Forest Management Decentralization in Kenya: Effects on
Household Farm Forestry Decisions in Kakamega
Wilfred Nyangena
School of Economics,
University of Nairobi,
Box 30197-00100, Nairobi, Kenya.
Email: nyangena_wilfred@uonbi.ac.ke
Maurice Juma Ogada
Kenya Institute for Public Policy Research and Analysis,
Bishops Garden Towers, Bishops Road,
Box 56445-00200, Nairobi, Kenya.
Email: mogada@kippra.or.ke or ogadajuma@yahoo.co.uk
Geophrey Sikei
EfD-Kenya/Kenya Institute for Public Policy Research and Analysis,
Bishops Garden Towers, Bishops Road,
Box 56445-00200, Nairobi, Kenya.
Email: goosikei@yahoo.com
The New Forest Management Regime in Kenya: Effects on
Household Farm Forestry in Kakamega
Abstract
This study investigates the effects of forest decentralization through the Kenya Forest Act
of 2005, on farm forestry investment decisions. The study uses household-level data
collected from Kakamega forest communities in March, 2010, and controls for selection
bias arising from incidental truncation by means of Heckman two-step approach.
Our results reveal that participatory forest management, among other factors,
significantly reduces the level of farm forestry investment among the poor rural
households. This indicates that, although co-management is useful in protecting the
existing government forests, a cocktail of other measures must accompany it for
increased forest cover to be realized. These measures could include: increased farmer
education, introduction of high-value fast-maturing farm trees and farm forestry incentive
schemes.
Keywords: Participatory Forest Management, Selection Bias, Farm Forestry
Development, Kenya
JEL Classification: Q12, Q28, D52
Forest Management Decentralization in Kenya: Effects on
Household Farm Forestry in Kakamega
1. Introduction
Forest decentralization programs have rapidly spread in developing countries in the last
twenty years (Agrawal, Chhatre and Hardin 2008). Support for the principle is derived
from grounds of economic efficiency, public accountability, community and individual
empowerment; as well as allocative efficiency (World Bank 2009). These reforms are
designed recognize forest and land rights to a diversity of local people eking a living
from these forests. These reforms are expected to reconcile both conservation and
livelihood needs. In particular forest decentralization is aimed at enhancing peoples’
livelihoods, poverty alleviation and preservation of the forest condition.
Decentralization policies do not affect forest users’ behaviour directly- rather they change
local incentive structures by altering security, access and the power structure of local
governance which in turn lead to behavioural change. The expected outcomes, of regime
change are mediated by forestry regulations that impose conditions for use of forest
resources, and by the capacities of small holders and communities to adopt to those
regulations. For instance the communities are required to implement workable systems of
governance for their collective lands, exclude third parties and engage in competitive
conditions with the forest markets. Indirectly, the outcomes of the reform are also
influenced by access to financial and non-financial services. In the absence of these
conditions, forest tenure reforms are unlikely to achieve their livelihood and conservation
goals.
Thus decentralization policies may change patterns of land use and local governance that
produce a variety of outcomes, some desirable and some not. For example, many of the
Community Forest Associations (CFAs) formed in Kenya were driven by expectations
beyond what the legislation provided for (Ongugo, 2007; Ongugo et al., 2007). Indeed
some CFAs anticipated converting forests into farmlands for production of cash and food
crops (Ongugo et al., 2004). In some cases if property rights are devolved to diverse
forest groups, the new property rights holders may hesitate to invest in forest resources if
they fear that these changes are temporary.
These diverse outcomes may be explained by several reasons. First, for groups that lack
experience and tradition of forestry operations such institutional arrangements are
complex. Second, in some instances, these new organizations have had tensions with
existing social organizations. The state assumes that these organizations have the capacity
to govern on their own and provide the required support to local people. Lastly, with
regard to market engagement, while some communities are capable to participate
aggressively and adopt entrepreneurial activities, others do not have the capacity to do so
(Monterroso 2008). Thus, to clearly understand the effects of forest decentralization, one
should examine the plethora of relevant variables in different contexts that might be
altered.
The links between national policy changes and their effects on the ground in terms of
local level behavior are mediated by a host of complex processes that inhibit policy
implementation (Sabatier 1986). Forests are usually located in remote and sometimes
marginal areas with many poor people. It is no surprise that evaluations of forest
decentralization have posted disappointing results (Agrawal and Ribot 1999). A solid
understanding of how local users’ behaviour change in response to these policies is also
critical and required.
Several studies have been conducted on community participation in forest management,
effects of PFM on household poverty and opportunity cost of forest conservation (Borner
et al., 2009; Colfer, 2005; Emerton, 1999; Guthiga et al., 2008; Mbuvi et al., 2007;
Mogaka et al., 2001; and Ongugo, 2007).. Research on forest decentralization, like other
forms of decentralization is plagued by analytical problems. First, decentralization is a
general term that is applied to a diversity of policies that may include some combination
of (a) increasing the decision making discretion of local level bureaucrats; (b) increasing
the decision-making authority of local users; and (c) moving government officials from
central locations to sites closer to forests (Cohen and Peterson, 1996). Second,
decentralization policies interact with numerous context specific pressures and
interactions to change governance institutions, forest user behavior and resulting forest
conditions and livelihood outcomes (Andersson et al., 2008). Lastly, while there are
several theoretical arguments relating benefits and costs of forest decentralization, but
these fail to generate consistent predictions (Andersson, et al., 2008). These studies
ignore behavioural changes resulting from decentralization among forest users in their
empirical investigations
This study seeks to address this gap by analyzing household farm forestry investment
decisions within the larger context of national decentralization policies. We seek to
understand how decentralization policies filter down to local forest users. Economic
theory does not provide clear predictions about the effects of decentralization policies on
forest users’ behaviour. Instead we must derive from studies of how such policies interact
with existing biophysical, socio-demographic variable such as age, gender and
educational variables, wealth and other factors change incentives at the local level.
We test the effects of forest decentralization, arguing that the effects of decentralization
need to be understood according to specific contexts. We investigate the effects of
decentralization drawing on data collected from Kakamega forest in Western Kenya. In
particular, we test the effects of decentralization on the farmers’ tree investment decisions
on their own farms. Increased forestry cover is a key policy requirement in Kenya, where
forest cover is only 3% far way below the recommended rate of 10%.
The rest of the paper is organized as follows: In the next section we review the history of
decentralized forest reforms in Kenya. In section 3 we draw on existing literature to
derive factors that influence household farm forest investment decisions. Section 4
examines methodological issues while Section 4 provides summary statistics of the
variables used. In section 6, we report our empirical results and discuss these findings and
in section 7 we conclude and draw policy recommendations.
2. Forest Decentralization trends in Kenya
The colonial government of Kenya created a forest department in 1902, which alienate
most prior existing community-managed forests. The Forest Department managed and
controlled all forests in the country with policy focused on conservation. Following
independence in 1963, a series of donor funded forestry programs focused on
afforestation and reforestation on farms, with the goal of alleviating fuelwood shortages.
The Forest department managed the forests without consultation outside the relevant
government ministry. Conflicts increased in the late 1980s between communities who
needed fuelwood from neighbouring forests, and the forest department (Ongugo and
Njuguna 2004).
The New Forest Act of 2005 saw the formation of the Kenya Forest Service (KFS), a
semi autonomous government agency with representation from various government
ministries. Under the Act, the KFS is expected to devolve powers to the private sector
and to forest conservation committees and community forest associations (CFAs).
Community participation is achieved primarily through CFAs, and integrated
management of forests are central principles motivating the new policy (Ongugo, et al.,
2007).
3. A review of farm forestry decisions by rural households
A clear understanding of tree planting decisions by households is required to estimate the
likely impact of participatory forest management approaches on households’ decisions to
establish farm forestry. To this end, this section reviews the link between participation in
community forest management groups and households’ decisions to establish farm
forestry. It also explores other factors that may motivate households to undertake on-farm
tree growing.
The body of literature on farmers’ tree growing decisions highlights the complexity of
factors involved in the behavioural function. The complexity arises from the diversity of
circumstances under which smallholder farmers operate. It is generally recognized in the
literature that a number of factors explain the differences in farm tree growing decisions
by smallholder farmers. However, the specific socio-economic and institutional variables
affecting the decisions differ across countries, regions, villages, and farms. Moreover, the
direction and significance of influence of a given variable is not often consistent across
studies.
Participation in forest management groups has been shown to influence decisions to plant
more trees on-farm (Emtage and Suh, 2004). This, however, does not imply that
households not participating in these groups are unlikely to plant. Participation in forest
management groups greatly enhances people’s attached value to forest ecosystems and
the need to protect them; which in turn results in their desire to increase forest cover on
their farms. Participation in community-based conservation groups enhances farmers’
access to diversity, quality and quantity of tree species (Boffa et al, 2005). Groups
increase farmers’ awareness on importance of trees and market access to the tree and
other agricultural products. The argument for market access, however, only applies where
markets work fairly well. In such cases, farm households would plant high value timber
and fruit trees, not only to satisfy their monetary needs but also subsistence requirements
(Emtage and Suh, 2004; Shively, 1999). This implies that closeness to market could also
accelerate participation of households in establishing farm forests through two distinct
channels—providing market for products of farm forests and creating alternative off-farm
income generating avenues which compels households to set their own sources of fuel
wood because of lack of time for collection from other sources far from households
(Dewees, 1995; Dove, 1995; Gilmour, 1995).
Besides Participation in community forest management, households’ decisions to plant
trees may be directly influenced by household-specific, plot-specific and institutional
factors. For instance, farm forests have enormous environmental advantages beyond
direct benefits to the farm households. To comprehend these indirect benefits, the
decision-maker at household level requires some education, either formal or informal,
obtained through schooling or extension services. Thus, better educated household heads
or households with access to government or farmer-farmer extension services are better
adopters of farm forestry (Muneer, 2008), either because they view tree planting as a
means of improving the land (Dewees, 1995) or because they are able to appreciate other
non-quantifiable
benefits
as
ambiance,
micro-climate
modification
or
carbon
sequestration. This also explains why households with good social networks may have a
higher possibility of planting trees because they are able to get extension services through
such networks (Gebreegziabher et al., 2010; Muneer, 2008).
Institutional factors have also been shown to influence the decision by households to
plant trees. Secure land tenure arrangements, for example, have been found to influence
tree planting decisions among farmer groups. Trees take a longer gestation period and
only farmers who are confident of continued use of a given plot would be encouraged to
plant them (Bannister and Nair, 2003; Deininger and Feder, 2001; Gebreegziabher et al.,
2010; Warner, 1995). However, some studies do not agree with the idea that secure
tenure may encourage tree planting and cite cases where communal ownership of land
has been more conducive for development of farm forestry (German et al., 2009).
Perhaps tree planting in areas with ambiguous land tenure system is a means used by
households to place a claim of legitimacy of ownership and/or access.
4. Conceptual framework
Kakamega forest region on which this study is based lies in the rural parts of Western
Kenya. Like other parts of developing countries, agricultural households are expected to
contend with either missing or highly imperfect markets characterized by high transaction
costs and constraints on marketed quantities (Thorbecke, 1993; Ellis, 1993). Specifically,
only a few households buy/sell forest products. Equally, very few households hire labour
for farm activities or extraction of forest products from the neighbouring government
forest. As a result, this study is premised on the non-separability of household production
and consumption decisions first introduced by Singh et al (1986) and advanced by others
like de Janvry et al (1991), Sadoulet et al (1996) and Taylor and Adelman (2003).
This study adopts a theoretical framework of utility maximization by a household
envisaged by Singh et al (1986). The household derives utility from consuming market
goods, home-produced goods and household leisure time. Thus, the household’s
optimization problem can be expressed as:
Max U =U(Ch, Cm, Th- Sh; Ηh)
(1)
s.t
Q=f(K, L, A)
(2)
PmCm= Ph(f(K, L, A)-Ch)-wL+ wSh+I
(3)
Th= Lh+ Sh
(4)
L- Sh ≥0
(5)
Ch- Q ≥0
(6)
The household maximizes the utility (Eq.1) subject to the production function (Eq.2),
household’s income constraint (Eq.3), the household’s total time constraint (Eq.4) and
market environment constraints (Eq.5 & Eq.6). Where U(.) is a well-behaved utility
function, Ch is consumption of home-produced goods and Cm is consumption of market
goods. Both Ch and Cm include agricultural crops and forest products. Lh is household
leisure hours while Ηh is household characteristics. Q is the home output of both
agricultural crops and forest products with f(.) being assumed to be increasing and
concave in all its arguments, K is the capital input, L is labour demand for production
while A refers to other factors that affect production. Pm is price of market goods, Ph is
market price of home-produced goods, Th is total time available to the household, w is the
wage rate, Sh is time spent on household production and off-farm wage earning
(household labour supply) and I is the exogenous household income from non-wage and
non-farm sources.
For output of forest products, the household has two options—own farm forestry and
government forest. However, in the area of study, households that are members of
Community Forest association (CFA) presumably have more access to the government
forest. The fact that a household is able to access government forest may not necessarily
mean that it ceases to plant trees because the range of products from government forests
is restricted.
The Lagrangian equation for this optimization problem becomes:
Z= U(Ch, Cm, Th- Sh ; Ηh) + λ[ Ph{f(K, L,A)-Ch}-wL+wSh)+I-PmCm]+γ[Ch-Q]+θ[L-Sh](7)
From FONC;
dU
dC
h
dU
dS
p
h


(8a)
 w    [ w   /  ]
(8b)

p
h
  =λ[
df
 w  / 
h dL
p
h
 ]
(8c)
The FONC reveal that, so long as the market environment constraints are binding, market
prices (Ph and w) cannot guide household decision-making. Instead the household is
guided by virtual/shadow prices (shown in parentheses in Equations 8a and 8b). Equation
8c also shows that the value of the marginal product of labour is not equal to the market
wage rate. Because the virtual prices reflect the true opportunity cost and benefits,
households respond to them rather than market prices while making utility-maximizing
choices (de Janvry et al., 1991; Singh et al., 1986). It is the sign of γ/λ and θ/λ that
determine the size of virtual prices and the relevant wage which would vary by household
depending on whether a household is self-sufficient in, net seller or net buyer of a
produce or labour (Sadoulet et al., 1996). These variations in prices and wages are caused
by transaction costs in buying and selling, household preferences, production technology
and access to employment opportunities. Thus, it is important to include the causes of
these variations in estimation of the production function. This is the approach adopted in
this study despite the existence of ways of estimating virtual prices and wages (see
Jacoby, 1993; Skoufias, 1994; Cooke, 1998; and Mekonnen, 1999). We are motivated by
the fact that agricultural production is a process that takes longer gestation periods
characterized by shocks, sequential decisions, irreversible choices, and complementarities
and substitutability among inputs (de Janvry and Sadoulet, 2003). This leads to large
inaccuracies in estimation of marginal productivities.
4.1 Analytical Model
The area of study has embraced community forest management introduced by the Forest
Act (2005). Individuals are free to choose whether to join Community Forest
Associations or not. However, only members of Community Forest Associations have the
privilege to access a specified range of forest products. Members of Community Forest
Associations, because of their access to forest resources from various parts of the larger
Kakamega forest, may have no drive to establish own farm forests or they could simply
reduce the number of trees planted on own farms. Alternatively, these people could
expand forest cover on their own farms due to increased appreciation of forests and the
need to protect them. This self-selection choice made by households which want to
belong to Community Forest Associations (CFAs), if ignored in the analysis, would lead
to inefficient parameter estimates and biased intercept (see Gronau, 1974; Lewis, 1974;
Heckman, 1974). The argument, here, is that the unobservable household factors that
drive participation in CFAs could be related to those that drive household tree planting
behaviour. Thus, correcting for selection bias is a precursor to our analysis of level of
farm forestry establishment by households in the area of study.
The study hypothesizes that besides conventional productive factors, individual,
household and institutional factors contribute to farm forestry development at various
levels of resource endowments. These factors express themselves through correlation of
self-selection of households into CFAs or not and level of farm forestry development, and
through inclusion of observable factors in explaining inter-household variation in levels
of farm forestry development. The main econometric task is to determine how selfselection and other factors influence levels of farm forestry among farm households in
Kakamega.
Following Heckman (1979) and Flood and Grasjo (1998; 2001), we specify our model,
ordered into two regimes. The first regime is defined by whether a household is a
member of a Community Forest Association or not. This is described by the selector
equation;
d  Z 
'
*
i
i
Where
d
i
i
 1 if
, μ~N (0, 1)
d
*
i
(9a)
 0, indicating that a household is a member of CFA, and zero
otherwise;
(9b)
Prob( d i  1)  ( 
where
Z
i
'
Z ) , Prob( d
i
i
 0)  1   ( 
'
Z ),
i
is a vector of explanatory variables, β is a vector of parameter estimates, μ i is
the white noise, Prob(.) is the probability function and Ф(.) is the cumulative distribution
function (CDF).
The level of farm forestry, measured by the number of trees planted by a household, is
determined in the second step, defined by the equation;
y  X  
'
i
i
(10),
i
applicable only if d i  1 .
Where
y
i
number of trees planted, α is a vector of parameter estimates, Xi is a vector of
exogenous variables and

i
is the error term.
The two error terms are assumed follow a bivariate normal distribution. That is,
(μi ,

i
 )~bivariate normal{0, 0, 1,
i

, ρ}; where ρ is the correlation coefficient. If μi and
are correlated (ρ≠0), there is enough justification to use Heckman Selection Model.
Model 9a is estimated by probit while model 10 is estimated by the standard OLS.
Transforming equation 10 into,
E[ y /Xi, di=1]=
'
i
X
i
+ E[ i / d i  1]  
'
X
i
   M i
(11),
where M is the inverse mills ratio. Failure to incorporate M implies under fitting the
model, leading to omitted variable bias in the estimation of α. Equations (9a) and (10) are
simultaneously estimated by two stage approach for consistent and efficient estimates.
The Log-Likelihood function for our specification, following Aristei et al (2007), can be
expressed as,
L= 0

'
 '

  Z i  ( yi   X i  1 
'

 
ln[(1-Φ(  Z i )]   ln [ 
2

  
1 





y  X
'
i

i


]

(12),
which can be simplified as,
L= 0 ln[(1-Φ( 
'

 '
1 
)]

ln
[


Z i    Z i   

y  X
'
i

i


]

(13),
if the error terms are not correlated (ρ=0)
The coefficients of the Heckman Model cannot be interpreted directly. Thus, we compute
the marginal effects as;
dE ( y / d  0, x)
*
dx
i
  i       (  Z )
i
(14),
where

i
is the parameter estimate from the outcome equation,

i
is the parameter
estimate from the Selector equation, ρ is the correlation coefficient of the Selector and the
outcome equations,


is the standard error of the residual in the outcome equation and
 (   Z ) is the inverse Mills ratio. This applies for the variables that appear in both
Selector and outcome equations (Sigelman and Zeng, 1999). For variables that appear
exclusively in the Selector equation, the marginal effects, according to Maddala (1983)
are computed as;
( 
'
Z
Z )  (
i
ik
 'Z )Z k
(15).
One main limitation in an estimation of this nature is that if there is a strong
multicollinearity between the independent variables and the inverse Mills ratio, the
parameter estimates cease to be efficient. We overcome this in our analysis by
introducing distance to the government forest and distance to the market for purchase of
firewood which have the potential of influencing participation of a household in CFA but
not the number of trees that the household may plant (see Sartori, 2003). Government
forest is important as an alternative source of wood fuel and other forest products but
only CFA members can access it. Also, those closer to markets for forest products may
not have the incentive to join CFAs. Thus, the two variables could easily determine
household participation in CFA but without directly influencing the number of trees
planted by a household.
5. The Study Area and Data
The study site for this survey was around Kakamega Forest, situated in Kakamega
District in Western Province of Kenya (Figure 1). It lies north-east of the Lake Victoria
between latitudes of 00°10’N and 00°21’N and longitudes of 34°47’E and 34°58’E at
about 1600 m above sea level. The forest area is drained by two main river systems, the
Isiukhu River to the north and the Yala River to the south.
The forest is the only
remaining rain forest in Kenya and is the furthest east remnant of the Guinea-Congolean
rain forest. According to the 1994 welfare monitoring survey, 52% of the population in
the district lives below the poverty line, meaning that they can hardly afford basic
necessities like food, shelter, clothing, and education. As such there is a heavy reliance on
the forest to supplement their daily necessities. This region has also been considered by
the Kenya Woodfuel and Agro-forestry Programme (KWAP) as one of the areas that
could benefit most from policies that target improvement of forestry projects due to its
high population and high agricultural potential.
Source: Biota Sub-project E13 data bank, 2006
Figure 1: Kakamega forest and its environs
The data for this study was collected from communities around Kakamega forest. A
random sample of 318 households was interviewed using a detailed semi-structured
questionnaire. The sampled households were randomly interspersed in the study area and
across the three management regimes. Table 5.1 shows a summary statistics for the
variables used in the paper.
Table 5.1: Basic descriptive statistics
Variable
Individual attributes
Age of head
Education of head (years of schooling)
Household size
Male headed households
Farm characteristics
Farm size in acres
Number of trees on farm
Households planting trees on farm
Value of total assets
Mean
Std. dev.
48
9
5
0.73
14
5.8
1.9
.44
2.2
41
0.76
22026
2.2
77
0.43
61038
Distance to nearest forest edge (km)
Market distance for selling forest products (km)
Access to extension services (1 if yes, 0 otherwise)
Access to credit facilities (1 if yes, 0 otherwise)
Institutional attributes
Membership in social groups
Membership in CFA (1 if yes, 0 otherwise)
Awareness on new forest laws (1 if yes, 0 otherwise
Tenure status
Own land with title
Own land no title
1.5
3.6
0.23
0.17
2.2
3.8
0.42
0.37
0.73
0.52
0.60
.44
0.50
0.50
38%
59%
The household characteristics show that the average age of the household heads is 48
years, with the average number of schooling years being 9. The low education level of
the household head is a contributor to their inability to secure more remunerative
employment opportunities elsewhere, thereby resorting to farming activities. Education
increases household’s off-farm employment opportunities. Furthermore, highly educated
members of the household tend to look for greener pastures in off-farm activities. This is
because of the traditional nature of farming activities within the region which many
people view as not competitively rewarding compared to non-farming activities. The
average household size measured in this survey is 5 members. Of the households
interviewed 73 percent were male headed.
On the farm characteristics, the average farm size in Kakamega is 2.2 acres; 76 percent of
households practice farm forestry while the mean value of household assets is KES.
22026. The average distance to the nearest forest edge is 1.5 km. On average, 23 percent
and 17 percent of the households had access to extension services and credit facilities,
respectively.
With regards to institutional attributes, 73 percent of households participate in social
groups while 52 percent participate in community forest management through CFAs. In
terms of tenure security, 38 percent of households owned land with title deeds while 59
percent owned non-titled land.
6. Results and Discussion
Table 5.1 presents a catalogue of results from the selector and outcome equations of the
Heckman regression of household farm forestry development.
Table 5.1 Heckman selection Results for Household Tree Planting
Variable
Coefficient
p-value
Outcome equation
Individual attributes
Education of household head
-0.341 (0.021)
0.109
Household Size
0.097 (0.041)
0.017**
Male-headed household
0.419 (0.167)
0.013**
Farm characteristics
Farm size in acres
0.092 (0.034)
0.008***
Log of value of total assets
-0.010(0.052)
0.841
Distance to Market for selling forest
0.073 (0.042)
0.083
products (km)
Market distance for selling forest
-0.001 (0.001)
0.251
products (km) squared
Institutional factors
No. social groups
0.185 (0.070)
0.008***
Access to extension services (1 if yes, -0.199 (0.156)
0.204
0 otherwise)
Access to credit facilities (1 if yes, 0
-0.027 (0.191)
0.889
otherwise)
Secure land tenure (own titled)
-0.504 (0.177)
0.004***
Constant
3.009 (0.532)
0.000***
Selection equation
Individual attributes
Education of household head
0.084 (0.034)
0.014**
Household Size
-0.051 (0.060)
0.392
Male-headed household
-0.453 (0.266)
0.089
Age of head
0.004 (0.01)
0.699
Farm characteristics
Farm size in acres
0.376 (0.116)
0.001***
Distance to nearest forest edge (km)
-0.331 (0.168)
0.049**
Distance to nearest forest edge (km)
0.033 (0.022)
0.129
squared
Distance to buying market for forest
-0.01 (0.069)
0.147
products
Distance to selling Market (km)
-0.240 (0.101)
0.018**
Distance to selling Market (km)
0.005 (0.002)
0.034**
squared
Institutional factors
Awareness on new forest laws (1 if
0.608 (0.239)
0.011**
yes, 0 otherwise)
Access to credit facilities (1 if yes, 0
-0.408 (0.284)
otherwise)
Secure land tenure (own titled)
0.686 (0.265)
Constant
-0.004 (0.722)
Mills Ratio
Lambda
-0.936 (0.722)
Rho
-0.980
Sigma
0.955
Wald test of independent equations
Chi 2= 69.25
** Significant at 5%, *** Significant at 1%
Standard errors in parentheses.
0.152
0.010***
0.996
0.004***
Prob>Chi2= 0.000
The results indicate that the error terms of the selection and outcome equations are
correlated as shown by the highly significant chi-square p-value of 0.000 and the
significance of the inverse Mills ratio. This justifies the use of Heckman procedure. The
fact that lambda is significant and negative shows that a household’s participation in
forest management through CFA reduces its level of investment in farm forestry. This
means there is correlation between households which self-select themselves into CFAs
and reduced investment in farm forestry due to unobservable effects. The most apparent
explanation to this is that households that participate in CFA have access to forest
products such as fuel wood, medicinal plant, wild fruits and vegetables, and certain kinds
of fodder. For such households, land would rather be put to uses other than farm forestry.
The rest of the results are discussed under respective equations:
6.1 The Selection Equation
In the selection equation, variables that significantly and positively influence selection
include the household’s land size, the square of the selling market distance, education of
the household head, household’s awareness of the new Forest Act and security of land
tenure. Those that have negative influence include distance to the government forest,
distance to the selling market, and male-headed households.
Land size, being a proxy of household’s wealth, could positively influence the choice to
join CFAs because such households are able to pay the requisite fees for participation.
Large landholders are also associated with large herds of cattle and this may increase
demand for grazing fields, attainable from forests through membership to CFAs. The
convex nature of the effect of distance to selling market on the selection could be
embedded in the fact that such a distance influences access to information on formation
and operation of CFAs, and for those who extract forest resources with the intention of
selling, longer distance to the market makes joining CFAs less attractive. Reversal of the
effect of market distance on selection after some limit could be explained by the fact that
markets tend to be near urban centres. Thus, very far away from these markets are most
likely typical rural settings with greater needs for primary livelihood products attainable
from forests. This creates the impetus to join CFAs.
Better educated household heads are likely to be aware of formation of CFAs and the
potential benefits. Consequently, where the head is better educated, the household has a
higher chance of participating in CFA. It is also evident that households that enjoy secure
land tenure are more likely to join CFAs. This is perhaps due to the fact that such
households are made of native residents, who find it easier to work together with the
other residents they have known for a longer time, some even being close relatives. Such
natives may also have special attachments to the government forest, either for religious or
cultural reasons. This creates the urge to join CFAs so as participate in protection of the
forest and gain access to it as and when need arises.
Households that are aware of the new Forest Act, which actually introduces community
participation in management of forests, are more likely to join CFAs. Such people
definitely understand that it is only through participation in CFAs that they stand to gain
from the government forest.
Distance to the government forest influences selection negatively because one of the
main reasons for participating in CFAs is to gain access to the forest for extraction of
specific forest products. As a result, those who live far from the forest have no motivation
to join CFAs. Similar results are associated with male-headed households most probably
because such households are able to raise own farm forests from which to obtain the
basic products which would have otherwise made it necessary to access government
forest through CFAs.
6.2 The Outcome Equation
In the outcome equation, distance to the selling market appears to have a concave
relationship although only the coefficient of distance to the market is significant at 10
percent level. The reason for this could be that those living far from the market have to
rely more on subsistence production of most of their domestic requirements. If they need
forest products, they may be compelled to plant their own trees. The geography of
Kakamega is such that the forest is closer to the town. The implication of this is that, if
town is the main market, those living far from the market are also far from the
government forest, the alternative source of forest products.
Of special interest in the outcome model are the coefficients of household heads,
household size, social capital and farm size. All are significant and positive; indicating
that, holding other factors constant, male-headed households, larger households, larger
land size and more participation in social organizations are associated with planting of
more trees. On the contrary, secure land tenure is associated with planting of fewer trees
or under-development of farm forestry.
The fact that male household heads are associated with planting of more own farm trees
is more cultural than economics. Planting of trees is viewed more as a man’s activity than
a woman’s. Where labour markets do not function properly or where the household is
less endowed with resources to hire labour, men could be better placed to plant trees
because this activity is more strenuous. Other activities like fencing and construction of
household’s dwelling units are also viewed as a man’s responsibility and these could
compel a man into planting more trees in anticipation of future needs.
Larger households tend to plant more own farm trees. There are two main angles to
this—larger households have larger requirements for forest products such as fruits, fuel
wood and medicinal plants. This could make it more prudent and economical for such
households to establish own farm forests. The second angle is that tree planting is labourintensive and larger household are capable of using own labour to accomplish the tasks
involved. Diversification of livelihood sources to meet the pressure of feeding a large
family may provide an alternative explanation.
Social capital is critical for farmer-farmer extension and also for raising resources,
physical, human and financial. Thus, households with wider interaction networks through
participation in social organizations plant more trees, either because they have drawn
lessons from others in their network loops or because they have benefited from rotational
labour common among rural households. Through the social networks, relatively poor
households can borrow the requisite farm tools or even obtain free tree seedlings from
other group members.
Households with large pieces of land, all else equal, plant more trees. This is because
trees compete with other crops for land. Those with smaller pieces of land may, therefore,
devote the entire parcel for food crop production as opposed to large land holders who
are able to either inter-crop food crops with trees or even set aside portions of land
exclusively for trees. For households that use trees to mark plot boundaries, it follows
that that the larger the piece of land, the more the trees planted.
The coefficient of security of land tenure looks perverse at first sight. One would expect
households with secure land tenure to be more motivated to plant trees because trees take
a longer time to manure. However, on second thought and taking cognizance of the
Kenyan situation, trees are a means of entrenching land ownership, either on plots that
are not regularly cultivated or on untitled land. Consequently, households with insecure
land tenure are predisposed to plant more trees, not as an economic investment but as
evidence of ownership should disputes arise.
6.3 Marginal Effects
Marginal effects computation combines effects of variables included in both the outcome
and the selection equations, conditional on the dependent variable being observed or
selected. Table 6.2 reports these results.
Table 6.2: Marginal effects of the Model
Variable
Individual attributes
Education of household head
Household Size
Male-headed household
Age of head
Farm characteristics
Farm size in acres
Distance to selling market (km)
Distance to selling market (km)
squared
Distance to nearest forest edge (km)
Distance to nearest forest edge (km)
squared
Distance to buying market for forest
products
dy/dx
p-value
0.019 (0.008)
-0.011 (0.013)
-0.090 (0.047)
0.001 (0.002)
0.016**
0.397
0.053
0.700
0.084 (0.020)
-0.053 (0.022)
0.001 (0.0005)
0.000***
0.014**
0.030**
-0.073 (0.038)
0.007 (0.005)
0.056
0.134
0.022 (0.015)
0.143
Institutional factors
Awareness on new forest laws (1 if
0.149 (0.063)
yes, 0 otherwise
Access to credit facilities (1 if yes, 0
-0.102 (0.078)
otherwise)
Secure land tenure (own titled)
0.141 (0.052)
** Significant at 5%, *** Significant at 1%
Standard errors in parentheses.
0.018**
0.188
0.007***
After accounting for selection, coefficient of distance to the selling market is negative
while the coefficient of the square of the distance to the market is positive, showing a
convex relationship. This implies that the number of trees planted by households falls as
distance to market increases up to an inflection point and then starts to rise for
households that lie very far from markets. Possible explanation to this is that households
plant trees not only to meet their domestic needs but also to generate income. As such,
the households near markets plant more trees because they are motivated by the possible
large profits. But very far from the market could be a reflection of typical village life
where demand for fuel wood, medicinal plants, fruits and fodder is very high yet each
household has to rely on own production. What was noticed during the survey is that
jaggery-making is a major economic activity in the remote villages of Kakamega—this
could be fanning demand for fuel wood which translates into increased planting of trees.
Secure land tenure is found to positively influence the number of trees that a household
plants. The underlying explanation to this is that trees take a longer time to mature and
one would invest in them only if he/she is certain of using the land for a longer time.
Similarly, size of landholding by the household turns out highly significant, increasing
the number of trees planted. A household with a small piece of land may devote the land
for production of subsistence crops and may use the land too intensively to allow planting
of trees. Conversely, households with large pieces of land are capable of either
undertaking agro-forestry or setting aside substantial portions of their land for growing
trees. Even if trees are confined to farm boundaries, a larger plot would still
accommodate a fairly large number of trees.
Other factors that increase the chance of planting more trees are the education level of the
household head and household’s awareness of the new Forest Act. Better educated
household heads are able to appreciate the role of forests—in soil and water conservation
at the farm level and carbon sequestration at the global level. Such household heads may
also be more inclined towards ensuring that the government forest remains intact, making
them develop their own sources of forest products. The Forest Act (2005) which
introduces community participation in forest management gives communities increased
access to the forest but only for a restricted, well-defined range of products. It is also
more punitive on the violators and the fact that it involves communities increases the
chance of the violators being caught. Thus, for those who understand this Act, it would be
more prudent to develop own forest if the interest goes beyond products allowable from
government forests.
For variables that appear only in the outcome equation, the parameter estimate is indeed
interpreted as the marginal effect. Therefore, social capital can be said to be positively
related with the number of trees planted by a household, other factors remaining
unchanged. This is attributable to flow of information among members of social groups
on the positives of planting trees. Other contributions of social capital could include interhousehold rotational use of labour and farm implements which is well developed among
rural households. Farm forestry is labour-intensive may require households to pool their
labour resources especially in instances where households are not well endowed with
financial resources to hire labour or where labour markets are not well developed.
7. Conclusion and Policy Recommendations
This study focuses on how participation in community forest management, introduced in
Kenya by the Kenya Forest Act (2005), and other factors impacts on development of
farm forestry which has been perceived as a vehicle through which to increase the
country’s forest cover. Using household data collected from Kakamega forest
communities and controlling for selection bias, the study find makes key findings with
various policy implications.
First, households that are involved in community forest management through Community
Forest Associations (CFAs) are associated with under-developed farm forestry possibly
because they are able to meet their demands for forest products from government forests.
The policy implication of this is that, while the new system of forest management may be
important for protection of existing forests, it could be counter-productive to increasing
the total forest cover in the country. An intensive campaign for farm forestry
development should therefore accompany it.
Secondly, secure land tenure is important for development of farm forestry because trees
take a longer time to mature. It is therefore important to guarantee security of access or
ownership of land in the country. Properly functioning land market could be one of the
options of ensuring that rights of land owners and land users are equally protected, and
that longer tenancy periods are enforceable. Thus, the government should work towards
developing land markets.
Third, small landholders are hesitant to plant trees possibly because the land cannot
accommodate both crops and trees. One thing that peasants should be made to understand
is that investing in trees could be more profitable than investing in food crops. Proper
education and introduction of high value fast maturing tree species could overturn the
current perception of the peasants.
Fourth, education level of the household head and his/her awareness of the Forest Act
(2005) favour development of farm forestry. This underscores the need to target
household heads in dissemination of information on importance of forests at household,
national and global level, and the content of the current Forest Act.
Lastly, social capital among households is useful in increasing forest cover. As a result, it
is important to develop infrastructures that promote formation and operation of social
groups. For instance, the government could provide tree seedlings or farm forestry
funding through community groups.
References
Agrawal,A.A.,Chhatre, A.,& Hardin, R. (2008). Changing governance in the worlds’ forests.
Science 320: 1460-1462.
Agrawal,A.A., and Ribot,J.C. (1999) Accountability in decentralization: a framework with south
Asian and West African cases. The Journal of Developing Areas 33:473-502.
Andersson, K.P.,Bauer,J., Jagger, P., Luckert,M., Meinzen-Dick, R.,Mwangi, E., and Ostrom, E.
(2008) Unpacking decentralization Paper presented at the SANREM CRSP Annual
Meeting, Los Banos, Phillipines, May 2008.
Aristei, D., F. Perali, and L. Pieroni. (2007). Cohort, Age and Time Effects in Alcohol
Consumption by Italian Households: A Double-hurdle Approach. Empirical Economics,
DOI 10.1007/s00181-007-0142-5.
Bannister, M.E. and P.K.R. Nair. (2003). Agroforestry adoption in Haiti: the importance of
household and farm characteristics. Agroforestry Systems 57: 149-157.
Boffa, J. M., L. Turyomurugyendo, J. Barnekow-Lillesø, R. Kindt. (2005). Enhancing Farm Tree
Diversity as a Means of Conserving Landscape-based Biodiversity: Insights from the
Kigezi Highlands, Southwestern Uganda. Mountain Research and Development 25 (3):
212–217
Bolfor. (2004). Bolivia sustainable forest management project report. Chemonics International
Inc., April 2004.
Borner, J., J. Mburu, P. Guthiga, and S. Wambua. (2009). Assessing opportunity costs of
conservation: Ingredients for protected area management in the Kakamega Forest,
Western Kenya. Forest Policy and Economics 11(7): 459-467
Cooke, P. (1998). The Effect of Environmental Good Scarcity on Own- Farm Labour Allocation:
The Case of Agriculture Households in Rural Nepal. Development and Environment
Economics 3: 443-469
de Janvry, A., and E. Sadoulet.( 2003). Progress in the Modeling of Rural Households' Behavior
under Market Failures. In: A. de Janvry and R. Kanbur, eds. Poverty, Inequality and
Development - Essays in Honor of Eric Thorbecke. New York: Springer, chap. 9, pp.
155-182.
de Janvry, A., M. Fafchamps, and E. Sadoulet. (1991). Peasant Household Behavior with
Missing Markets: Some Paradoxes Explained. Economic Journal, 101: 1400-1417.
Deininger, K. and G. Feder. (2001). Land Institutions and Land Markets: Handbook of
Agricultural Economics (B.Gardner and G.Raussser, Eds) 288-331.
Dewees, P.A. (1995). Trees on farms in Malawi: Private Investment, Public Policy, and Farmer
Choice. World Development 23(7):1085-1102.
Dove, M.R. (1995). The theory of social forestry intervention: the state of the art in Asia.
Agroforestry Systems 30(3):315-340.
Ellis, F. (1993). Peasant Economics: Farm Households and Agrarian Development. Cambridge
University Press.
Emerton, L. (1999). Mount Kenya: The Economics of Community Conservation. Evaluating
Eden Series, Discussion Paper No. 4.
Emtage, N., and J. Suh. (2004). Socio-Economic Factors Affecting Smallholder Tree Planting
and Management Intentions in Leyte Province, the Philippines. Small-Scale Forest
Economics, Management and Policy 3(2):257-271.
FAO. (2003). Country profile, Kenya.
Flood, L. and U. Grasjo. (1998). Regression Analysis and Time Use Data A comparison of
Microeconometric Approaches with Data from the Swedish Time Use Survey (HUS),
Göteborg University, School of Economics and Commercial Law, Working Papers in
Economics, No 5.
Flood, L. and U. Grasjo. (2001). A Monte Carlo Simulation Study of Tobit Models. Applied
Economics Letters, 8 (581-584).
Gebreegziabher, Z., A. Mekonnen, M. Kassie, and G. Kohlin. (2010). Household Tree Planting
in Tigrai, Northern Ethiopia: Tree Species, Purposes and Determinants. EfD Discussion
Paper Series. RFF.
German, G., F. K. Akinnifesi, A. K. Edriss, G. Sileshi, C. Masangano3 and O. C. Ajayi. (2009).
Influence of property rights on farmers’ willingness to plant indigenous fruit trees in
Malawi and Zambia. African Journal of Agricultural Research 4 (5), pp. 427-437
Ghate, R. (2004). Un-commons in the commons: Community Initiated Forest Resource
Management. Concept Publishing Company. New Delhi, India.
Gilmour, D.A. (1995). Rearranging trees in the landscape in the middle hills of Nepal. In:
Arnold, J.E.M. and P.A. Dewees (Eds). Tree management in farmer strategies: Responses
to agricultural intensification. Oxford.
Guthiga, P.M., J. Mburu, and K. Holm-Mueller. (2008). Factors Influencing Local
Communities’ Satisfaction Levels with Different Forest Management Approaches of
Kakamega Forest, Kenya. Environment Management 41:696-706.
Government of Kenya (GoK). (2007). Kenya Vision 2030. Government Printers. Nairobi,
Kenya.
Government of Kenya (GoK). (1994). Kenya Forestry Master Plan Development
(Implementation Programmes). Nairobi, Kenya.
Gronau, R. (1974). Wage camparisons: A selectivity bias. Journal of Political Economy
81:1119-1143.
Heckman, J. J. (1979). Sample Selection Bias as a Specification Error. Econometrica 47 (l): 153161.
Heckman, J. J. (1974). Shadow Prices, Market Wages, and Labour Supply. Econometrica
42:679-694.
Jacoby, H. (1993). Shadow Wages and Peasant Family Labour Supply: An Econometric
Application to the Peruvian Sierra. Review of Economic Studies 60: 903-921.
Kumar, C. (2003). Institutional reforms in Joint Forest Management: reflecting on experience of
Haryana Shivaliks. Paper presented at RCSD Conference: Politics of the Commons:
Articulating Development and Strengthening Local Practices. July 11-14, 2003, Chiang
Mai, Thailand
Lewis, H.G. (1974). Comments on selectivity bias in wage comparisons. Journal of Political
Economy 82(6):1145-1155.
Maddala, G.S. (1983). Limited-Dependent and Qualitative Variables in Econometrics.
Cambridge: Cambridge University Press.
Mbuvi, M.T.E., P.O. Ongugo, J.O. Maua, and C.K. Koech. (2007). Emerging values of forests: a
Blessing for Participatory Forest Management (PFM)-The case of Kenya. Paper
presented to the 3rd International PFM Conference. Addis Ababa, Ethiopia.
Mekonnen, A. (1999). Rural Household Biomass Fuel Production and Consumption in Ethiopia:
A Case Study. Journal of Forest Economics 5(1): 69-97.
Mogaka, H., G. Simons, J.Turpie, L.Emerton, and F. Karanja. (2001). Economic Aspects of
Community Involvement in Sustainable Forest Management in Eastern and Southern
Africa. IUCN —The World Conservation Union, Eastern Africa Regional Office,
Nairobi, Kenya.
Muneet, S.E.T. (2008). Factors Affecting Adoption of Agroforestry Farming System as a Mean
for Sustainable Agricultural Development and Environment Conservation in Arid Areas
of Northern Kordofan State, Sudan. Saudi Journal of Biological Sciences 15(1) 137-145.
Ongugo, P.O. (2007). Participatory Forest Management in Kenya: Is there anything for the poor?
Proceedings: International Conference on Poverty Reduction and Forests. September,
2007. Bangkok, Thailand.
Ongugo, P.O., M.T.E. Mbuvi, J.O. Maua, C.K. Koech and R.A. Othim. (2007). Emerging
Community Institutions for PFM process Implementation in Kenya. Paper presented to
the 3rd International PFM Conference. Addis Ababa, Ethiopia.
Ongugo, P.O and J.W. Njuguna. (2004). The Potential Effects of Decentralization Reforms on
the Conditions of Kenya’s Forest Resources. Paper presented to the Biennial Conference
of the International Association for the Study of Common Property (IASCP) on 3-13th
August, 2004. Oaxaca, Mexico
Oyono, P.R. 2004. Resources Management Decentralization in Damerooon. The Journal of
Modern African studies. 42 (1): 91-111.
Sabatier,P.A.,(1986) Top-down and bottom-up approaches to implementation research: acritical
analysis and suggested synthesis. Journal of Public Policy 6:21-48.
Sadoulet, E., A. de Janvry and C. Benjamin. (1996). Household behaviour with imperfect labour
markets. Working Paper No. 786. Department of Agricultural and Resource Economics,
Division of Agriculture and Natural Resources, University of California at Berkeley.
Sartori, A. (2003). An Estimator for Some Binary-Outcome Selection Models Without Exclusion
Restrictions. Political Analysis 11:111-138
Shively, G.E. (1999). Prices and Tree Planting on Hillside Farms in Palawan. World
Development 27(6):937-949.
Sigelman, L. and L. Zeng (1999). Analyzing Censored and Sample-Selected Data with Tobit and
Heckit Models. Political Analysis 8(2): 167- 182.
Singh, I., L. Squire, and J. Strauss, eds. (1986). Agricultural Household Models: Extensions,
Applications, and Policy. Baltimore MD: The Johns Hopkins University Press.
Skoufias, E. (1994).Using shadow wages to estimate labour supply of agricultural households.
American Journal of Agricultural Economics 76: 215-227.
Taylor, J. E., and I. Adelman. (2003). Agricultural Household Models: Genesis,
Evolution and Extensions. Review of Economics of the Household, 1(1).
Thenya, T., E. Nahama, and B. Wandago. (2007). Participatory Forest Management Experience
in Kenya (1996-2006). Proceedings of the 1st National Participatory Forest Management
Conference, 6-8 June 2007. Nairobi, Kenya.
Thorbecke, E. (1993). Impact of State and Civil Institutions on the Operation of Rural Markets
and Nonmarket Configurations. World Development 21(4): 591-605.
Warner, K. (1995). Patterns of tree growing by farmers in Eastern Africa. In: Arnold, J.E.M. and
P.A. Dewees (Eds). Tree management in farmer strategies: Responses to agricultural
intensification. Oxford.
World Bank (2009) Community driven development. http://go.worldbank.org/24k81HVVso
(accessed 8th December, 2009).