Fertilizer subsidies and private market participation: The case of Kano State, Nigeria.
Lenis Saweda O. Liverpool-Tasie
Michigan State University
[email protected]
Prepared for presentation at the CSAE conference March 2014
ABSTRACT
This article estimates the effect of a fertilizer voucher program on farmer participation in the
private fertilizer market in Nigeria. Using a double hurdle model (to address corner solution
challenges with estimating input demand) and a control function approach (to account for the
endogeneity of subsidized fertilizer receipt), the study finds evidence that receiving subsidized
fertilizer in Kano, Nigeria increased both the probability and extent of participation in the private
fertilizer market. Findings demonstrate that under certain circumstances e.g. where input dealers’
links to farmer are weak; there could be significant gains from the temporary use of voucher
programs to strengthen such links.
JEL: Input vouchers, crowding in, crowding out, windfall gains, Nigeria, sub Saharan Africa
1
INTRODUCTION:
Historical increases in agricultural productivity have often been linked with significant increases
in chemical fertilizer use. The combination of low yields and low rates of fertilizer use in African
countries has resulted in periodic calls for government subsidies (Ellis, 1992; Sachs, 2004;
Crawford et al, 2006). Responding to failures of past fertilizer schemes, current policy debates
and initiatives focus on the use of input vouchers as smart subsidies which are targeted and
temporary (Gregory 2006; Minot and Benson 2009).
Though “market smart” subsidies should foster (crowd in) rather than hamper (crowd out)
private sector fertilizer markets, the effectiveness of these new subsidy programs in Africa and
their impact on the private sector has remained debated, with limited empirical evidence. Xu et
al (2009) 1 and Ricker-Gilbert et al (2011) are exceptions; exploring the effect of input vouchers
in Zambia and Malawi respectively. They find different impacts of vouchers on private market
activity depending on the type of private market and level of targeting. This study contributes to
this limited literature by exploring the Nigerian situation; where no empirical studies of the effect
of input vouchers on private market participation have been conducted till date.
Most empirical studies on input vouchers and private input markets have focused on contexts
where input vouchers are received by relatively well-off farmers and fertilizer use is relatively
1. Mason and Jayne (2013) explore this issue in Zambia again updating Xu et al (2009) and using a
similar approach as Ricker-Gilbert et al (2011).
2
low (Ricker-Gilbert et al, 2011; Mason and Jayne, 2013). This study focusses on a context (a
state within Nigeria) where fertilizer use is relatively common with a mean of about 165kg per
hectare in the 2010 Nigerian living standards measurements survey- Integrated survey on
agriculture. Furthermore Liverpool-Tasie, Banful and Olaniyan (2010) 2 show that where wealth
was significantly different between participant and non-participants in the pilot voucher program
being studied in Nigeria, participants tended to have fewer assets. They were also more likely to
be renting agricultural land rather than owning it. As far as we are aware, no studies have been
conducted to explore if lower private market participation will obtain if input vouchers are used
in this kind of scenario. Consequently, this study contributes to this discussion; informing on
how the effects of targeted subsidies on farmers’ participation in private markets can vary across
different contexts. Throwing more light on this question is important, from both an analytical and
policy perspective.
Nigeria is the most populous country in Africa (about 158 million) and the second largest
economy (World Bank 2010). Nigeria alone accounts for 23 percent of the entire fertilizer
consumption in sub Saharan Africa in 2008/2009(IFDC 2009).Given its size, the regional
fertilizer market in West Africa is significantly affected by what goes on in Nigeria. The
potential effects of national policies geared to increase farmer use of fertilizer often extend
beyond Nigeria’s border. Consequently understanding the effects of input vouchers on the
private fertilizer market in a country like Nigeria is expedient for any strategic effort to improve
the sustained use of inputs in sub Saharan Africa (SSA).
2
They use the same data that is used in this study.
3
Section 2 of this article provides a brief discussion of fertilizer vouchers and private market
development generally and in Nigeria. Section 3 provides a description of the 2009 voucher
program in Kano State linking its design and implementation to the potential effects it could
have on the private market. Section 4 and 5 discuss the conceptual and empirical framework for
this study. Section 6 describes the data while section 7 provides the study results. Section 8
presents further considerations of the empirical findings and section 9 concludes.
FERTILIZER VOUCHERS AND PRIVATE FERTILIZER MARKET DEVELOPMENT
Generally, when private sector enterprises are negatively affected by public sector distribution of
subsidized inputs (e.g. under a parallel distribution system), the concern of crowding out or
private sector displacement occurs. Sales diverted from farmers who would have otherwise
purchased their inputs from the private sector but participate in a government subsidy program
could negatively affect the private sector. Market diversion serves as a disincentive for private
investments in the fertilizer subsector. Lack of investments can have far reaching effects on the
ability to stimulate a competitive commercial fertilizer market; necessary for sustained
agricultural growth in many African nations.
In recent years, various forms of input voucher programs have been developed and implemented
in developing countries. Malawi has used input vouchers in its nationwide fertilizer and seed
subsidy programs since 1999. Zambia (in 2003), Tanzania (in 2008) and Ghana (2008 and 2009)
are other examples where input vouchers have been recently used (Longley et al 2003; Gregory
2006; Minot and Benson 2009; IFDC 2010). Limited empirical evidence of the effect of these
4
targeted subsidies in sub Saharan Africa exists. The effect of targeted input subsidies on crop
production and consequently growth and poverty reduction has been studied (Dorward et al,
2008). Others have evaluated the effect of targeted input vouchers on farmer’s timely access to
affordable inputs (Liverpool-Tasie (2013) or the political motivation behind the allocation of
vouchers (Banful, 2011)). However, Mason and Jayne (2013), Ricker-Gilbert et al (2011) and Xu
et al, (2009) are among the few empirical studies which have looked at the effect of fertilizer
voucher programs on private market participation. Takeshima et al (2013) has looked at the
effect of a traditional fertilizer subsidy program (not vouchers) on private market participation in
Nigeria.
In most of these countries, fertilizer subsidies have been shown to crowd out private
commercial fertilizer sales (Mason and Jayne (2013); Takeshima et al (2013); Ricker-Gilbert,
Jayne and Chirwa 2011; Xu et al 2009).
The extent of distortionary effects of fertilizer subsidies depend on several factors. These include
the size of fertilizer demand (Dorward 2009) and the level of development of the private sector
(Xu et al 2009). Xu et al (2009) find that where the private sector is well developed, crowding
out effects are larger. However, in poorer areas where private commercial markets are weak,
subsidies tend to crowd in commercial sales as they create demand. Ricker-Gilbert, Jayne and
Chirwa (2011) find that input vouchers reduce farmer likelihood of participating in the
commercial fertilizer market as well as the quantity purchased. They attribute this largely to poor
targeting of voucher programs to relatively less poor households. Mason and Jayne (2013) find
similar results to Malawi in their study in Zambia.
While these explanations are possible, several factors could cause a fertilizer voucher to increase
farmers’ participation in the private fertilizer market.
5
For example, input vouchers could
stimulate fertilizer use among farmers for which a credit constraint on fertilizer demand is
binding.
Learning by farmers’ is another potential mechanism through which access to
subsidized fertilizer might be positively correlated with commercial purchases from the private
market. Farmers’ positive experience (higher output or healthier crops) with subsidized fertilizer
could increase their demand for fertilizer from commercial markets. Yet, another mechanism on
the demand side is an income effect where higher revenue from higher outputs when subsidized
fertilizer is used expands farmers demand for fertilizer and this demand is
met in the private
market. Alternatively, the guaranteed demand from a voucher program could have economies of
scale effects on fertilizer suppliers (Dorward, 2009), encouraging them to expand their
operations in rural areas. This expansion could increase fertilizer provision to areas previously
excluded and reduce costs associated with fertilizer procurement (e.g. due to shorter distances to
access the input), expanding fertilizer use generally.
One contribution of this article is its ability to capture the effect of a subsidy program that
appears to have reached relatively poorer households. Arguably, these are the farmers who
would normally not participate in the market at all or whose participation in terms of quantity
might be small. Since we do not see a clear targeting system focused on the poor, we assume that
the program might have reached poorer farmers because it made lots more people have access to
a significant amount of fertilizer, through the nature of voucher distribution than was the case in
the past.
Distinct from most of the relevant literature on this topic, this study recognizes that rather than a
situation where two parallel distribution channels are used for an input (as was the case in Xu et
al (2009), we are dealing with a case where both the subsidized and commercial fertilizer were
6
being distributed by the private sector. Consequently, the private sector as a whole is not
necessarily expected to be negatively affected under such an arrangement. What could obtain
however is an efficiency challenge in the subsidy administration process, if poor targeting results
in the windfall gain associated with participating in the program going to farmers who need it
less.
What is more problematic in this scenario is if certain suppliers are excluded from
participating in the program as such suppliers might be crowded out of the market (Krausova and
Banful (2010). In addition, if the size of the subsidy program is large, this can create a
dependency problem and limit the development of the private fertilizer sector independent of the
government. In this particular study context, the effect of a subsidy program on private sector
participation consequently the private sector is not obvious apriori. This study does not explore
the effect of the voucher program on the entire private fertilizer market but focusses on
identifying if the program had any effect on farmer’s participation in the private fertilizer market
and possible explanations for this.
THE 2009 VOUCHER PROGRAM IN KANO STATE, NIGERIA
Studies have shown that policy inconsistency, diversion of fertilizer from farmers, late access to
fertilizer; high prices and poor fertilizer quality are major constraints to fertilizer use in Nigeria
(Banful et al, 2010, Liverpool-Tasie et al, 2010). The use of vouchers in Nigeria was proffered as
a potential solution to the shortcomings of subsidized fertilizer which was largely distributed by
the government 3. The 2009 voucher program was piloted in two Nigerian states; Kano and
Taraba. Kano, located in North Western Nigeria is the most populous Nigerian state with about
3
The traditional system of government procurement and distribution of subsidized fertilizer in Nigeria has been
characterized as persistently delivering fertilizer late with the diversion of fertilizer from the intended beneficiaries
(Nagy and Edun, 2002). Leakages of the product into the regular market were common, distorting the market
price and providing arbitrage opportunities.
7
9.4 million residents (NPC 2006). The primary activities in Kano are commerce and agriculture
and the poverty rate in Kano is about 75percent (NBS 2009). Fertilizer use in Kano is not new
and according to the 2010 LSMS-ISA data, fertilizer use in Kano State is significantly higher
than many other Nigerian states.
The 2009 voucher program in Kano was a collaborative effort between the government (Federal
and State), the private sector suppliers and dealers and the International Center for Soil Fertility
and Development (IFDC). The program (designed to deliver subsidized fertilizer to small holder
farmers across the 44 local government areas (LGA’s) of Kano) was meant to model smart
subsidies that allow for targeting of rural smallholder farmers and develop a private sector supply
and distribution channel (IFDC, 2010).
Three fertilizer suppliers and over 150 private sector agro dealers participated in the program.
Participating farmers were provided with vouchers, which were redeemable at certified
agricultural input dealers within their LGA of residence. The value of the voucher was a N2000 4
discount per bag on two bags of Nitrogen Phosphorous Potassium (NPK 15:15:15) and one bag
of Urea (46% Nitrogen). Farmers were required to pay the difference between the market price
and the N2000 discount per bag.
Participants in the 2009 voucher program in Kano were required to be members of a farmer
group 5. Each participating farmer group received a single voucher that entitled each of its
members to the N2, 000 discount per bag on three bags of fertilizer. According to the Nigeria
Agri Markets Information Service (NAMIS), fertilizer prices in central markets in Kano were
4
This is about $13.50 at N150=$1
Previous studies described how credit and subsidized inputs in Nigeria were funneled to and captured by
“absentee farmers, retired civil servants, and soldiers Olayide and Idachaba (1987), the focus of the voucher
program was to reach farmers and not necessarily the poorest as is often cited as an appropriate targeting
criterion in subsidy programs
5
8
about N3000 for a 50Kg bag of NPK 15:15:15 and N3, 200 for a 50Kgbag of Urea. Thus the
voucher value was slightly over 60 percent and 65 percent of the NPK and Urea market price
respectively. For verification, farmer groups were required to bring their certificate of
registration to verify their group’s authenticity (IFDC 2010).Once verified, any representative of
the group leadership (Chairman, Secretary or Treasurer) could redeem the groups voucher, as
individual members were not required to be present when the voucher was being redeemed
(IFDC 2010).This structure of the 2009 voucher program in Kano created an opportunity for
differential experience of the voucher program across farmers of the same group. Anecdotal
evidence drawn from complaints on the field indicated that individuals got different quantities
than the three bags they had been promised. Liverpool-Tasie, Banful and Olaniyan (2010) show
that the friends or relatives of the farm group president received more bags of fertilizer than
those without such relationships. This article capitalizes on this characteristic of the program in
its identification strategy.
CONCEPTUAL FRAMEWORK
Our study area is rural Nigeria, where farmers face limited access to various input and output
markets. Consequently, we recognize the existence of several market failures and model the
fertilizer demand decision as a constrained utility maximization problem for a household as
characterized in Sadoulet and de Janvry (1995) and de Janvry and Sadoulet( 2006). Here,
household characteristics are expected to affect household demand for fertilizer. In addition to
household characteristics, we incorporate the fact that farmers have two channels through which
they receive fertilizer; regular fertilizer through the private market and subsidized fertilizer from
9
the government 6. Consequently our input demand model following Singh, Squire and Straus
(1986) can be expressed as:
𝑄𝐹𝑒𝑟𝑡𝑝 = 𝑓(𝑄𝐹𝑒𝑟𝑡𝑠 , 𝑃𝐹𝑒𝑟𝑡, 𝑃𝑜𝑢𝑡𝑝𝑢𝑡 , 𝐾, 𝐴, 𝑍 )
(I)
where 𝑄𝐹𝑒𝑟𝑡𝑝 refers to the quantity of fertilizer purchased from the private sector distribution
channel 𝑄𝐹𝑒𝑟𝑡𝑠 refers to the quantity of subsidized fertilizer received 𝑃𝐹𝑒𝑟𝑡 and 𝑃𝑜𝑢𝑡𝑝𝑢𝑡 refer
to the prices of fertilizer and the major crop (maize and/or rice ) produced by over 70 percent of
households. K and A are access to credit and land size respectively and Z refers to other
household characteristics and socioeconomic variables that the previous literature document
affect demand for inputs. They include credit availability (Coady 1995; Croppenstedt, Demeke
and Meschi 2003; Odhiambo and Magandini 2008), transactions and transportation costs which
have been demonstrated to affect input demand ((Morris et al. 2007; Winter-Nelson and Temu
2005) as well as household participation in both input and output markets (de Janvry et al. 1991;
Key et al. 2000; Bellemare and Barrett 2006; Croppestedt, Demeke, and Meschi 2003; de Janvry,
Fafchamps, and Sadoulet 1991; Holloway, Ehui, and Teklu 2008; Holloway, Barrett, and Ehui
2005;Key, Sadoulet, and de Janvry 2000).
EMPIRICAL FRAMEWORK AND ESTIMATION STRATEGY
To estimate the degree to which subsidized fertilizer affects farmers’ participation in the private
commercial fertilizer market, the conceptual model above can be specified as follows:
𝑄𝐹𝑒𝑟𝑡𝑝𝑖 = 𝛽𝑋𝑖 + 𝛿𝑄𝐹𝑒𝑟𝑡𝑠𝑖 + 𝑢𝑖
6
(1)
Note that both subsidized and unsubsidized fertilizers were distributed via private sector dealers.
10
where 𝑄𝐹𝑒𝑟𝑡𝑝𝑖 refers to the quantity of fertilizer purchased by farmer i from the private market
in 2009. 𝑋𝑖 is a vector of controls that affect private sector fertilizer demand. It includes
household demographic information, socioeconomic characteristics and variables to capture
access to credit and transactions cost. It also includes the price of fertilizer as well as the price of
output crops on which fertilizer is applied. 𝑄𝐹𝑒𝑟𝑡𝑠𝑖 refers to the quantity of subsidized fertilizer
that farmer i received in 2009. 𝑢𝑖 is a farmer specific error term and 𝛽 and 𝛿 are parameters to be
estimated. Our primary interest is in estimating the parameter δ which indicates the presence and
extent to which access to subsidized fertilizer affects farmers’ participation in the private
fertilizer market.
Two key problems in estimating the effect of subsidized fertilizer on commercial fertilizer
demand are the endogeneity of the quantity of subsidized fertilizer received and the corner
solution nature of input demand. When the decision not to use inputs is an optimal choice
(where use is not profitable at prevailing market prices as in Coady (1995)), a zero observation
for the quantity of an input used reflects an optimal choice rather than an unobserved one. This
precludes the use of OLS estimations as the constraint of y≥0 automatically implies that we do
not have constant marginal effects and 𝐸(𝑌|𝑋, 𝑌 > 0) ≠ 𝛽X. It also precludes Heckman,
LaLonde and Smith (1999) type selection models designed for incidental truncation where the
zeros are unobserved values. Like Ricker-Gilbert et al (2011), This article uses a double hurdle
approach which allows the process that determines the decision to participate in the private
market to be different from that which determines the extent of participation. Variables like
distance may affect the decision to participate but distance need not play a significant role in the
quantity of inputs purchased. However, it is possible that even after the decision to participate
has been made, individuals further away from the market (and with sufficient capital) might
11
prefer to buy more (to minimize the average cost of inputs) than a farmer who is closer to the
market.
In equation (1), following Roy (1951) and Cameron and Trivedi (2005; 2009) , 𝑄𝐹𝑒𝑟𝑡𝑝𝑖 = 0 is
determined by the density 𝑓1 (. )
such that 𝑃(𝑄𝐹𝑒𝑟𝑡𝑝𝑖 )=0 = 𝑓1 (0) and 𝑃(𝑄𝐹𝑒𝑟𝑡𝑝𝑖 )>0 is
determined by 𝑓2 (𝑄𝐹𝑒𝑟𝑡𝑝𝑖 |𝑄𝐹𝑒𝑟𝑡𝑝𝑖 > 0)= 𝑓2 (𝑄𝐹𝑒𝑟𝑡𝑝𝑖 )|1 − 𝑓2 (0)7 .The associated likelihood
function whose log is maximized can be expressed as:
𝐿 = ∏𝑖|𝑄𝐹𝑒𝑟𝑡𝑖 =0{𝑓1 (0) } ∏𝑖|𝑄𝑓𝑒𝑟𝑡𝑖 ≠0 �
1−𝑓1 (0)
𝑓 �𝑄𝐹𝑒𝑟𝑡𝑝𝑖 ��
1−𝑓2 (0) 2
(2)
Though concerned with estimating delta (𝛿) in equation (1), the quantity of subsidized fertilizer
received by a farmer 𝑄𝐹𝑒𝑟𝑡𝑠𝑖 in (1) above is potentially endogenous if there are unobserved
characteristics that affect the quantity of subsidized fertilizer a farmer receives that are also
likely to affect their demand for non-subsidized fertilizer. For example, if more motivated
farmers are better coordinated with other farmers, they would be more likely to participate in the
voucher program 8 (and receive more subsidized fertilizer) but also be more likely to already be
using fertilizer and thus purchase more fertilizer from the private market. Alternatively, if high
performing farmers are influential in their community, they could receive more bags of
subsidized fertilizer than their peer but could also be more likely to have already been using
7
This is multiplied by P(Qfert pi )>0 to ensure that the sum of probabilities sum to one.
8
Farmers had to be in organized farm groups in Kano to be able to participate in the voucher
program.
12
fertilizer and thus more likely to participate in the commercial fertilizer market. Either of these
would make estimates of the effect of the quantity of subsidized fertilizer on the quantity of nonsubsidized fertilizer purchased by a farmer biased.
For the control function approach and given the corner solution nature of𝑄𝐹𝑒𝑟𝑡𝑝𝑖 , the study
estimates:
𝐸(𝑄𝐹𝑒𝑟𝑡𝑝 |𝑄𝐹𝑒𝑟𝑡𝑠 , 𝑍, 𝑟) =𝐸(𝑄𝐹𝑒𝑟𝑡𝑝 |𝑄𝐹𝑒𝑟𝑡𝑠 , 𝑍1 , 𝑟) = Ф (𝑌𝛽+𝑟)
𝑄𝐹𝑒𝑟𝑡𝑠 = 𝑍𝛾 + 𝑣
(3)
(4)
where 𝑌 is a general, nonlinear function of (𝑄𝐹𝑒𝑟𝑡𝑠 , 𝑍1 ) and 𝑟 is an omitted factor which is
likely correlated with 𝑄𝐹𝑒𝑟𝑡𝑠 . The exclusion restriction associated with the first part of (3) is that
a subset of 𝑍, (𝑍1 ) appears in𝐸(𝑄𝐹𝑒𝑟𝑡𝑝 |𝑄𝐹𝑒𝑟𝑡𝑠 , 𝑍, 𝑟). Following Wooldridge (2007) and
Wooldridge (2008), we estimate a first stage regression of the quantity of subsidized fertilizer
received by a farmer (𝑄𝐹𝑒𝑟𝑡𝑠𝑖 ) using a Tobit model. Then the generalized residual is constructed
as:
𝑔𝑟
�𝑖 = −𝜏̂ 1[𝑄𝐹𝑒𝑟𝑡𝑠𝑖 = 0] λ (−𝑍𝑖 𝛾�) + 1[𝑄𝐹𝑒𝑟𝑡𝑠𝑖 > 0](𝑄𝐹𝑒𝑟𝑡𝑠𝑖 − 𝑍𝑖 𝛾�)
(5)
Where 𝜏̂ and 𝛾� are the Tobit MLEs and λ is the inverse Mills ratio. Then the generalized
residuals are included in the second stage estimations (Wooldridge 2008).
For the second stage estimations, a double hurdle model is used. The double hurdle model used
in the second stage estimations follows Cameron and Trivedi (2009). The first step is the
decision to purchase fertilizer from the private market or not while the second step is the decision
about the quantity of fertilizer to purchase from the private market once the decision to
13
participate has been made. The double hurdle model relaxes the assumption that the zeros and
the positive values come from the same data generating process. The model assumes that the
zeros are determined by the density 𝑓1 (. ) such that 𝑃(𝑄𝐹𝑒𝑟𝑡𝑝𝑖 )=0 = 𝑓1 (0) and 𝑃(𝑄𝐹𝑒𝑟𝑡𝑝𝑖 )>0 is
determined by
1 − 𝑓1 (0). The
positive quantities come from the truncated density
𝑓2 (𝑄𝐹𝑒𝑟𝑡𝑝𝑖 |𝑄𝐹𝑒𝑟𝑡𝑝𝑖 > 0)= 𝑓2 (𝑄𝐹𝑒𝑟𝑡𝑝𝑖 )/{1-𝑓2 (0)}, which is multiplied by 𝑃(𝑄𝐹𝑒𝑟𝑡𝑝𝑖 )>0 to
ensure that the probabilities sum up to 1. Since the two parts of the model are functionally
independent, maximum likelihood estimation of the hurdle model is achieved by separately
maximizing the two terms in the likelihood, one corresponding to the zeros and the other to the
positive values. Consequently, the first part uses the full sample while the second part only uses
the positive observations. For the first hurdle, we estimate a probit model for the entire sample
and for the second stage we estimate a truncated regression. In both hurdles, the generalized
residuals are included as covariates.
Identification when using the control function approach requires an instrument used to estimate
(4) which are appropriately excluded from our primary equation (1). This article uses a variable
that captures if the respondent was a relative of the farm group leader, secretary or treasurer as an
instrument for the quantity of subsidized fertilizer that a farmer received. While being related to
the farm group leadership would affect the quantity of fertilizer received by a farmer, being
related to the group leadership is not likely to be correlated with unobserved characteristics that
are likely to drive commercial fertilizer demand such as farmer ability. Therefore, this study
assumes that upon controlling for other characteristics that could be correlated with a households
relationship to the farm group leadership in the demand for commercial fertilizer, the variable
satisfies the exclusion restriction of not being correlated with the error term of equation (1) and is
14
considered an appropriate instrument. Being related to the farm group leadership is consequently
excluded from our estimation of equation (1). In all second stage estimations, p values are
estimated via bootstrapping at 500 repetitions (except where otherwise stated) to account for the
fact that the generalized residual came from a first stage regression estimation. With a cross
sectional dataset, we are not able to completely rule out the effect of time invariant
characteristics (like farmer ability) which could simultaneously be correlated with their ability to
secure subsidized fertilizer, while also affecting their demand for fertilizer in the private market.
However, in addition to the instrumental variable approach adopted, additional controls to
capture possible variation across farmers are included in the estimations and a falsification test is
conducted to further demonstrate that it is unlikely that there is some unobservable variable
correlated with relationship to the leadership that also causes market purchases of fertilizer.
District dummies are included in all estimations to capture time invariant (or slowly changing)
geographic factors that are likely to affect input demand. These include soil characteristics,
weather and infrastructural development.
Data
The data used in this article come from a survey of 640 households in Kano (North-West
Nigeria). Interviewed households were selected from 10 randomly selected Local Government
Areas (LGAs); administrative units constituting the third tier of the administrative structure in
Nigeria. Households were randomly selected from the randomly selected local governments in
Kano. Details of the sampling and survey procedures are included in the appendix. The survey
respondents were largely household heads, their spouses and other adult household members.
Respondents were interviewed about their participation in various farm groups and other
community associations, their leadership positions in their farm group and local communities,
15
their farming practices (input use, sources and prices) and about their participation in the 2009
voucher program. Household demographic information was also collected. Because more than
one household member could have participated in the voucher program, standard errors are
clustered at the household level in all estimation s.
Table 1 demonstrates that participation in the private fertilizer market differs significantly across
respondents who participated in the voucher program and those who did not. The quantity of
fertilizer purchased in the private market was significantly higher among those who participated
in the voucher program than those who did not. However, there was substantial participation in
the private market irrespective of the program, as about 70 percent of respondents who did not
participate in the voucher program still purchased fertilizer from the private market. Table 2
shows that the average extent of participation ranged from about 0.3bags (15kg) from the lowest
10percent of subsidized fertilizer recipients in the sample (0 bags were bought by the lower
5percent) to about 12 bags (600Kg) for the highest 10percent.
Table 1. Private sector participation by participation status in the 2009 voucher program
Non Participants
Voucher program
participant
Purchased
fertilizer from
market(1/0)
the private
0.720
0.937*
(0.449)
(0.243)
Number of 50Kg bags of non-subsidized 3.009
5.004*
fertilizer (bags) purchased
(4.177)
(7.683)
Received subsidized fertilizer
0.387
1.000*
(0.488)
(0.00)
16
Non Participants
Voucher program
participant
Number of 50kg bags of subsidized
fertilizer received (bags)
1.224
3.699*
(3.581)
(5.962)
Source: Generated by author with data from the fertilizer voucher program evaluation survey
* means the means are statistically significantly different at 5% or less
Table 1 also shows that a significant portion of respondents received subsidized fertilizer even
though they did not participate in the voucher program. Slightly less than 40 percent of
respondents who did not participate in the 2009 voucher program still received subsidized
fertilizer. In addition to the Federal Subsidy, farmers had access to another fertilizer subsidy
program at the local government level which gave farmers one bag of fertilizer each at a fixed
price of N1000. While the average number of bags of fertilizer received by program participants
was about 3.6 bags (compared to 3 bags the program stipulated), it was about 1.2 bags for nonparticipants; what we would expect if they participated in this alternative program mentioned
above.
The average respondent was a member of a farm group that tended to purchase inputs together
and lived about 33 kilometers away from the major trading market in the state. Similar to RickerGilbert, Jayne and Chirwa (2011), this study assumes a naive expectation of the average real
maize price for the six months prior to the planting season, which was November to April in
Kano in 2009. The price ranged from N52/kg for respondents at the lowest 10percent of the price
distribution to N55/Kg for those in the highest 10percent. The reported price paid for fertilizer
17
purchased in the private market was about N4000 a bag (N80/kg). The average household size
was about 4 people headed by a married male.
Table 2. Descriptive statistics of study variables
Variable
Mean
Number of 50Kg bags of non-subsidized
fertilizer (bags) purchased
0.00
1.00
3500
4600
4.87
57.38
49.50
55.00
0.00
1.00
2.00
7.00
0.28
(0.48)
Household size
1.00
52.06
(18.80)
Member of a group that provides credit to
members in 2008 (1/0)
0.00
33.01
(20.19)
Naïve price based on representative price 4
months prior to farming season (Naira per Kg)
6.00
4038.19
(575.23)
distance of respondent to major market in the
state (KM)
0.00
0.66
(0.47)
Reported price paid for fertilizer in 2009
(Naira per 50Kg bag)
12.00
0.89
(0.32)
Farmer related to the farm group leadership
(1/0)
0.30
2.85
(5.32)
Farmer member of a group that purchased
fertilizer together in 2008
Highest
10%
6.51
(12.99)
Number of 50kg bags of subsidized fertilizer
received (bags)
Lowest
10%
3.89
(1.47)
18
Variable
Mean
Age in years
34.24
(13.33)
Sex (1=Male)
Lowest
10%
Highest
10%
19.00
53.00
0.00
1.00
6.00
12.00
0.00
1.00
0.75
10.00
0.00
1.00
0.00
1.00
0.00
1.00
0.51
(0.50)
Number of years of education
7.60
(3.25)
Marital Status (1/0)
0.85
(0.36)
Land area in 2008 (hectares)
4.36
(6.06)
Used improved seed in 2008(1/0)
0.52
(0.50)
Rented Agricultural land (1/0)
0.10
(0.30)
Household owns a motorcycle (1/0)
0.065
(0.48)
Total Livestock Units (TLU)
6.86
0.00
11.10
(20.20)
Source: Generated by author with data from the fertilizer voucher program evaluation survey
Note: Standard deviation are in parenthesis.
19
ANALYSIS AND RESULTS
To identify the factors affecting the quantity of subsidized fertilizer respondents received,
equation (4) is estimated using a Tobit model to account for selection into the voucher program.
Results of this are shown in table 3. The first stage results reveal that whether the recipient was
in a farm group that purchased fertilizer together was highly correlated with the quantity of
subsidized bags received 9. Wealth and education variables do not appear to be a significant
determinant of the number of bags recipients received but being married and the local
government where farmers lived was. The instrumental variable; “Whether the respondent was a
relative of the farm group president, secretary or treasurer” is significantly and positively
associated with the number of bags received. The average partial effects (APE) were estimated 10
and reveal that being a relative of one of the farm group leaders increased the number of bags
received by about 1.6 and this was significant at 1%. As there is no known test for IV strength in
non-linear models, we test the strength of our IV by the partial correlation of the “being a relative
of the farm group leadership” in the reduced form equation (Ricker-Gilbert, Jayne and Chirwa
2011). The high t statistic (2.03) and p value of 0.04 are evidence that the IV is strongly
correlated with the endogenous variable. We argue that it is highly unlikely for one’s relationship
to the farm group president to affect participation in the private market after conditioning on
other factors. Thus it is considered appropriately excluded from the second stage estimations.
9
We check for multicollinearity between the control variables and find none to have a correlation coefficient
above 0.4. We also fail to drop any of our control variables in the application of a multicollinearity test on the
variables prior to estimation
10
This was estimated using the margins command in STATA and the associated standard errors and p values were
generated using the delta method.
20
Table 3. First stage regression results of the Tobit model
Number of bags of subsidized fertilizer
Farmer member of a group that purchased
fertilizer together
Price of fertilizer
Distance to main market
Price of Maize (N/kg)
Member of a farm group that gives credit to
members
Household size
Age of respondent (in years)
Male (1/0)
Years of education
Married (1/0)
Land area ( in hectares)
Uses improved seed (1/0)
Rents land (1/0)
someone In household owned a motorcycle
(1/0)
Total Livestock Units
Respondent is related to the farm group
leadership
LGA Dummies included
Number of observations
Pseudo R
2
Coefficient
p value
0.764
0.288
0.376
17.543
29.87
0.001
0.007
0.004
-0.074
0.912
0.184
-0.008
0.009
-0.023
0.837
-0.052
-0.379
1.449
0.702
0.590
0.976
0.713
0.050
0.417
0.599
0.259
0.259
0.674
-0.007
0.368
1.677
0.045
YES
1569
0.06
Source: Generated by author using STATA Note: *,** and *** indicate p values significant at 10%, 5%
and 1% respectively
Next, the generalized residual (𝑔𝑟
�𝑖 ) is constructed using the maximum likelihood estimates
from the Tobit regression and the inverse mills ratio as defined in (5). The generalized residual is
then included in the second stage estimation. Following Imbens and Wooldrige (2007) and Garen
(1984), the generalized residual is also interacted with the quantity of subsidized fertilizer
21
received to recover consistent estimates of (𝛿) in equation (1). Consequently, in the second stage
we estimate:
𝑄𝐹𝑒𝑟𝑡𝑝𝑖 = 𝛽𝑋𝑖 + 𝛿𝑄𝐹𝑒𝑟𝑡𝑠𝑖 + 𝜌1 𝑔𝑟
�𝑖 + 𝜌2 𝑔𝑟
�𝑖 ∗ 𝑄𝐹𝑒𝑟𝑡𝑠𝑖 + 𝑢𝑖
(6)
Where all variables are as earlier defined and 𝜌1 and 𝜌2 are the coefficients associated with the
generalized residual and its interaction with the quantity of subsidized fertilizer received. The
significance of these variables in the second stage results both tests and controls for the
endogeneity of the quantity of subsidized fertilizer received (Rivers and Vuong 1988; Smith and
Blundel 1986; Vella 1993).
Results from the second stage regressions are displayed in table 4. All results are expressed in
terms of the average partial effects. The results indicate that an additional bag of subsidized
fertilizer increased the probability of participating in the private commercial market by 0.09.
Though statistically significant, this effect is quite small. This is not surprising given the large
fraction of participants in the private market generally and thus the limited variation in the
private market participation variable. Data on fertilizer use in Nigeria have indicated that over 50
percent of households in Kano reported using some fertilizer (Fadama III baseline data 2010).
Having so many respondents in our sample participated in the private commercial market in
2009 might be further evidence of the impact of the voucher program on fertilizer availability
across the state. By nature of the programs implementation, suppliers were required to establish
presence in all the LGAs at certain periods to distribute the subsidized fertilizer as determined by
the program. If suppliers supplied quantities beyond the amount needed in the program in the
22
various LGAs, this could explain the high level of participation revealed in the data 11. Input
suppliers being physically present in the LGA could have had significant implications on
fertilizer access for many farmers. In addition to just making the product available in areas where
input suppliers might have been previously absent, it also could have significantly reduced the
transactions and transportation costs associated with procuring fertilizer.
As expected, fertilizer price is negatively associated with participation in the private market
while access to credit is positively correlated. Farmers with more livestock and more education
are also more likely to participate in the private market while households who rent rather than
own their land are less likely to participate in the private market. Renting in land likely captures
a wealth effect.
11
It should be noted that the extent of participation varies significantly in the sample. Many
farmers purchased less than half of a 50Kg bag. The quantity purchased ranges from a third of a
50Kg bag at the lowest10% of private market fertilizer purchases to 12 bags for the highest 10%
of the distribution.
23
Table 4. Second stage estimation results of private fertilizer market participation
Hurdle 1
Coefficients are Average partial effects (APE)
Number of subsidized bags of fertilizer received
Hurdle 2
Quantity
of
fertilizer
Participation in the
purchased
from
the
private market
market
Coefficient p value Coefficient p value
0.090
0.002
0.257
0.001
Farmer member of a group that purchased 0.062
fertilizer together
-0.069
Price of fertilizer
Distance to main market
Price of Maize (N/kg)
0.002
0.392
0.000
-0.013
-0.042
0.050
0.224
0.000
-0.109
0.096
10.8
0.172
-1.671
0.110
0.008
-0.088
0.233
0.123
0.063
0.112
0.030
0.314
0.070
0.469
-0.021
0.330
-0.019
0.679
0.064
0.013
0.065
0.233
-0.030
0.227
-0.037
0.582
-0.020
0.089
0.031
0.000
-0.027
0.287
-0.077
0.247
-0.030
0.441
-0.343
0.007
-0.020
0.460
0.038
0.408
0.023
0.009
0.083
0.010
-0.047
0.100
-0.052
0.517
0.002
0.005
0.002
0.347
Member of a farm group that gives credit to 0.058
members
-0.033
Household size
Age of respondent (in years)
Male (1/0)
Years of education
Married (1/0)
Land area ( in hectares)
Uses improved seed (1/0)
Rents land (1/0)
someone In household owned a motorcycle (1/0)
Total Livestock Units
Generalized residual
Generalized residual*bags of subsidized fertilizer
LGA Dummies included
Number of observations
YES
925
24
YES
880
Source: Generated by author using STATA Note: ∗, ∗∗, ∗∗∗ indicates that the corresponding coefficients
are significant at the 10%,5%,and 1% levels, p-values obtained via bootstrapping at 250 repetitions in
hurdle 1 to account for first-stage estimation;
Next, we focus on the extent of participation in the private market, once the decision to
participate has been made. The double hurdle model results provide evidence that access to
subsidies is positively correlated with the extent of farmer participation in the private sector.
Once the decision to participate in the private fertilizer market has been made, the quantity of
subsidized fertilizer received is significantly positively correlated with the quantity of fertilizer
purchased in the private market.
Though the DH model allows the process determining
participation in the private market to be different from that determining the extent of
participation, we recognize that a group member could go to the agro dealer once or several
times. If they go once, then the two decisions are likely made at the same time. We do not have
any information on how exactly subsidized fertilizer and or commercial fertilizer was purchased.
From the DH model, the average partial effect results imply that for every 50Kg of subsidized
fertilizer received an additional 13kg was added to the average quantity of fertilizer purchased in
the private market. They appear to indicate that the importance of fertilizer use is not unknown to
farmers in Kano. Rather, farmers tend not to have access to the product. One main benefit of the
voucher program was ensuring that fertilizer was available at the various local governments of
the states. This likely reduced the cost associated with fertilizer procurement encouraging farmer
purchases. Private dealers were also encouraged to expand their market in the rural areas since
the program had subsidized their cost of distribution and was providing them a guaranteed
market. These results are consistent with Xu et al (2009) who find evidence of crowding in
25
where private markets are weak. Though participation in the private fertilizer market in Kano
was generally high in 2009, anecdotal evidence indicates that this was not the norm and not the
case in previous years. Furthermore, additional information from the implementing agency
(IFDC) indicates that input suppliers in Kano sold quantities larger than the amount sold through
the voucher program. The program indirectly subsidized the transactions cost associated with
fertilizer distribution and studies have shown transportation costs to account for about 25 percent
of fertilizer costs in West Africa (Bumb, Johnson and Fuentes 2012).
Other factors that affect the extent of participation in the private market are distance, ,
landownership status, livestock ownership and land area which is not unusual. Apart from facing
higher transactions cost for input access, respondents located far away from the major trading
markets are also likely to face lower output costs (farm gate) and higher transactions cost to
market their output. Wealth and larger landholdings indicate potentially larger quantity of
fertilizer needed. As earlier mentioned, renting rather than owning land also likely indicates a
wealth effect which is expected to be negatively associated with the quantity of input farmers can
purchase. . Where significant, the interaction term between the generalized residual and the
quantity of subsidized fertilizer reveals that it is endogenous and simultaneously corrects for it in
the second stage estimation model. Though not consistently statistically significant, we find a
positive effect of fertilizer price on the extent of participation in the second hurdle. While this
might be capturing quality differentials as fertilizer quality has been cited as an important
problem in the Nigerian fertilizer market (Banful et al, 2010), it could also be reflecting the
positive pressure of this higher demand on fertilizer prices.
These results for Kano appear to reveal that it is possible for a fertilizer subsidy program
to increase farmer participation in the private sector. We do not find evidence of negative effects
26
of subsidized fertilizer on private market participation like Ricker-Gilbert et al (2011) or Mason
and Jayne (2013). The conclusions based on comparison of the different results should be
tempered by recognition that the studies used different methods and so results are not completely
comparable. While Ricker-Gilbert et al (2011) and Mason and Jayne (2013) use panel data ,
applied to the entire country, this study is focused on only one State; Kano and uses a cross
sectional dataset.
There are several possible explanations for the positive correlation between subsidized fertilizer
and fertilizer purchased from the private market. One could be the context of the study; a state in
Northern Nigeria where the use of fertilizer has long been an established practice. Farmers in
Kano are already likely to be aware of the importance of fertilizer but probably face challenges
getting access to the product. Thus rather than encouraging the adoption of fertilizer, the
voucher program seems to have made it possible for an already existing but unmet demand for
fertilizer to be at least partially addressed. Other factors likely to be important are the size of the
subsidy and associated transactions cost. If the subsidy is too small, then it might not be enough
to make it worth the fixed costs of participating. Thus our findings might indicate that the
voucher distribution in this study made sure lots of people got access to enough subsidized
fertilizer to stimulate further demand from private markets to augment the amount received 12.
Another possible explanation for the positive effects in this study might be the close
oversight to the subsidy implementation process provided by the 2009 voucher program. Since it
was a pilot project meant to demonstrate smart subsidies (and quite limited in scope compared to
12
According to the 2010 LSMS data, the average quantity of fertilizer used per hectare in Kano was bout 165Kg,
which amounts to slightly above 3 bags of fertilizer per hectare. As households tended to cultivate between 3.5
and 4 hectares, their fertilizer demand could be estimated to be between 10 and 15 bags of fertilizer. The voucher
program provided farmers access to 3 bags of fertilizer at the subsidized price which does not sufficiently cover the
fertilizer needs of farmers.
27
the entire country), IFDC monitored the process very closely to minimize the diversion of the
project from beneficiaries and to ensure that the program design was adhered to as much as
possible. It is also possible that by channeling public fertilizer through the private sector, the
voucher program naturally connects farmers to fertilizer dealers and encourages private dealers
to invest in the private fertilizer market. Furthermore, through providing input suppliers a
guaranteed market in rural areas, input suppliers have an incentive to establish sales depots in
rural, more distant locations than they might ordinarily do (given demand uncertainty and costs
of setting up shop) . Such an approach to developing private input markets has potential
sustainability advantages. As input suppliers become aware of the extent of demand in various
rural locations, they will naturally satisfy those markets even in the absence of the voucher
program. However, the success of this will most likely depend on the transactions costs
associated with the distribution of inputs to rural farmers in the absence of the program 13. Where
poor infrastructure persists and the voucher program or other support to cushion associated costs
for input suppliers are absent, the high transactions costs associated with providing fertilizer to
farmers in distant and less accessible locations could still pose challenges to fertilizer access for
some farmers.
As mentioned earlier, other factors besides the largely supply side story above could
explain these positive correlations between subsidized fertilizer and private market participation.
Lower transactions costs of procurement, farmer’s positive experience with the subsidized
product or higher income from better yields could also explain these positive correlations.
13
Another important factor that relates to a potential crowding out effect of the fertilizer subsidy program in Nigeria
is the fact that only 3 fertilizer supply companies participated in the 2009 voucher program. Excluding the
participation of other private dealers has the potential to crowd them out of the private market. Though not the focus
of this study, this is an important issue in any study on crowding out and other distortionary effects of subsidy
programs on private input markets.
28
FURTHER CONSIDERATIONS
As mentioned earlier, alternative sources of subsidized fertilizer existed in Kano in 2009. To
confirm that the voucher program played a significant role in the positive correlations found
between subsidized fertilizer and private market fertilizer purchases, we explore the extent of
participation in the private market by excluding all respondents who did not participate in the
2009 voucher program but received subsidized fertilizer 14. Table 5 shows that the positive effect
of subsidized fertilizer receipt on private fertilizer demand is maintained. Relevant variables
from the analysis with the entire sample also maintain.
We also run the estimations by
including a dummy variable to account for all those who did not participate in the 2009 voucher
program but received subsidized fertilizer. We include this variable and its interaction with the
quantity of subsidized fertilizer. These estimation results are included in the appendix and reveal
that the positive correlations between subsidized fertilizer and private market participation on
average still maintain. If significant, one could expect that participating in this alternative
subsidy program might have a negative effect on private market purchases since this program ran
parallel to the private sector but did not involve the private dealers. This appears to be revealed
in the data. Farmers, who received subsidized fertilizer through alternative means than the
voucher program being studied here, were significantly more likely to participate in the private
market on average. However, every additional bag of subsidized fertilizer received for this group
reduced their likelihood of participating in the private market as well as the quantity of fertilizer
purchased (Table A2). These results supports the findings of Takeshima et al (2013) which finds
that participants in the traditional government fertilizer program in Nigeria tended to be those
14
By removing those who participated in the government program, this estimation enables us focus only on those
who participated in the 2009 voucher program. Anecdotal evidence indicates that the other subsidy program was a
very small program by local governments
29
already likely to participate in the private market and receiving subsidized fertilizer reduced
their participation in the private market.
Though we recognize that participation in the alternate program is likely endogenous, including
these respondents in our estimations should actually have yielded lower effects of subsidized
fertilizer, making our estimates a lower bound. The fact that our results maintain (with a larger
coefficient) whether they are included or not, indicates that they are not driving the positive
correlations observed between quantity of subsidized fertilizer received through the voucher
program and quantity of fertilizer purchased from the private market.
Table 5. Second stage estimation results of private fertilizer market participation excluding
respondents who did not participate in the 2009 voucher program but received subsidized
fertilizer
Hurdle 1
Coefficients are Average partial effects (APE)
Hurdle 2
Quantity of fertilizer
participation
in
purchased
from
private market
market
Coefficient p value Coefficient p value
Number of subsidized bags of fertilizer received
0.025
0.000
0.482
0.000
0.267
0.401
0.000
0.000
-0.036
0.004
0.236
0.000
-0.099
0.572
0.103
0.026
-0.072
0.364
-0.063
0.000
0.041
0.325
0.068
0.044
0.118
0.189
-0.030
0.231
-0.060
0.255
0.083
0.000
0.109
0.011
Farmer member of a group that purchased fertilizer 0.032
together
-0.064
Price of fertilizer
Distance to main market
Price of Maize (N/kg)
Member of a farm group that gives credit to members
Household size
Age of respondent (in years)
Male (1/0)
Years of education
30
Married (1/0)
Land area ( in hectares)
Uses improved seed (1/0)
Rents land (1/0)
someone In household owned a motorcycle (1/0)
Total Livestock Units
Generalized residual
Generalized residual*bags of subsidized fertilizer
LGA Dummies included
Number of observations
-0.085
0.004
-0.061
0.534
-0.003
-0.090
0.032
0.000
-0.016
0.538
-0.066
0.397
-0.082
0.032
-0.365
0.040
-0.005
0.887
0.035
0.518
0.024
0.020
0.096
0.000
-0.333
0.000
-0.408
0.001
0.010
0.000
0.011
0.004
YES
702
YES
Source: Generated by author using STATA Note: ∗, ∗∗, ∗∗∗ indicates that the corresponding coefficients are
significant at the 10%,5%,and 1% levels, p-values obtained via bootstrapping at 250 repetitions in hurdle 1 to
account for first-stage estimation;
To account for diminishing returns to fertilizer application since multiple members of a
household could have received subsidized fertilizer; the double hurdle model is run including a
dummy indicating if the respondent belonged to a family with multiple recipients of subsidized
fertilizer. Though APEs are lower (from 0.26 to about 0.19) , the positive correlations maintain
(shown in table 6).
31
Table 6. Second stage estimation results of private fertilizer market participation
accounting for multiple participants per household
Number of subsidized bags of fertilizer received
Hurdle 1
Hurdle 2
Participating in the
market
Coefficient p value
0.047
0.102
Quantity of fertilizer
purchased
Coefficient p value
0.194
0.025
Farmer member of a group that purchased fertilizer 0.062
together
-0.059
Price of fertilizer
0.001
0.394
0.000
0.017
-0.030
0.181
0.244
0.000
-0.087
0.159
-0.855
0.248
-6.768
0.132
0.085
0.004
-0.090
0.150
-0.034
0.080
0.060
0.210
0.030
0.341
0.069
0.357
-0.023
0.360
-0.020
0.702
0.065
0.006
0.066
0.241
-0.025
0.323
-0.028
0.757
-0.003
0.027
0.031
0.000
-0.030
0.230
-0.081
0.139
-0.024
0.515
-0.329
0.014
-0.020
0.375
0.037
0.408
0.021
0.001
0.079
0.010
-0.006
0.855
0.001
0.232
0.008
0.450
0.005
0.702
Household has multiple members participating in the 0.064
voucher program (1/0)
LGA Dummies included
YES
Number of observations
Pseudo R2 or adjusted correlation coefficient
0.149
0.046
0.095
0.111
Distance to main market
Price of Maize (N/kg)
Member of a farm group that gives credit to members
Household size
Age of respondent (in years)
Male (1/0)
Years of education
Married (1/0)b
Land area ( in hectares)
Uses improved seed (1/0)
Rents land (1/0)
someone In household owned a motorcycle (1/0)
Total Livestock Units
Generalized residual
Generalized residual*bags of subsidized fertilizer
YES
Source: Generated by author using STATA Note: ∗, ∗∗, ∗∗∗ indicates that the corresponding coefficients are significant at the
10%,5%,and 1% levels, p-values obtained via bootstrapping at 250 repetitions in hurdle 1 to account for first-stage estimation;
32
It can be argued that our instrument; “one’s relation to the leadership of their farm group” could
be correlated, in unseen ways, with their integration in markets and liquidity access. Though all
estimations already capture differential liquidity access through the credit variable, further
estimations of all the models were rerun including controls for a farmer’s level of integration into
the market. The study uses the number of farm groups a farmer belongs to, the number of trade
associations he belongs to and a dummy variable to identify respondents who hold leadership
positions in the community. While trade associations are directly related to trading activities, a
farmer engaging in multiple farm groups usually occurs where group activities are specific to a
crop or activity (say marketing output). Thus these two variables give some indication of
potentially different levels of market integration. The positive correlations more in line with a
crowding in argument remain and are included in Table A3 in the appendix 15. Furthermore, a
falsification test was conducted using the other program state Taraba (where vouchers were
given to individual farmers rather than the farm group leadership) to confirm that that there
wasn't some unobservable variable correlated with relationship to the leadership that also causes
market purchases of fertilizer. We test for whether being related to farm group leadership is: (a.)
correlated with the number of subsidized bags received (b.) correlated with the number of bags
purchased from the private market. Our hypothesis is that the null hypothesis of correlation in (a)
and (b) should be rejected in Taraba, if there are no other unobservables driving the number of
bags received, also driving market purchases. We reject both (a) and (b) thus increasing our
confidence in our instrumental variable. This test and its results are also included in the
appendix.
15
We recognize that farmers who were friends of the farm group leadership could potentially go with the leaders
to pick up the subsidized fertilizer and buy more commercial fertilizer while they are there for convenience. To
avoid any omitted variable problems, we include a dummy for whether a respondent was a friend of the farm
group leader. The coefficient on subsidized bags of fertilizer remains strongly positively correlated with commercial
market purchases. For space consideration, these results are available upon request.
33
CONCLUSION
With the resumed emphasis on input subsidies, this article contributes to the limited empirical
evidence on the unsolved question about the impact of input voucher programs on private
markets in SSA. We estimate the effect of a fertilizer voucher program in Nigeria on farmer
participation in the private fertilizer market. The article used a double hurdle model (to account
for the corner solution nature of fertilizer demand) and a control function approach to address the
endogeneity of the quantity of subsidized fertilizer received by farmers. The article finds
evidence that receiving subsidized fertilizer in Nigeria increased both the probability and extent
of participation in the private fertilizer market. Once the decision to participate had been made,
every bag of subsidized fertilizer increased the quantity of fertilizer purchased from the private
market by 0.26 of a bag which is about 13kg.
The results of this study indicate that one benefit of the 2009 voucher program was to develop
links between rural farmers and input suppliers. These results indicate that where fertilizer
subsidy programs involve the private sector and ensure that many farmers (including the poor)
receive a significant amount of fertilizer to warrant participation but not enough to satisfy their
demand, there is less likely to be distortionary effects of fertilizer subsidies. Also, if access to
subsidized fertilizer increased yields and incomes, this could also explain the positive effect of
subsidized fertilizer on private market fertilizer demand. The study’s results also indicate that
where private fertilizer market links are weak, there could be significant gains from the
temporary use of voucher programs to strengthen links between input suppliers and farmers. The
design of these programs is very important. For example, since the pilot program cushioned
some of the transactions costs associated with reaching some remote locations, it is important
34
that Governments fulfill their role to provide the necessary infrastructure to minimize these costs.
This will help ensure that the private sector can still profitably meet the demand of rural farmers
after the exit of a government or development organization led program. However, designing
input subsidy programs subsidizes input suppliers in a way that encourages the development of
private markets rather than purely subsidizing farmers is a potentially useful way to develop
sustainable private markets and make such input programs more effective. Furthermore, as such
subsidy programs are expanded; increased attention has to be paid to the targeting mechanism to
ensure that elite capture of the process is minimized as the program becomes better understood in
the rural areas.
Going forward, it is important to gather additional information from other actors in the fertilizer
supply chain to better understand the effect of these programs on private market development.
One approach could be an analysis of the agro-input dealers, as was done by Krausova and
Banful (2010) for Ghana.
REFERENCES
1. Banful, A. 2010. Old Problems in new solutions? Politically Motivated Allocation of
Program
Benefits
and
the
“New”
Fertilizer
Subsidies.
Discussion
Paper
1002. Washington, D.C. International Food Policy Research Institute
2. Banful, A. B., and O. Olayide. 2010. Perspectives of varied stakeholders in Nigeria on
the federal and state fertilizer subsidy programs. Nigeria Strategy Support
Program Working paper. Washington, D.C. International Food Policy Research
Institute
35
3. Bellemare, M., and C. Barrett. 2006. An Ordered Tobit Model of Market Participation:
Evidence from Kenya and Ethiopia. American Journal of Agricultural Economics
88(2): 324-337.
4. Bumb, B.L., M. Johnson and P. Fuentes. 2012. Improving Regional Fertilizer Markets
in West Africa. Policy Brief 20. Washington, D.C. International Food Policy Research
Institute.
5. Cameron, A.C. and Trivedi, P.K.2005. Microeconometrics, Methods and Applications.
New York: Cambridge University Press.
6. Cameron, A.C. and Trivedi, P.K. 2009. Microeconometrics Using STATA. Texas:
Stata Press.
7. Coady, D.P. 1995. An Empirical Analysis of Fertilizer Use in Pakistan.
Econometrica 62(246): 213-224.
8. Crawford E.W., T. J. Jayne and V. Kelly. 2006.
Alternative Approaches for
Promoting Fertilizer Use in Africa. World Bank Agriculture and Rural Development
Discussion Paper 22.
9. Croppenstedt, A., M. Demeke and M. Meschi. 2003. Technology adoption in the
presence of constraints: The case of Fertilizer demand in Ethiopia. Review of
Development Economics, 7: 58–70.
10. de Janvry, A., M. Fafchamps, and E. Sadoulet. 1991. Peasant Household Behaviour
with Missing Markets: Some paradoxes Explained. Economic Journal 101(409):
1400-1417.
36
11. de Janvry, A., and E. Sadoulet. 2006. “Progress in Modeling of Rural Household’s
Behavior Under Market Failures. In Poverty, Inequality and Development Vol. 2 ed
A. de Janvry and R. Kanbur, 155-181. New York: Springer
12. Dorward A. 2009. Rethinking agricultural input subsidy programmes in a changing
world. Paper presented for the Trade and Markets Division. Food and Agriculture
Organization of the United Nations.
13. Dorward, A., E. Chirwa, V. Kelly, T. S. Jayne,R. Slater, and D. Boughton. 2008.
Evaluation of the 2006/07 Agricultural Input SubsidyProgramme, Malawi. Final
Report.Lilongwe, Malawi
14. Ellis, Frank. 1992. Agricultural Policies in Developing Countries. Cambridge, UK:
Cambridge University Press
15. Fadama III Baseline survey .2010. Dataset collected by the Fadama Development
Project, a World Bank funded project across Nigeria’s 36 states.
16. Garen, J., 1984. The returns to schooling: A selectivity bias approach with a
continuous choice variable. Econometrica. 52: 1199–1218.
17. Gregory, I. 2006. The Role of Input Vouchers in Pro-Poor Growth. Selected Sections
from a Background Paper Prepared for the African Fertilizer Summit June 9-13,
2006 Abuja, Nigeria.
37
18. Heckman, J., R. LaLonde and J. Smith. 1999. The Economics and Econometrics of
Active Labor Market Programs," in Handbook of Labor Economics .Vol.III, .pp.
1865-2097.
19. Holloway, G., C. Barrett, and S. Ehui. 2005. The Double-Hurdle Model in the
Presence
of Fixed Costs. Journal of International Agricultural Trade and
Development 1:17-28.
20. Holloway, G., S. Ehui, and A. Teklu. 2008. Bayes Estimates of Distance-to-Market:
Transactions Costs, Cooperatives and Milk-Market Development in the Ethiopian
Highlands. Journal of Applied Econometrics 23:683-696.
21. International Fertilizer Development Center (IFDC). 2010. Project Implementation
Manual (PIM) For 2009 Fertilizer Voucher Programme in Kano and Taraba
States.
22. International Fertilizer Development Center (IFDC). Fertilizer Reaching Farmers’
Hands.
http://www/ofdc/prg/changing_lives_case_studies/fertilizers_
reaching_farmers_hands,_Nigeria_Ferti (accessed April 1, 2012)
23. International Fertilizer Development Center (IFDC). 2009. Update of fertilizer supply
and demand –Sub Saharan Africa. Presentation at the IFA Africa Forum, Cairo –
Egypt, February 2009.
24. Imbens, G and J.Wooldridge. 2007. What’s New in Econometrics? Lecture notes
from Summer Institute 2007. http://www.nber.org/minicourse3.html ( accessed
January 5, 2012).
38
25. Key, N., E. Sadoulet, and A. de Janvry. 2000. Transactions Costs and Agricultural
Household Supply Response. American Journal of Agricultural Economics 82:24559.
26. Krausova, M. & Banful, A. (2010). Overview of the agricultural input sector in Ghana.
Discussion Paper 01024. Washington, DC: International Food Policy Research
Institute.
27. Liverpool-Tasie, L. S. O. (2013), Do vouchers improve government fertilizer distribution?
Evidence from Nigeria. Agricultural Economics. doi: 10.1111/agec.12094
28. Liverpool-Tasie, L. S. O . 2012. Farmer Groups, Input Access and Intragroup
Dynamics: A Case Study of Targeted Subsidies in Nigeria. Discussion Paper 01197.
Washington, DC: International Food Policy Research Institute
29. Liverpool, L.S.O. and A. Winter-Nelson. 2010. Poverty status and the impact of
formal credit on technology use and wellbeing among Ethiopian smallholders. World
Development. 38(4):541–554.
30. Liverpool-Tasie, L. S. O., A. B. Banful, and B. Olaniyan. 2010. An assessment of the
2009 fertilizer voucher program in Kano and Taraba, Nigeria. Nigeria Strategy
Support
Program Working paper 0017. Washington, D.C. International Food
Policy Research.
31. Longley, C., R. Kachule, M. Madola, I. Maposse, B. Araujo, T. Kalinda, and H.
Sikwibele. 2003. Agricultural Input vouchers: Synthesis of Research Findings from
Malawi. Mozambique and Zambia, FANRPAN.
39
32. Mason, N. M., Jayne, T. S.2013. Fertilizer subsidies and smallholder commercial
fertilizer purchases: Crowding out, leakage, and policy implications for Zambia. J.
Agric. Econ., in press
33. Minot N., and T. Benson. 2009. Fertilizer subsidies in Africa: Are vouchers the
answer? Policy Brief 60. Washington, D.C. International Food Policy Research
Institute.
34. Morris, M., V. Kelly, R. Kopicki, and D. Byerlee. 2007. Fertilizer use in African
agriculture: Lessons learned and good practice guidelines. Washington, D.C.
World Bank.
35. Nagy, J. G., and O. Edun. 2002. Assessment of Nigerian government of policy and
suggested alternative market friendly policies. Confidential draft, September 26,
2002.
International
Fertilizer
Development
Center.
http://www.usaid.gov/ng/downloads/reforms/assessmentoffertilizerpolicy.pdf.
(accessed May 10, 2010).
36. Nigeria, National Bureau of Statistics (NBS). 2009. Annual Abstract of Statistics,
2009. Abuja.
37. National Population Commission, Nigeria (NPC).2006. National Population Census.
http://www.population.gov.ng/index.php?option=com_content&view=article&id=89
( accessed April 1, 2012).
40
38. Odhiambo, J. J., and V. N. Magandini. 2008.An Assessment of the Use of Mineral and
Organic Fertilizer by Smallholder Farmers in Vhembe District, Limpopo Province,
South Africa. African Journal of Agricultural Research 3(5): 357–362.
39. Ricker-Gilbert, J., T. Jayne, and E. Chirwa. 2011. Subsidies and Crowding Out: A
Double-Hurdle Model of Fertilizer Demand in Malawi. American Journal of
Agricultural Economics. 93(1): 26-42.
40. Rivers, D., and Q.H. Vuong. 1988. Limited Information Estimators and Exogeneity
Tests for Simultaneous Probit Models. Journal of Econometrics 39: 347-366.
41. Roy, A. 1951. Some Thoughts on the Distribution of Earnings, Oxford Economic
Papers, 3, 135-145.
42. Sachs, Jeffrey D. 2004. “The Case for Fertilizer Subsidies for Subsistence Farmers.”
Working Paper, the Earth Institute at Columbia University.
43. Sadoulet, E., & de Janvry, A. 1995. Quantitative development policy analysis.
Baltimore: The Johns Hopkins University Press.
44. Singh, I., L. Squire, and J. Strauss. 1986. Agricultural Household Models. Baltimore:
Johns Hopkins University Press.
45. Smith, R.J., and R.W. Blundell. 1986. An Exogeneity Test for a simultaneous
Equation Tobit Model with and Application to the Labor Supply. Econometrica
50(3):679-685.
41
46. Takeshima, H., E. Nkoya and S. Deb (2013). Impact of Fertilizer subsidies on the
Commercial Fertilizer Sector in Nigeria. Nigeria Strategy Support Prgram II, IFPRI
working paper No. 23.
47. Vella, F. 1993. A Simple Estimator for Simultaneous Models with Censored
Endogenous Regressors. International Economic Review 34(2): 441-457.
48. Winter-Nelson, A., and A. Temu. 2005. Impacts of Prices and Transactions Costs on
Input Usage in a Liberalizing Economy: Evidence from Tanzania, Agricultural
Economics 33:243-253.
49. World Bank. 2010. Data retrieved May 12, 2012, from World Development
Indicators Online (WDI) database.
50. World Bank. 2008. Data retrieved March 20, 2012, from World Development
Indicators Online (WDI) database.
51. Wooldridge J, M. 2008. Quasi-Maximum Likelihood Estimation and Testing for
Nonlinear Models with Endogenous Explanatory Variables. Working paper. Dept. of
Econ., Michigan State University, East Lansing
52. Xu, Z., Burke, T. Jayne, and J. Govereh. 2009. Do input subsidy programs
“crowd in” or “crowd out” commercial market development? Modeling fertilizer
demand in a two-channel marketing system. Agricultural Economics, 40: 79–94.
42
APPENDIX
A.1.Sample selection
The domain for this analysis was smallholders in Kano and Taraba States; the subpopulations
for which we want survey estimates of the outcome of participation in the voucher program. We
randomly selected 10 local government areas each in both states. To ensure a level of
generalization was possible from our survey, we confirmed that the 10 LGAs selected
represented potential LGA variation such as proximity to state capitals (Kano city and Jalingo),
population and accessibility in terms of road availability and quality.
Our measurement units are the households and household members surveyed in both states. The
key variables of interest that were used to determine the minimum sample size necessary for our
analysis are quantity of subsidized fertilizer used as well as price of fertilizer purchased. We used
the formula given in the sampling guide provided by the Food and Nutrition Technical
Assistance (FANTA) for calculating the minimum necessary sample size. Our calculations were
done to ensure with 95% confidence that estimated differences between program participants and
non-participants (or participants over time) are not purely by chance and to have 80% confidence
that an actual change or difference will be detected (power of the test) (Magnani 1997).
Data on fertilizer consumption by states was not readily available. Thus our minimum sample
size requirements were estimated using approximations from available data as follows: For
quantity of fertilizer used, Banful and Olayide (2010) reveal that the average quantity of fertilizer
that farmers in Kano and Taraba states would have if subsidized fertilizer were equally
distributed across households would be 97kg and 117kg respectively. However, Nagy and Edun
(2002) estimate that only about 30 percent of subsidized fertilizer reaches small farmers at the
43
subsidized price. Thus we can estimate that farmers in Kano and Taraba on average receive
about 29.1kg and 35.1kg each of subsidized fertilizer through the traditional distribution
mechanism. The goal of the voucher program was to increase the quantity of subsidized
fertilizers farmers received through the use of vouchers rather than the previous government
controlled distribution mechanism. Participating farmers in Kano and Taraba should have
received 3 bags (150kg) and 4 bags (200kg) respectively. Using these figures, we can estimate
that the sample size needed to identify the changes due to the program required samples of
between 30 and 35 households on the quantity of subsidized fertilizer used in each state using the
following FANTA formula:
n=D[(Z_α+Z_β )^2*(〖sd〗_1^2+〖sd〗_2^2 )/(X_2-X_1 )^2 ]
Where:
n = required minimum sample size per survey round or comparison group
D = design effect for cluster surveys indicating the factor by which the sample size for a
cluster sample would have to be increased in order to produce survey estimates with the same
precision as a simple random sample (We use the default value of 2 as suggested by Magnani
1997.
X1 = the estimated level of fertilizer a household has access to prior to the program
X2 = the expected level of subsidized fertilizers households have access to after participation
sd1 and sd2 = expected standard deviations for the indicators for the comparison groups being
compared
44
Zα = the z-score corresponding to the degree of confidence with which it is desired to be able to
conclude that an observed change of size (X2 - X1) would not have occurred by chance
(statistical significance), and
Zβ = the z-score corresponding to the degree of confidence with which it is desired to be certain
of detecting a change of size (X2 - X1) if one actually occurred (statistical power).
NOTE: For the standard deviation, we used estimates on the ratio of mean to standard deviation
of fertilizer use from a subsample of largely cereal producing households in another northern
state, Kaduna in 2008 (IFPRI, 2008). The mean to standard deviation ratio was 1.07. This ratio
was applied to our mean quantity of subsidized fertilizer before and after the voucher program to
get the associated standard deviations. Even if there was no diversion of subsidized fertilizer in
both states, applying the same formula indicates that we need between 170 and 250 respondents
in Taraba and Kano respectively.
For further confirmation, the minimum sample calculation was also conducted using secondary
data from other studies. A 2007 study cites 41kg/ha as the average fertilizer use for Kano state
(Maiangwa et al 2007). With average land size in Kano of about 1.9 ha (KNARDA monitoring
and evaluation department, 2010), this amounts to about 78kg per household. Using the same
standard deviation as above, we estimated the new minimum size necessary to satisfactorily
capture a change in quantity of fertilizer used from 78 kg per household to about 150 kg (the
three subsidized bags to be available through the program). It is estimated that a sample size of
118 is necessary.
For price of fertilizer, we used the August 2009 price of Urea (as that was the date at which
about 80percent and 90percent of the vouchers had been distributed in Taraba and Kano
45
respectively. The price of Urea at Dawanau market in Kano was about N3200 per 50 kg bag
(N64/kg) . The vouchers were individually worth a total value of N 2,000/50 kg bag. Thus, the
benefit of receiving the voucher should translate to a N2000 difference in the price of Urea.
Using this in the above formula to calculate the minimum sample size with standard deviation
calculated again using the ratio of the mean to standard deviation of prices paid by farmers in
Kaduna, we estimate that the minimum sample size would be about 80 households in Kano.
Recognizing that farmers in more remote rural areas are likely to pay higher prices for their
fertilizer, we simulated the price estimates and find that even if Urea prices were 50percent
higher in the rural areas (N4500 per bag), the minimum sample size would be about 210 .
46
Table A1. Distribution of sample households across the 10 Local government areas in Kano
Kano
Local
Number
Government
of
Local
Number
households
Government
households
Area
surveyed
Area
surveyed
Bagwai
64
Ungoggo
91
Takai
64
Gezawa
83
Dambatta
60
Gabasawa
60
Dala
82
Rano
49
Karaye
37
Kura
50
Total
640
of
Source: Generated by author from the fertilizer voucher program evaluation survey
47
Table A2. Second stage estimation results of private fertilizer market participation including a
dummy for respondents who did not participate in the 2009 voucher program but received
subsidized fertilizer and an interaction between the dummy and subsidized fertilizer
Hurdle 1+
Hurdle 2
Quantity of fertilizer
participation in private
purchased
from
market
market
Coefficient p value
Coefficient p value
0.069
0.090
0.382
0.000
Farmer member of a group that purchased fertilizer 0.037
0.552
0.261
0.021
Number of subsidized bags of fertilizer received
together
Price of fertilizer
-0.080
0.259
0.248
0.038
Distance to main market
0.198
0.000
-0.174
0.050
Member of a farm group that gives credit to members
0.075
0.014
-0.131
0.136
Household size
-0.041
0.037
0.116
0.024
Age of respondent (in years)
-0.031
0.523
-0.019
0.830
Male (1/0)
0.011
0.737
-0.001
0.899
Years of education
0.056
0.036
0.082
0.105
Married (1/0)
-0.004
0.919
-0.011
0.915
Land area ( in hectares)
0.001
0.663
0.042
0.000
Uses improved seed (1/0)
-0.039
0.149
-0.020
0.766
Rents land (1/0)
-0.040
0.345
-0.442
0.015
Price of Maize (N/kg)
48
someone In household owned a motorcycle (1/0)
-0.019
0.564
0.007
0.926
Total Livestock Units
0.021
0.002
0.056
0.021
not a participant in voucher program but received 0.576
0.022
0.721
0.890
0.000
-0.338
0.080
subsidized fertilizer
not a participant in voucher program but received -0.248
subsidized fertilizer*subsidized fertilizer received
Generalized residual
-0.068
0.040
-0.160
0.069
Generalized residual*bags of subsidized fertilizer
0.003
0.140
0.004
0.225
LGA Dummies included
Yes
YES
Number of observations
925
761
Source: Generated by author using STATA Note: ∗, ∗∗, ∗∗∗ indicates that the corresponding coefficients are
significant at the 10%,5%,and 1% levels, p-values obtained via bootstrapping at 250 repetitions in hurdle 1 to
account for first-stage estimation;
49
Table A3. Second stage estimation results of private fertilizer market participation including
additional controls to capture market integration
Hurdle 1+
Hurdle 2
Quantity of fertilizer
participation
in
private
purchased
from
market
market
Coefficient
p value
Coefficient p value
0.0790***
0.001
0.260***
0.010
together
0.029
0.274
0.218*
0.085
Price of fertilizer
-0.040
0.198
0.254**
0.031
Distance to main market
-0.115
0.104
-0.313**
0.045
Price of Maize (N/kg)
-0.871
0.243
-6.102
0.140
Member of a farm group that gives credit to members
0.058*
0.087
-0.103
0.257
Household size
-0.032**
0.048
0.121**
0.016
Age of respondent (in years)
-0.033
0.195
-0.026
0.801
Male (1/0)
0.005
0.762
0.003
0.961
Years of education
0.027*
0.060
0.070
0.270
Married (1/0)
0.007
0.742
0.018
0.847
Land area ( in hectares)
-0.015
0.111
0.427***
0.000
Uses improved seed (1/0)
-0.032**
0.039
0.088
0.375
Rents land (1/0)
-0.050*
0.082
-0.485***
0.000
someone In household owned a motorcycle (1/0)
0.000
1.000
0.032
0.514
Total Livestock Units
0.012*
0.071
0.047
0.127
Generalized residual
-0.069**
0.005
-0.219**
0.029
Number of subsidized bags of fertilizer received
Farmer member of a group that purchased fertilizer
50
Generalized residual*bags of subsidized fertilizer
0.002**
0.006
0.004
0.116
Number of trade groups
0.039
0.647
0.094
0.635
holds a position in community
0.022
0.814
-0.001
0.996
Number of farm groups
-0.021
0.339
0.042
0.720
LGA Dummies included
YES
YES
Number of observations
925
761
Source: Generated by author using STATA Note: ∗, ∗∗, ∗∗∗ indicates that the corresponding coefficients are
significant at the 10%,5%,and 1% levels, p-values obtained via bootstrapping at 250 repetitions in hurdle 1 to
account for first-stage estimation;
51
Falsification Test for instrument “related to farm group president”
This falsification test was conducted to further check that our results are not driven by
unobservable variables correlated with relationship to the leadership that also caused farmers to
be more likely to purchase from the private market. We use the other program state Taraba
where vouchers were given to individual farmers rather than the farm group leadership to
confirm this
We test for whether being related to farm group leadership is:(a.) correlated with the number of
subsidized bags received (b.) correlated with the number of market bags purchased. Table 3
demonstrates (a) to be true in Kano and (b) is implied by the control function estimator results in
Kano. However in Taraba, being related to the farm group leadership doesn't affect subsidized
fertilizer receipt there, so any correlation between market purchases and relationship with the
leadership would be due to other variables. Thus our hypothesis is that the null hypothesis of
correlation in (a) and (b) should be rejected in Taraba if there are no other unobservable factors
driving the number of bags received also driving market purchases.
Table A4 demonstrates that being related to the farm group leadership does not significantly
affect the quantity of subsidized fertilizer received by respondents in Taraba as one would expect
given the program set up. This is consistent with Liverpool-Tasie (2012) which (using the same
data as this study) demonstrates that where fertilizer was given to a group representative for
further distribution to members, respondents with close links to their farm group leaders received
more bags of fertilizer than did those without. When fertilizer was given directly to farmers (in
Taraba state), such results did not occur. Table A5 demonstrates that being related to the farm
group leader is not a significant factor in determining the number of bags of fertilizer purchased
52
in the private sector. These results maintain whether additional controls are included or not. In
A5, location dummies are included in the estimation.
Consequently, we reject the null that being related to farm group leadership is:(a.) correlated
with the number of subsidized bags received (b.) correlated with the number of market bags
purchased in Taraba confirming that it is not likely that being related
to the farm group
leadership is correlated with other unobserved factors likely to affect farmers purchase from the
private market.
53
Table A4. Determinants of quantity of subsidized fertilizer received in Taraba State
Bags of subsidized fertilizer
Coefficient t statistics
P>t
Participated in the voucher program
8.042
4.010
0.000
together
2.880
2.130
0.034
Holds a leadership position in the village
1.117
0.600
0.548
Age of respondent (in years)
0.025
0.620
0.534
Male (1/0)
1.515
1.390
0.165
Married (1/0)
0.166
0.260
0.796
Years of education
-1.099
-0.620
0.537
Land area in hectares
0.008
3.510
0.000
Related to farm group leadership (Instrument)
1.043
1.130
0.259
Someone In household owned a motorcycle (1/0)
-0.606
-0.530
0.593
Total Livestock Units (TLU)
-0.004
-1.000
0.316
Rents land (1/0)
-1.114
-0.590
0.555
Constant
-13.589
-3.180
0.002
N
738
Pseudo R2
0.152
Farmer member of a group that purchased fertilizer
Source: Generated by author using STATA Note: *,** and *** indicate p values significant at 10%, 5%
and 1% respectively.
54
Table A5. Effects of being related to farm group leadership on private fertilizer market
participation in Taraba State
Tobit
Bags of fertilizer
purchased in the
private market
Coefficient
OLS
t
P>t
statistics
Coefficient
t statistics
P>t
Being related to the
farm group president
0.668
0.860
0.391 0.690
1.320
0.186
Constant
3.033
1.930
0.055 4.676
4.060
0.000
Number of
observations
527
527
R squared/pseudo r
squared
0.05
0.16
Source: Generated by authors using STATA. Location dummies are included and models were
run with other controls and location dummies and the insignificant effect of being related to the
farm group president maintain
55
56