).
DCE rely on random utility theory which states that a respondent’s utility function is
comprised of a deterministic, observable component (V) and a random, unobservable
component (ε) (Christie et al., 2004):
Uijt = Vijt + εijt = βXijt + σiXijt + εijt ,
Where Uijt represents the utility a respondent i derives from choosing alternative j on choice
situation t, Xijt is a vector of k observed attributes for the resettlement strategies (k being the
number of attributes), β is the vector of preference parameters associated with the attributes,
σi is a vector of k standard deviation parameters, and εijt is a stochastic error term,
independently and identically distributed (iid) according to a Gumbel distribution (Louviere,
Hensher, &Swait, 2000). One choice set comprises of several resettlement scenarios.
Choosing one alternative over the others implies that the utility of the chosen alternative
exceeds the utility derived from the other alternatives (Ben-Akiva & Lerman,1985).
Respondents’ preferences are generally estimated through maximum likelihood in logit
models (Ben-Akiva & Lerman, 1985). Parameter estimates are derived from the loglikelihood function associated with the logit model.
7
3.2 Choice Experiment Design
To assess individual preferences for resettlement plans we conducted a survey that
included socio-demographic questions as well as a choice experiment. The construction of a
choice experiment includes three important stages: the identification of the attributes
describing the alternatives within each choice set, the identification of the attribute levels and
the experimental design. To identify the attributes, semi-structured interviews were conducted
among six sub-county chiefs, an officer of the Office of the Prime Minister and the National
Environmental Management Authority. In addition, four focus-group discussions (FGD), each
consisting of five women and five men, were organized. In Nametsi and Bukalasi sub-county
one FGD was organized while in Bibiita sub-county two FGD took place. Finally a workshop
was organized in Mbale with the aim of testing the feasibility of the attributes and the
attribute levels which were identified as important during the semi-structured interviews and
FGD. All sub-county chiefs as well as representatives of the Red Cross, the Ugandan Wildlife
Authority, the District of the Local Governments and the UNDP were invited to participate in
the workshop. A screening of the resettlement literature together with the information
gathered during the interviews, the FGD and the workshop allowed identifying six
resettlement attributes and the relevant attribute levels (Table 1). 24 choice cards divided over
2 blocks were designed using the NGene software and tested in a pilot survey which allowed
to further fine-tune the survey and choice experiment. An example of a choice card is shown
in Figure 2.
The duration of one interview including the choice experiment and the survey
questionnaire took on average 25 minutes. Each time a translator started with a small
introduction to emphasize that the research had nothing to do with the government. Second,
each attribute with its levels was thoroughly explained, to make sure that the respondent
understands everything. Just before the experiment started, the translator reminded the
respondents of 3 issues which were important to fill in the choice experiment correctly: (1) If
he chooses one of the 2 scenarios, he agrees to relocate to a new house in a new location. He
will remain owner of his land but isn’t allowed anymore to live on it. (2) If none of the 2
scenarios convinces him to consider a resettlement in the future, he can choose to stay under
the same conditions and face the landslide risk with the potential consequences in the future.
Choosing for the status-quo is not a choice against a resettlement, but only indicates that he
finds the given scenarios insufficient to consider a resettlement. (3) Finally we included a
‘cheap talk’ script. People sometimes answer in a way that they think will influence
8
government’s future decisions or give answers to favor the interviewer. A method to diminish
this kind of biases is cheap talk (Cummings and Taylor, 1999). Cheap talk is included to
convince the people to think carefully about whether they really would do what they say and
to answer as if this was a real choice with real consequences. To make sure that the person
understood the choice experiment, we first gave him a test card. The test card had the same
structure as the 12 choice cards which followed. If anything was unclear, he could ask the
interviewer for clarifications.
4. RESULTS
4.1 Descriptive statistics
During the period of August – October 2013, we interviewed a total of 307 household
heads in the entire Bududa district. The last population census was carried out in 2002 and
population predictions were established for the next 10 years. These predictions consist of
detailed population estimations for each sub-county, and the total population of Bududa
district was estimated at 182 867 in 2012. The numbers for each sub-county were
consequently used and projected on a target sample of 300 respondents. The amount of
respondents interviewed in each sub-county is therefore proportional to the sub-county’s
population size. The final survey was carried out in two randomly selected villages of all the
15 sub-counties and 1 town council of Bududa district. This sampling method enables us to
use a random and proportional sample that takes into account the different landslide risk and
susceptibility zones of Bududa district.
In Table 2 we summarize the main sample characteristics. Descriptives are given for
the full sample as well as split over the self-reported steepness of the respondents’ own land.
Looking at the socio-demographic characteristics of our sample, two important inferences can
be made dependent on the steepness of the farmers’ land. First, people living on the gentle
and steeper slopes are significantly younger compared to people living in the flatter areas2.
Second, there is an inverse relationship between household income and the steepness of your
land. People who live in the flatter areas have a significantly higher income than people living
2
Differences in continuous variables where tested with two sample t-tests with unequal variances while Pearson
chi-square tests were run for categorical variables.
9
on steeper land. Besides, people living on the steepest slopes have a significant lower income
than people living on gentler slopes.
Other demographics show that household heads are mainly males; farming is the
major driver for income generation; there is a high average household size; almost all farmers
cultivate coffee; and 81% of respondents are living in huts or semi-permanent houses (no
bricks involved). Besides, respondents in our sample are proportionally divided over the
different landslide zones/areas of Bududa district as classified by Kitutu et al. (2009).
Landslides and steepness of the land are clearly positively correlated as people living on
steeper slopes report to have experienced significantly more landslides in their village as well
as in their surroundings. Concerning resettlement, as a first indication, 73% of respondents in
our full sample expressed a willingness to resettle defined as willing to leave current land and
get new safer land elsewhere. This number increases significantly to 90% for people living on
the steepest slopes. Finally, 84% of respondents in our sample were willing to give up part of
their land for landslide risk mitigation such as planting deep-rooted trees as Eucalyptus trees
or the Cordia Africana.
4.2 Social drivers to stay living on the risky slopes
Before studying resettlement as a key strategic element to mitigate landslide risk in
Bududa, it is important to know the underlying drivers why the Bududa people do not leave
the hazardous places even if they have been affected by catastrophic events in the past
(Lübken, 2012). Previous literature on environmental migration has shown that a bundle of
drivers affect migration decisions which can be identified as economic, political,
social/cultural, demographic and environmental drivers (Black et al., 2011). The cultural and
emotional attachment to place is often one of the most important forces for people to stay
although they are threatened by severe natural events (Lübken, 2012). Besides, affected
people often lack the personal and financial resources to resettle even if they face the risk of
repeated landslides (Hatton and Williamson, 2002). Finally, Lübken (2012) points out that
this family of drivers can shift in emphasis so that at some point cultural tradition may take a
privileged position in the affected people’s discourse and at another stage in the process it
may all be about resources they can’t leave behind such as fertile soils.
In what follows we summarize the main drivers the Bududa people put forward in both
the focus group discussions and the experimental survey that keep them to live on the slopes
10
of the extinct Mount Elgon volcano (Table 3). These drivers are consequently structured into
three categories: economic, anthropologic and perception drivers.
Looking at the economic drivers, soil fertility is the main factor to stay in the Bududa
district. The volcanic soils combined with the abundant rainfall create very fertile soils for
farming. These soils serve as protection against food insecurity and ensure that people are
able to grow cash crops such as coffee in order to generate an income. Besides, farmers
mentioned that although they knew they were facing landslide risks in the future, they could
not generate enough (financial) resources to migrate to lower risk areas.
The anthropologic and sociologic drivers can be classified into three closely related
categories. First, Bududa people have a strong cultural attachment to their place, as this place
is linked to their language, the cultural acceptance of polygamy and the cultural tradition of
male circumcision. These cultural habits are been frowned upon by other population groups in
Uganda and people therefore fear to lose their license to practice their cultural values once
they are resettled. Second, the familiarity with the climate and the resulting agricultural
practices that have been hand down by generations results in a strong land attachment.
Because of the tempered climate and the altitude, Bududa people are neither faced with
diseases such as malaria. Third, the largest driver people mentioned is linked to a strong
historical attachment to their ancestral land. These ancestral values pertain to former family
members being buried on their family land as well as ancestral sayings such as “If your
resettle, never go down, always go up.” Another example is after that a landslide displaced
some 3000 people in Bushiyi sub-county in August 2013, the Ugandan Daily Monitor wrote:
“Some hesitated, preferring they would rather die on their fertile ancestral land to which they
profess a sturdy bond”.
The last driver is linked to the current negative perception Bududa people have of
resettlements. When asked whether the risk of resettlement was higher or lower than the risk
of staying to live in a landslide area, 61% of respondents indicated that they perceived a
resettlement to be riskier than to stay living in the risky area. In several focus group
discussions it became clear that people rationally weighted the pros and cons of a resettlement
against these of a potential landslide risk. The percentage of 61% of respondents indicating
that a resettlement embodies a higher risk, originates from the perception Bududa people have
of the first landslide induced resettlement by the Ugandan government in 2010 after the
devastating landslide of Nametsi killing more than 350 people. More than 600 households,
11
4031 people, were resettled to Midwestern Uganda in Kiryandongo with the government
pledging to assist in housing, land and livelihood recreation. An analysis of this previous
resettlement by the authors in 2013 with the Impoverishment Risks and Reconstruction model
of Cernea (1999) showed that at least some essential resettlement risks where not adequately
tackled by the government and promises such as building houses for every household were
not kept (Internal report). Because of this, many internally displaced people returned to
Bududa and spread the negative experiences they encountered with the resettlement affecting
people’s perception throughout the Bududa area.
To conclude, voluntary migration is not happening because the poor have limited
access to capital, there is limited land available to settle elsewhere within the district, there is
a negative perception of previous resettlements, and people have strong cultural/historical
attachment to their place. However a resettlement serves as a key element to limit landslide
risks and thus loss of life and property holdings. Nevertheless, households link a resettlement
with other risks such as a decrease in soil fertility, an increase in morbidity due to changing
climate and impoverishment due to loss of livelihood creation. Therefore a compensation
incentive scheme, as constructed in our choice experiment, should be provided to induce a
more voluntary and community accepted resettlement scheme.
4.3 Choice experimental results
4.3.1 General random parameters model
The results of estimating the random parameter logit model for the full sample are
reported in Table 4. The alternative specific constant Staying (coded as 1 for staying to live in
their current place) implies there is a general willingness to resettle compared to staying to
live in the risky areas for the full sample model. People prefer to be compensated by a
monetary compensation, and more is better. There is a strong willingness to be resettled
within the Bugisu region, and especially within Bududa district compared to being resettled
outside the Bugisu area. This could indicate that people do not want to be resettled outside
their ancestral and cultural grounds due to the strong land attachment. For the housing
attribute we find that both barracks and a single house are less preferred than being resettled
into a multi-story building. This result might seem surprising at first. However a couple of
explanations could be hypothised. First, respondents might link the construction of flats to a
general development of the area (increase in services) since electricity, running water, etcetera
are needed to be constructed when people live in multi-story buildings. Second, due to the
12
population pressure in Bududa district, people possibly realize that there is no space to resettle
into a single house within the area. Last, from government consultations we inferred that
government plans are being developed to create semi-urban centers in flat areas to absorb the
population that lives on the steep slopes. It could well be that these government plans already
circulated among the affected population. Both health center and schooling attributes are
insignificant for the respondents. It seems that when asking people to make a resettlement
choice and trade-off different attributes against each other, health and schooling services
become of secondary importance compared to primary needs such as location and
compensation. Finally, regarding the compensation of land, they prefer to be compensated
with the same amount of land as they were holding before the resettlement. The significant
standard deviations of the parameter distributions show that preference heterogeneity is
present for all normally distributed random parameters. Therefore, to exploit this
heterogeneity, we run two split sample models based on two key variables linked to landslide
risk.
4.3.2 Split sample results
Although the full model can give us general insights, it is important to understand
whether the willingness to resettle is dependent on the probability of a landslide occurring and
whether the compensation asked by affected households differs under these different landslide
risk situations. Therefore, we run split sample random parameter logit models on two key
variables related to landslide risk. The first variable is the steepness of their agricultural land
as this is a key precondition factor for a landslide to happen3. Within one of the three areas
classified by Kitutu et al. (2009), there are still large variations in risk and an area variable
would therefore not capture sufficient variation. The second variable is whether a landslide
already happened in their village. This variable measures the factual (more chance another
one happens) and perceived possibility of a landslide happening in their surroundings. These
two variables constitute an ideal interplay between objective facts and own perception which
both have been shown to influence choice behavior in other fields.
Results of both split sample models (Table 4) show that the choice behavior of the
affected people is dependent on the probability of a landslide occurring. This means that
landslide risk significantly affects how people perceive a resettlement and consequentially
3
In case participants had multiple scattered plots, we asked them to record the steepness of their agricultural land
where their house was located. So steepness records both the steepness of the location of their house and their
surrounding plots.
13
their willingness to resettle. On top of that, the compensation asked by affected people differs
significantly depending on the landslide risk. Significantly less compensation is asked by
people living on steeper slopes or by people that have experienced a landslide in their village
in the past. Looking at each model of steepness separately4, people living in flat areas have no
willingness to resettle (insignificant ASC) unless they are extra compensated by both money
and land. For people living on gentle slopes, there is a limited willingness to resettle, which
increases with monetary and land compensation. Finally, people living on the steepest slopes
(the riskiest areas) have a strong willingness to resettle, even without any extra compensation
of land and money (both insignificant). For the people having experienced a landslide in the
past, again the general willingness to resettle and importance of monetary and land
compensation differs between the two subsamples.
5. DISCUSSION AND POLICY RECOMMENDATIONS
Implementing a resettlement choice experiment and asking people to reconsider a
completely new livelihood compared to their current situation simply by choosing 12 choice
cards may be considered overly simplified and unrealistic. Besides several authors have
shown that people find it hard to think about uncertainty, probability and risk, that humans
have limited cognitive capacity to process low probabilities and that people construe future
events differently from events in the present (Kahneman and Tversky, 1984; Chanteau, 1992;
Trope and Liberman, 2003). Therefore we do not claim that our choice experiment is an exact
representation of the reality taking into account the complex links between the different
economic, anthropologic, and environmental drivers. The authors acknowledge that it is
impossible to recreate livelihoods via the representation of 6 attributes and that it may be
difficult for participants to deliberate these questions. However, what we do claim is the
practicality and robustness of using a choice experiment vehicle to understand household
preferences for key resettlement attributes to inform and influence government policies which
have not been implemented yet.
4
Although you cannot compare parameters directly between logit models due to issues with the scale parameter,
we will use likelihood ratio tests from Swait and Louviere (1993) that allow to compare these coefficients. For
now, simple LR tests not accounting for the scale parameter show that these three parameters (staying and
compensation) do differ between the subgroups.
14
One of the key elements for a future resettlement policy to take into consideration will
be the choice of resettlement location. Although the willingness to resettle in our general
model is significantly positive, it becomes clear that the affected people prefer to be resettled
within the Bududa district or at least within the wider Bugisu region. In every model, the
location attribute was the most significant attribute and stayed significant throughout these
different models. It seems that ancestor land, culture, and land attachment will play a key role
in a future resettlement of the affected population. However resettling people within the
region would be a daunting task as the whole region is characterized by intense population
pressure and chronic land shortage (Claessen et al., 2007; Knapen et al. 2006). Therefore
government should pay particular emphasize to the risk of social disarticulation and the loss
of cultural determinants (Cernea, 2000).
The willingness to resettle increases with monetary and land compensation. Besides,
landslide risk has a significant influence on the willingness to resettle. Significantly less
compensation is asked by people living on the steeper slopes or by people that have
experienced a landslide in the past. It will be difficult to convince households from flat/gentle
areas and households where a landslide never happened to resettle without compensating
them considerably. We calculated the implied compensation5 that is needed for people to be
resettled around 400.000 UGX per month.
If resettlement is judged to be a feasible coping strategy, government should create
targeted resettlement campaigns since a lot of heterogeneity exists. When we control for
income, age and type of dwelling it becomes clear that people want to keep their current level
of livelihood. Older people, people with a higher income or living in a permanent house are
significantly less willing to resettle and ask a significantly higher compensation. Herein lays a
key opportunity from the policy maker’s point of view. The low-income and younger
households are faced with the highest landslide risk as they are being pushed upwards onto
steeper, unstable slopes. However these people ask less compensation to be resettled. For
policy makers, this entails an interesting opportunity since households that risk most to be
affected by a landslide can be resettled to the least cost (which does not mean they shouldn’t
be compensated at all). Besides, resettling younger families could also imply a more longterm and structural solution to curb future population pressure on the unstable slopes.
5
This implied compensation was estimated through the calculation of the implicit income that made people
switch from wanting to resettle to not wanting to resettle.
15
6. CONCLUSION
In this study, we use a discrete choice experiment to examine whether a resettlement is
a feasible coping strategy to mitigate landslide risks on Mount Elgon, and if so under which
conditions and compensatory schemes. Besides we investigate whether the willingness to
resettle depends on the probability of a landslide occurring. We model heterogeneity in these
preferences using a random parameters logit model and a split sample approach. Our study
provides the first analysis of resettlement related preferences of people that are affected by
environmental change namely recurrent landslides on the slopes of Mount Elgon.
Resettlement can be an effective landslide risk management strategy for reducing landslide
risk for large-sized recurrent landslides (Dai et al., 2002). However, we stress that the primary
goal of any resettlement process should be to prevent impoverishment and to improve the
livelihood of resettlers (Cernea, 1999).
Our results enable us to give valuable policy advice for different resettlement policies
which are not yet implemented in a very-cost effective manner. We identify a key opportunity
for policy makers if resettlement is opted for as a coping strategy. The low-income and
younger households are faced with the highest landslide risk as they are being pushed
upwards onto steeper, unstable slopes. However these people ask less compensation to be
resettled. This entails an interesting opportunity since households that risk most to be affected
by a landslide can be resettled to the least cost. Besides, resettling younger families could also
imply a more long-term and structural solution to curb future population pressure on the
unstable slopes. Our findings suggest that the choice of resettlement location will be
important for the resettlement to be a success. Although the willingness to resettle in our
general model is significantly positive, it becomes clear that the affected people prefer to be
resettled within the Bududa district or at least within the wider Bugisu region. However
resettling people within the region would be a daunting task as the whole region is
characterized by intense population pressure and chronic land shortage (Claessen et al., 2007;
Knapen et al. 2006).
In the future, it will be important to create susceptibility maps of the area and link
these with the local reality of the affected households. That way, future management
strategies could be targeted more directly. Since it is expected that climate change will result
in increased environmental degradation, studies suggest this will lead to increased permanent
migration (Black et al., 2011). Therefore, policy makers will need to act pro-actively in order
16
to secure the livelihoods and lives of their population in light of increased environmental
change. Besides, policy makers will need to be informed by knowledge of how migration and
displacement affect vulnerability as physical displacement, relocation and resettlement are
widely acknowledged as posing enormous social risk (Owen and Kemp, 2014; Stal and
Warner, 2009). Therefore government should not only look to how a resettlement may reduce
the landslide risk but should pay equal attention to eliminating any social vulnerability in
destination areas such that the solving of one problem does not create a bigger one (Stal and
Warner, 2009).
Given that the process of creating an optimal resettlement scheme as a mitigating
strategy is slow and complicated, it is all the more important for the relevant research to
develop methods to pre-test the behavioural impacts of hypothetical resettlement scenarios
through systematic experimentation. Our paper made such an endeavour and our finding that
people who are living in the riskiest areas are most willing to be resettled can serve as a piece
of scientific evidence for public authorities to further explore and implement a sustainable
community-based resettlement scheme.
17
REFERENCES
BDPU, 2012. Bududa district stastical abstract. Bududa.
Cernea, M. M. (1999). Why economic analysis is essential to resettlement: a
sociologist's view. Economic and Political Weekly, 2149-2158.
Cernea, M. M. (2000). Risks, safeguards and reconstruction: a model for population
displacement and resettlement. Economic and Political Weekly, 3659-3678.
Claessens, L., Knapen, A., Kitutu, M. G., Poesen, J., & Deckers, J. A. (2007).
Modelling landslide hazard, soil redistribution and sediment yield of landslides on the
Ugandan footslopes of Mount Elgon. Geomorphology, 90(1), 23-35.
Crozier, M. J., & Glade, T. (2006). Landslide hazard and risk: issues, concepts and
approach. Landslide hazard and risk. Wiley, West Sussex, 1-40.
Dai, F. C., Lee, C. F., & Ngai, Y. Y. (2002). Landslide risk assessment and
management: an overview. Engineering geology, 64(1), 65-87.
Deckers, S.A., Nachtergaele, F.O., Spaargaren, O.C. 1998. World Reference Base for
Soil Resources.
Gorokhovich Y., Doocy S:, Walyawula F., Muwanga A, Nardi F., (2013), Landslides
in Bududa, Eastern Uganda: Preliminary Assessment and Proposed Solutions, In: Landslide
Science and Practice, edited by Springer Berlin Heidelberg, pp 145-149, DOI 10.1007/978-3642-31337-0_19, ISBN 978-3-642-31336-3.
Jenkins, D. H., Harris, A., Abu Tair, A., Thomas, H., Okotel, R., Kinuthia, J., ... &
Quince, M. (2013). Community-based resilience building: normative meets narrative in
Mbale, 2010/2011. Environmental Hazards, 12(1), 47-59.
Kahneman, D. and A. Tversky. 1984, “Choices, values, and frames,” American
Psychologist, 39, 341-350.
Kitutu, M. G., Muwanga, A., Poesen, J., & Deckers, J. A. (2009). Influence of soil
properties on landslide occurrences in Bududa district, Eastern Uganda. African journal of
agricultural research, 4(7), 611-620.
Kitutu, M. G., Muwanga, A., Poesen, J., & Deckers, J. A. (2011). Farmer’s perception
on landslide occurrences in Bududa District, Eastern Uganda. African Journal of Agricultural
Research, 6(1), 7-18.
Knapen, A., Kitutu, M. G., Poesen, J., Breugelmans, W., Deckers, J., & Muwanga, A.
(2006). Landslides in a densely populated county at the footslopes of Mount Elgon (Uganda):
characteristics and causal factors. Geomorphology, 73(1), 149-165.
Owen, J. R., & Kemp, D. (2014). Mining-Induced Displacement and Resettlement: A
Critical Appraisal. Journal of Cleaner Production.
18
Oxfam. 2008. Turning up the Heat: Climate change and poverty in Uganda. Research
report prepared by Magrath John. ISBN: 9781848140394
Rukundo, P. M., Iversen, P. O., Oshaug, A., Omuajuanfo, L. R., Rukooko, B.,
Kikafunda, J., & Andreassen, B. A. (2014). Food as a human right during disasters in Uganda.
Food Policy, 49, 312-322.
Shanteau, J. 1992. Decision making under risk: Applications to insurance purchasing.
In J. F. Sherry & B. Sternthal (Eds.), Advances in consumer research. Chicago: Association
for Consumer Research.
Stal., M., Warner., K. 2009. Research Brief based on the Outcomes of the 2nd Expert
Workshop on Climate Change, Environment, and Migration 23 - 24 July 2009, Munich,
Germany
Trope, Y., & Liberman, N. (2003). Temporal construal. Psychological review, 110(3),
403.
Note: Incomplete due to time restriction.
19
Tables and Figures
Table 1. Attributes and their corresponding attribute levels
ATTRIBUTES
Location
LEVELS
Within subcounty
Within Bududa district
Within Bugisu region
Outside Bugisu
Housing compensation Single house
Baracks
Multi-story
Health services
Health Center 4
Health Center 3
Health Center 2
Education services
Primary + Secundary+ Vocational
Primary + Secundary
Primary
Monetary compensation 250.000UGX / 12months
250.000UGX / 6months
250.000UGX / 0months
Land compensation
Same amount of land
Half the amount of land
No amount of land
Table 2. Summary statistics for full sample and subsamples
Steepness of own land
Characteristics
Full Sample
Flat
Gentle
Steep
Socio demographics
N° of Respondents
307
97
94
111
% of male household heads
93%
94%
94%
93%
Age (years)
42
44
42***
39***
Average HH size
7.9
7.6
8.2
7.9
Literacy level
54%
54%
63%
47%
Main occupation: farminga
83%
85%
78%
86%
% of farmers growing coffee
94%
93%
97%
94%
Monthly income (UGX)
387 452
448 104 371 196*** 352 545***
Land cultivated (acres)
2.39
2.29
2.32
2.48
Avg. livestock holdings (LU)
2.05
2.31
2.12
1.78*
Semi-permanent dwelling (hut)a
81%
77%
78%
86%
Geography / Landslides
Zone/area
Central
19%
25%
16%
15%
East
40%
35%
40%
45%
West
41%
40%
44%
40%
Landslide in villagea
57%
35%
60%
75%***
Landslide in surroundingsa
74%
65%
72%
83%***
Resettlement
Willingness to resettlea,b
73%
61%
67%
89%***
Give up land for risk mitigationa,b
84%
83%
78%
88%
a
Percentage of sample possessing the specific characteristic, b Yes/No survey question
20
Table 3. Drivers inhibiting current voluntary resettlement
Drivers
Importance
Ancestor land
34%
Climate
24%
Soil fertility
23%
Income generation
21%
Less malaria
8%
Political pressure
1%
21
Table 4. Random paramet logit model results for both the general model as the split sample results.
General model
Steepness of own land
Landslide happened in own village
Flat
Gentle
Steep
Yes
No
Variables
Coeff.
Se.
Coeff.
Se.
Coeff.
Se.
Coeff.
Se.
Coeff.
Se.
Coeff.
Se.
Staying
0.351 -0.974*** 0.301 -2.397*** 0.297 -1.896*** 0.220
0.255
-0.990*** 0.159
-0.252
-0.310
Monetary comp 9.88e-05*** 0.000 0.00016*** 0.000 0.00015*** 0.000 1.7e-05 0.000 6.0e-05** 0.000 0.00016*** 0.000
Bugisu region
1.476*** 0.141 1.734*** 0.403 1.376*** 0.265 1.114*** 0.157 1.193*** 0.152 1.483*** 0.226
Bududa district 1.873*** 0.160 3.251*** 0.435 2.054*** 0.320 1.949*** 0.202 1.980*** 0.198 2.665*** 0.290
Sub-county
1.732*** 0.161 2.404*** 0.337 1.151*** 0.270 1.613*** 0.198 1.580*** 0.192 1.733*** 0.271
Barracks
-0.226*** 0.085 -0.333* 0.185 -0.399** 0.160 0.00937 0.111 -0.168* 0.098 -0.378** 0.151
House
-0.186** 0.089 -0.669*** 0.232
-0.264
0.163 0.0891 0.117 -0.172* 0.097
-0.213
0.164
HealthCenter3
0.000978 0.089 -0.00739 0.185
0.278*
0.168 -0.101 0.117 0.0718 0.103 -0.0960 0.153
HealthCenter4
0.0307
0.083
-0.176
0.201 0.368** 0.163 -0.0993 0.115
0.103
0.095 -0.0936 0.150
Prim+Sec
-0.0936
0.082
0.156
0.167
-0.190
0.165 -0.0618 0.105 -0.0868 0.097 -0.0168 0.134
Prim+Sec+Voc
0.121
0.086
-0.145
0.205
0.0572
0.167 0.266** 0.117 0.174* 0.098
-0.115
0.157
Half land
0.217**
0.106
0.358
0.261 0.526*** 0.198 -0.123 0.128
0.164
0.112
0.302
0.193
Same land
0.132 0.378*** 0.122 0.895*** 0.196
0.516*** 0.109 0.676*** 0.234 1.031*** 0.229
0.150
SDbugisu
1.645*** 0.154 3.196*** 0.551 1.603*** 0.260 0.731*** 0.212 1.226*** 0.182 1.783*** 0.212
SDbududa
2.389*** 0.198 3.491*** 0.449 2.205*** 0.303 1.107*** 0.188 1.705*** 0.202 2.540*** 0.279
SDsubcounty
2.020*** 0.165 3.426*** 0.426 2.111*** 0.327 -1.423*** 0.224 1.837*** 0.214 2.516*** 0.319
SDbarracks
-0.594*** 0.127
0.229
0.528 0.654*** 0.198
0.222
0.197 0.367** 0.157 -0.718*** 0.222
SDhouse
0.666*** 0.117 1.425*** 0.252 0.768*** 0.231
0.292
0.213
0.289
0.185 1.100*** 0.188
SDHC3
0.625*** 0.131 0.661*** 0.250 -0.660*** 0.236 0.377** 0.153 0.541*** 0.142 0.643*** 0.191
SDHC4
0.394*** 0.142 1.161*** 0.236 0.637*** 0.231 0.0660 0.180 0.0107 0.159 0.817*** 0.222
SDPS
-0.527*** 0.121
-0.228
0.567 0.899*** 0.231 -0.0900 0.214 -0.491*** 0.151 -0.585*** 0.207
SDPST
0.398**
0.166 0.821*** 0.303 0.851*** 0.203 -0.272* 0.162 0.290** 0.140 0.823*** 0.214
SDhalfland
0.913*** 0.122 1.115*** 0.273 -0.694*** 0.260 -0.276 0.196
0.301
0.208 1.249*** 0.189
SDsameland
0.939*** 0.115 1.337*** 0.255 0.833*** 0.202 0.473*** 0.158 0.571*** 0.146 1.169*** 0.187
LL
-2649.9
-821.02
-821.22
-849.73
-1403.92
-1164.23
Observations
11,052
3,492
3,492
3,384
3,384
3,996
3,996
6,300
6,300
4,680
4,680
Note: *** significant at 1% level. SD: standard deviations of random parameters. Both ASC and price attribute are kept non-random.
Models have been estimated with 500 Halton draws. All variables were dummy coded.
22
Figure 1. Location of the study area, Bududa district, in Uganda
23
Figure 2. Example of a choice card with two resettlement scenarios and a status quo.
24