Land Management and Technology Adoption in Eastern Uganda

Land Management and Technology Adoption in Eastern Uganda.

A Farm-Based Bio-Economic Modeling Approach.

By: Johannes Woelcke

Contribution to Final GTZ-Project Report “Policies for Improved Land Management in Uganda”

Zentrum für Entwicklungsforschung Center for Development Research

Universität Bonn

2

1. Introduction Under the regimes of Idi Amin (1971-79) and Milton Obote (1980-1985) Uganda’s economy

plunged into a prolonged crisis with negative real GDP growth rates (Baffoe, 2000). In 1987

the Ugandan Government under Musevini introduced an Economy Recovery Program in

cooperation with the IMF (International Monetary Fund) and World Bank aiming at market

liberalization, privatization and decentralization. Although, these reforms have positive

impacts on Ugandan economy (GDP real growth has averaged 6 per cent per annum), the

productivity in the agricultural sector has either stagnated or declined (APSEC, 2000). The

agricultural sector is the mainstay of Uganda’s economy accounting for 43 % of the GDP, 85

% of the value of exports and providing 80 % of employment (FAO, 1999). Land degradation

is assumed to be a major factor contributing to declining agricultural productivity, poverty

and food insecurity. It was estimated that soil nutrient losses were among the highest in sub-

Saharan Africa in the early 80s and average annual nutrient losses were predicted to reach 85

kg/ha of N, P2O5, K2O by the year 2000 (Stoorvogel and Smaling 1990). Recent studies in

eastern and central Uganda have given high negative nutrient balances for most of the

cropping systems (Wortmann and Kaizzi, 1998).

The “critical triangle of development goals” by Vosti and Reardon (1997) implies that it is a

major objective for researchers and politicians to find technologies, institutions, and policies

and to make the three goals of growth, poverty alleviation, and sustainability more

compatible. It is obvious that the three goals are compatible in the long run. Sustaining the

natural resource base will help agricultural productivity growth and this will lead to poverty

alleviation. In the short run there might be trade-offs among the three goals taking into

account the short-term perspective of the individual farmer to satisfy the basic needs of the

household. Farmers need to have the incentive and the capacity for a sustainable

intensification of agriculture. Several factors such as policies, technologies, institutions,

population pressure, and agro-climatic conditions can affect the links between sustainability,

growth and poverty alleviation by influencing the choices of households and communities.

These factors have the potential to increase the compatibility of the three objectives.

Addressing these issues of sustainable intensification of agriculture, the Ugandan Government

has published a “Plan for Modernization of Agriculture” in 2000 as part of the “Poverty

Eradication Action Plan (PMA)” with the vision of “poverty eradication through a profitable,

competitive, sustainable and dynamic agricultural and agro-industrial sector.” The priority

areas for action are: improving access to rural finance, improving access to markets, research

3

and technology development, sustainable natural resource utilization, and new ways of

management and better education in agriculture for farm households.

2. Research Objectives

The proximate causes of land degradation (e.g. very low use of inorganic fertilizers and

limited use of organic inputs, declining fallow periods, deforestation, crop production on

steep slopes with limited investments in terraces or other conservation measures) are

relatively well known, but the core of the land degradation problem is of economic nature.

Poor rural households in developing countries have to cope with a situation where land

productivity and therefore household income are stagnant or declining. Financial constraints

and imperfect market conditions are leading to livelihood strategies that contribute to nutrient

depletion since the majority of farm households does not perceive sustainable intensification

of agriculture as a suitable strategy (Barbier, 1997). The majority of rural households depend

on agricultural production as their main source of income, but the importance of off-farm

incomes increases as the average farm size declines. Consequently, labor, land, and cash

constraints are limiting the ability to invest in land improvements. It is an important and

difficult task to design effective policy strategies, which make environmentally sound

technologies affordable and adoptable for the farmers, including poor farmers. Some studies

(Feder, Just and Zilberman, 1985) have been conducted analysing the determinants, which

influence the adoption of technology (e.g. farm size, tenure, age, education and risk). How

farm households react to alternative policy strategies and how the adoption of a technology

affects the environment and the productivity simultaneously is less clear though.

Consideration of the problem presented above led to the following research objectives:

1. Improve understanding of key economic determinants affecting land management

decisions at the farm household level;

2. Analyze the likely impacts of land use policies aiming at more productive, sustainable,

and poverty-reducing land management in Uganda.

4

With respect to research objective 1 the following hypothesis will be considered:

• Labor shortages, capital constraints, imperfect capital markets, distorted input and

output prices, transaction and information costs are the most binding factors affecting

land use practices and adoption of new technologies.

With respect to research objective 2 the following scenarios will be developed:

• Policy and institutional interventions mentioned as priority areas in the PMA

(development of local credit markets, promotion of improved technologies, labor

exchange institutions etc.), which affect farmers` choices of land management

practices.

• Potential impacts of promoted technologies on household welfare and sustainability

indicator.

3. Integrated Approach to Bio-Economic Modeling

Figure 1 illustrates the concept of the Integrated Approach to Bio-Economic Modeling in

Eastern Uganda. A model can be defined as a representation of an actual phenomenon to

explain real world systems. In order to depict the links between the problem statement, the

real world situation, and the final bio-economic modeling approach, the modeling process is

divided into four stages: the modeled system, the conceptual model, the representational

model and the computational model (Parrott and Kok, 2000). Each model step is addressing

specific objectives, reaching from the overall description of the study area and its comparative

advantages to the assessment of the likely impacts of relevant policy options at the farm

household level. Furthermore, the stepwise approach illustrates the concept of the sampling

procedure, which includes: a listing of the households in the study area, community surveys

(conducted by International Food Policy Research Institute, IFPRI) of both villages under

consideration, and two interconnected comprehensive household/plot level surveys. After

identification of the predominant development domains in Uganda, IFPRI selected

communities representing these domains for the community surveys. Two villages in Iganga

District, which were representing the development domain of interest and which were

captured by the community surveys were selected for this study.

5

Figure 1: Integrated Approach to Bio-Economic Modeling

Modelled System • Development Pathways/ 1. Overall description of Development Domain Farming Systems study area (IFPRI)

2. Comparative advantages 3. Identification of potential

pathways analysed Stratified Random Sampling

Conceptual Model • Farm Households within 1. Identification/Description Community Surveys

Socio-Economic + Agro- of external factors (IFPRI) Ecological Environment 2. Identification of economic/

ecological constraints

represented Stratified Random Sampling Representational Model • Farm Household Models 1. Identification of represent. Household Survey 1

(Singh) household types 2. Capturing differences in resources/wealth endowm./

encode economic opportunities Principal Component/ Cluster Analysis Computational Model • Mathematical Progr./ 1. Understanding of household Household Survey 2

Neural Networks decision-making process Nutrient Balances 2. Economic/ecological impact of promoted technologies 3. Impact of relevant policy options

Aggregation Procedure

Modeling

Process

Objectives Sampling Procedure

6

A census was conducted in both villages to get a complete list of the households, which

served as a sampling frame. For the first household survey stratified random sampling was

performed, in order to include the correct proportion of households conducting farm trials in

cooperation with CIAT (International Center for Tropical Agriculture) and Africa 2000

Network (A2N). After the identification of representative household types, households most

closely to the cluster centers were selected for the second household/plot level survey.

3.1 Farming Systems and Development Domains in the Study Area (Modelled System)

The concepts of farming systems and development domains help to define the overall socio-

economic and agro-ecological environment of the particular study area. Development

domains are characterized by their population density, market access, and agricultural

potential (Pender et al., 2001). Since different domains have different impacts on land

management, productivity, natural resource conditions, and welfare outcomes, comparative

advantages can be identified. These comparative advantages in turn lead to constraints and

opportunities for a sustainable development and to the identification of potential policy

measures.

Iganga District is located in the Lake Victoria Basin in Eastern Uganda. Iganga town is about

120 km north-east from Kampala (capital city of Uganda) and about 100 km away from

Busia, a important trading center in Kenya. The district belongs to the “Intensive Banana-

Coffee Lake Shore System” (Bashaasha, 2000) with high agricultural potential, high

population density, and high market access (Pender et al., 2001). Primary goal of production

is home consumption for the majority of the farm households. Traditional food crops are:

maize, bananas, sweet potatoes, cassava, beans, millet, sorghum, and Irish potatoes. The

traditional cash crops are coffee and cotton. The location of the district has an altitude of

1070-1161 meters above sea level and covers an area of about 11.113 km 2. The bimodal

rainfalls are varying from 1250 to 2200 mm per annum (Esilaba et al., 2001). Orthic

Ferralsols are the predominant soils in Iganga District (FAO, 1977). Magada and Buyemba,

the villages under consideration, are part of Imanyiro sub-county, which is located at 00 35´N,

32029´E.

In the last two decades the process of market liberalization and decentralization has

influenced land management at the farm level in Uganda in many different ways. Subsidies

were abolished and farmers are not assured of government sourced input supplies any more.

Additionally, minimum prices for the commodities are not guaranteed as in the past. Prices

are low and fluctuating, but markets have been developed for the majority of the products.

7

Another consequence of the economic and policy reforms is the shifting of responsibilities for

extension services to the district level. Many NGOs have therefore intensified their efforts to

advice farm households on agriculture-related issues. Numerous NGOs are particularly active

in Iganga District. Different organizations, e.g. CIAT, A2N or Sasakawa Global 2000, are

promoting various types of technologies helping to overcome the interrelated problems of

land degradation, poverty and food insecurity. Their main objective is to encourage the

adoption and diffusion of soil productivity enhancing technologies (e.g. inorganic fertlizer,

organic fertilizer and soil and water conservation methods).

Considering the characteristics mentioned above, Iganga District falls into the category of a

program-induced development pathway with high population density, high market access, and

high agricultural potential. Taking into account the comparative advantages, a potential

profitable pathway involves intensive production of high value perishable crops, perennial

crops or livestock, or development of off-farm activities. The fact that agriculture in Iganga

District is predominantly a low-intensity, low-input and low-output system reveals that farm

households have not realized yet the comparative advantages of their region. These

considerations lead to another interesting research problem, which has to be addressed in this

study:

Why is agriculture production in a region belonging to a development pathway with high

population density, high market access and high agricultural potential predominantly

characterized by low intensity, low-input and low output systems?

3.2 Farm Households within their Socio-Economic and Agro-Ecological Environment

(Conceptual Model)

The farm households as the decision-making entities within their socio-economic and agro-

ecological environment can be defined as the conceptual framework (Upton, 1996; Ruben et

al., 1997). The agro-ecological and socio-economic environments are considered to be the

most important external factors determining farm household decision-making. The agro-

ecological environment (soil quality, climatic conditions etc.) defines the potential

agricultural production activities from which the households can select. On the other hand the

socio-economic environment (markets, service, and infrastructure) gives incentives or

disincentives to select from these activities. Policy interventions lead to changes in the socio-

economic environment resulting in different (dis)incentives for the farm households. The final

outcome of the decision making process of the household is reflected in the production

pattern, productivity, social well being of the household, and the impact on sustainability.

8

Therefore, the farm household framework can be used to assess the implications of different

policy measures for crop and technology choice, production, market exchange, labor use, and

farm household welfare. Differences in risk behaviour (Roe and Graham-Tomasi, 1986),

market failures or missing markets (de Janry et al., 1991), and inter-temporal choice

(Fafchamps, 1993) can be taken into account as well.

IFPRI selected communities for the community surveys using a stratified random sample of

the identified development domains in Uganda. Out of the communities captured by IFPRI,

Magada and Buyemba were selected for this study representing the pathway defined above.

Therefore, mainly IFPRI survey data were used to characterize the socio-economic and agro-

ecological environment the farm households are facing in Magada and Buyemba.

Asked for the area that has the most positive impact on life since 1990, representatives of the

villages reported that security and peace improved substantially giving the safety needed for

non-restrictive social and economic activities.

The market access is relatively high for both villages, with trading centers for basic

commodities and inputs being away 1,5 km. Iganga town, one of the major towns in Eastern

Uganda, is about 20 km away from Magada and about 30 km from Buyemba. Moreover, the

relatively short distance to Kampala, Mbale and Busia presents attractive trading

opportunities.

The nearest tarmac road is about 10 km away from Magada and 20 km from Buyemba. The

centers of both villages are crossed by seasonal mud roads. The distances to the next primary

and secondary school are 500 m and 5 km respectively in Buyemba, and 800m and 2 km in

Magada. The next health center is 500 m away from Buyemba’s center and 300 m from

Magada’s center.

The majority of the households is selling their products at the farm gate to middle men in the

villages. Farmers are complaining about the low and fluctuating level of output prices. High

transaction costs (caused by lack of information) and imperfect competition are the

determining factors. The middle men are selling the products to local buyers in trading centers

and Iganga town. Part of this amount is sold to traders from Kenya or Mbale, Busia and

Kampala. Low level of agricultural input use was justified with high prices, especially for

inorganic fertilizer. Inefficiency in procurement, high transportation costs, and absence of

competitive pressure are leading to unreasonably high fertilizer prices (IFDC, 1999).

The labor markets can be characterized by a declining importance of labor exchange and an

increasing importance of hiring labor within the last 10 years. Increased off-farm

commitments were mentioned as the main determining factor, although off-farm employment

9

opportunities – especially on a permanent basis - are limited. Labor exchange is a traditional

form of labor acquisition, where men or women with the same level of education form

working groups, based on the idea of overcoming labor constraints of its members. It is not

common among group members to pay for labor, occasionally labor is paid in-kind. The wage

rate for one hour of hired labor can be set at around 500 USh on average1. Factor markets of

livestock exchange for land preparation do not exist.

The average farm size declined from 1990 to 2000, whereas the proportion of cultivated land

increased in comparison to other land use categories such as fallow, grazing areas, and natural

forest/woodland. The consequence is a declining availability of cropland with declining

fallow periods. The most important methods of obtaining access to agricultural land are

getting land as a gift from relatives, purchase, and fixed rental. The price for land with annual

crops is three times higher in 1999 than it was in 1990. As indicated above, the dominant land

tenure system is freehold.

A fast growing population is identified as the main reason for the developments on the land

market with its increasing prices and declining land availability. The number of households

increased by 8 % per year from 1990 to 2000 in Buyemba, where a fast growing population

within the village, and in-migration due to the attractiveness of the trading center were

indicated as the major factors.

The access to credit is limited. One micro-finance institution is mentioned as the only formal

source. The Entandikwa fund provided by the Ugandan Government was stopped in 1997

because of bad repayment moral. Informal sources are used more frequently, it was reported

that 50 % of the households used relatives and friends as a credit source in 1999, 25 %

borrowed from traders, and 10 % from money lenders.

The perceived changes in resource conditions are a major deterioration for most natural

resource items, such as soil fertility, soil moisture holding capacity, soil erosion, and quality

of natural water sources. A major decrease in yield levels since 1990 was reported for main

crops, such as maize, millet, and bananas. For other crops a minor decrease or stagnation was

identified as the yield trend.

Summing up, market imperfections (e.g. low output prices, high input prices and missing

credit markets), declining soil fertility, prolonged dry seasons, as well as pests and diseases

were identified as major constraints for agricultural production.

1 1 US $ are 1776 USh (exchange rate from 9.4.2002).

10

3.3 Farm Households (Representational Model)

Farm household models offer a promising perspective for the analysis of production and

consumption decisions at the farm level (Singh et al., 1986). Farm households are considered

to be the central decision makers regarding agricultural production. Individual farmers have to

decide which commodities to produce in which quantities, by which method, and in which

seasonal time periods. It is the objective of the farmers to maximize their utility, which

deviates from pure profit maximizing behaviour in many cases. For example, leisure and the

provision of enough food for household consumption are important goals, which have to be

taken into account, too. The decision-making procedure is subject to physical and financial

constraints (e.g. acres of land, days of labor and limited credit availability). Linkages between

production and consumption decisions, characteristic for farm households operating under

imperfect markets, have to be included. Due to the possibility of analysing both, production

and consumptions decisions, the farm household model approach represents a useful starting

point for the analysis of the effectiveness of economic policy instruments supposed to

enhance a sustainable intensification of land management.

The main objective of the first survey was to identify representative household types of the

selected two villages in Iganga District. Agricultural producers differ in their wealth,

economic opportunities, and resource endowments. Therefore, the investigation of producer

response to policy changes requires the identification of typical farm households and their

inclusion in the modelling approach (Hazell and Norton, 1980). The procedure of stratified

random sampling was performed in order to reflect the proportion of non-trial households and

trial-households in the sampling universe.

A listing done in 2000 indicated that 44 out of 608 households are conducting agricultural

technology trials in cooperation with CIAT and Africa 2000 Network. 11 different

technologies are promoted aiming at a sustainable intensification of agriculture, including

inorganic fertilizer (NPK), farmyard manure, trenches, green manure, and improved fallow.

Out of the first strata of 564 non-trial farmers 62 were selected randomly. Out of the second

strata of 44 trial farmers 5 were selected randomly for the subsequent analyses for

identification of representative household types. The other 39 trial farmers were also covered

in order to collect reliable data for technical coefficients of different technologies.

11

A Principal Component Analysis with a subsequent Cluster Analysis are used to select

representative household types in the study region. The objectives of the Principal

Component Analysis were:

1. to analyse the structure of correlation among variables by defining a common set of

underlying factors,

2. to differentiate relevant from irrelevant variables for the subsequent Cluster Analysis

(variables which are not distinctive across the households can be eliminated).

3. the subsequent Cluster Analysis can be conducted with uncorrelated factor scores

(Backhaus, 1994). A Factor Analysis reduces highly correlated variables to one factor.

If highly correlated variables are used for the subsequent Cluster Analysis, some

characteristics have a stronger impact than others when it comes to the clustering of

objects.

4. data can be reduced for the subsequent Cluster Analysis (Hair, 1998).

One side effect of this multivariate data analysis is certainly the associated loss of statistical

information.

Various variables, captured in the first survey, are used as inputs for numerous Principal

Component Analyses. Because of their correlation structure and their relevance the following

8 variables were selected for the final Principal Component Analysis:

time of adoption (of improved cassava variety) compared to opinion leader, number of

inorganic fertilizers/agrochemicals, number of trial types conducted, number of different

types of training, values of residence buildings and other structures of the household, values

of radios, perceived walking time to output market, and percentage of quantity disposed on

total production. Different measures (e.g. Kaiser-Meyer-Olkin Measure of sampling

adequacy, Bartlett Test of Sphericity) indicate that this set of variables is appropriate for a

Principal Component Analysis (see table A 5 in appendix). The most common criteria for the

number of extracted factors is the eigenvalue. Three factors have an eigenvalue greater than 1,

explaining together 67 % of the variance (see A5). Table 1 illustrates the rotated component

matrix with the factor loadings. The factors could be titled as: innovativeness, household

assets, and market orientation.

12

Table 1: Principal Component Analysis

Rotated Component Matrix a

,90 ,19 ,19

,88 ,06 ,23

,84 ,20 -,10

-,51 ,25 ,26

-,04 ,84 -,04

,35 ,77 -,10

,00 ,07 ,73

-,08 ,22 -,68

number of inorganicfertilizer/agrochemicalnumber of trial typesconductednumber of differenttypes of trainingtime of adoptioncompared to opinionleadervalues of residencebuildings and otherstructures of thehouseholdvalue of radioswalking time outputmarket(minutes)% of quantity sold ontotal production

1 2 3Component

Extraction Method: Principal Component Analysis. Rotation Method: Varimax with Kaiser Normalization.

Rotation converged in 6 iterations.a.

Source: own calculations Computed factor scores are used as the input for the subsequent Cluster Analysis. The main

reasons for performing this type of multivariate analysis are 1) identification of homogenous

household groups and 2) to provide a criterion for the selection of households for the in-depth

interviews in the second household survey. Regarding the clustering algorithm, a combination

of hierarchical and non-hierarchical methods was chosen in order to fine-tune the results and

to have a validity check. For the hierarchical clustering the similarity measure chosen is the

commonly used Squared Euclidean Distance, which is computed by the following formula:

2

1

2 )( jk

p

kikij

xxD −= ∑=

,

where

Dij

2 = squared distance between object i and j ,

xik = value of kth variable for the ith object,

xjk= value of the kth variable for the jth object,

p = number of variables.

Regarding the clustering technique, Ward’s Method was applied, which belongs to the

agglomerative methods. In the agglomerative methods each object starts as its own cluster. In

subsequent steps the two closest clusters (or objects) are combined into a new aggregate

13

cluster. The Ward’s Method forms clusters by maximizing within-clusters homogeneity.

Within-group sum of squares is used as the measure for homogeneity. Therefore, Ward’s

Method tries to minimize the total within-clusters sum of squares (Sharma; 1996).

Non-hierarchical methods have some advantages over hierarchical, e.g. the results are less

susceptible to outliers in the data, the distance measure used, and the inclusion of irrelevant

variables (Hair et al. 1998). Since these benefits are only realized with specified seed points, a

combination of both methods was chosen. Furthermore, the hierarchical analysis determines

the appropriate number of clusters. After the initial seed points are selected (these are the

cluster centres of the hierarchical method), all objects (households) within a pre-specified

threshold distance are included in the resulting cluster. Then for each cluster new centres are

computed and the objects are assigned again. Objects may be reassigned if they are closer to a

new cluster centre. No standard procedure for the number of clusters to be formed exists. A

simple example for a stopping rule is to look on large increases in the agglomeration

coefficient. A large increase indicates that two very different clusters are being merged. Table

A 6 in the appendix illustrates the calculated percentage change in the coefficient for 10 to 2

clusters. There is a large percentage increase when the number of clusters is reduced from 4 to

3 clusters. Therefore, forming 4 clusters seems to be an appropriate solution.

Finally, the following four clusters were identified (see Figure 3): subsistence farm

households (30 %), semi-subsistence farm households (52 %), commercial farm households

(10 %), and innovative trial farm households (7 %).

Figure 2: Identified Farm-Household Groups

52%30%

10% 7%Semi-Subsistence Farm-HouseholdsSubsistence Farm-Households

Commercial Farm-Households

Innovative Trial Farm-Households

Source: own calculations (based own survey 2000)

14

beans

maize

cassava s.pot. coffee

bananasgnuts

other

0

10

20

30

40

50

%land

Figure 3: Average proportion of cultivated land under major crop types

Source: own survey 2000

Descriptive statistics provide information on household characteristics, household assets (see

appendix A 1 – A 4), major crop types (Figure 3), main differences between the different farm

household groups (see Table 2), major information sources for applied technologies (Figure

4), and reasons for non-adoption of technologies (Figure 7).

The average farm size in the villages is about 2.3 ha; the dominant type of tenure is freehold.2

The main crops grown in the area are maize, beans, cassava, sweet potatoes, bananas, coffee,

fruits, and vegetables (Figure 2). The majority of the farms have few or no livestock and the

mean numbers are 1 bovine livestock, 2 goats and 5 chickens per farm. Agricultural

production can be characterized by low levels of fertilizer application and poor agronomic

practices based on rain-fed agriculture with hand hoe cultivation.

The innovative trial farm households belong to the early adopters of a mosaic resistant

cassava variety. They are (by definition) the only group, which is conducting trials in

cooperation with CIAT. These households apply the highest number of inorganic fertilizers

and other agrochemicals, and they are the only group with a reasonable number of different

types of agricultural training.

The commercial farm households achieve the highest mean values for the following variables:

value of residence and other structures of the household, value of agricultural equipment per

person involved in farming, total value of agricultural production, value of agricultural

production per acre cultivated land, and quantity sold on total agricultural production.

Interesting is the fact that they belong to the late adopters of the mosaic resistant cassava

variety. There are two explanations: 1) cassava is not important as a cash crop and therefore

not of major interest for commercial farm households 2) wealthier households are excluded

from the communication process of the average farm households. Subsistence and semi-

2 The average proportion of land under freehold status is 72 % (see appendix A 4).

15

subsistence farm households attain relatively low mean values for the following variables:

years of schooling of household head, value of household assets, quantity sold of total

agricultural production, value of agricultural production (total and per acre cultivated land),

and number of inorganic fertilizers and other agrochemicals applied. Furthermore, the

subsistence farm households face very long walking times to the next output market and

belong to the group of late adopters.

Table 2: Characteristics of the Identified Clusters

Semi-Subsistence Farm Households

Subsistence Farm Households

Commercial Farm Households

Innovative Trial Farm Households

Household Characteristics Years of schooling head 4.4 5.5 12.4 7.7 Years of schooling wife 4.3 3.3 8.1 5.6 Number of different types of training (since 1990)

0.7 0.3 1.0 4.2

Household Assets Value of residence and other structures (103USh)

837 1267 7601 1951

Value of radios (103USh) 22 16 74 43 Value of agricultural equipment per person involved in farming (USh)

5261 4358 9739 5778

Crop Production Total value of agricultural production (103USh)

833 455 1635 1066

Value of agricultural production per acre cultivated land (103USh)

182 182 224 207

Quantity sold on total production (%)

52 23 64 35

Perceived walking time to output market (minutes)

45 142 64 81

Intensity of land use3 1.2 0.9 1.4 1.1 Labor-land ratio4 131 260 165 159 Innovativeness Time of adoption (improved cassava variety) compared to opinion leader (years)

0.7 4.6 5.8 -2

Time of adoption (improved cassava variety) compared to personal network (%)

0.41 0.66 0.73 0.33

Number of technologies adopted last 10 years

5 5 6 8

Number of trial types conducted

0 0 0 7


3 intensity of land use is defined as ratio between land area cultivated last 12 months and total land size 4 labor-land ratio is defined as ratio between labor use on farm (person days) and cultivated land size

16

Peers and friends are after CIAT the most important information sources for applied

technologies. Figure 4 illustrates the enormous impact NGOs apparently can have on the

livelihood strategies of farm households. Shifting the responsibilities for extension services to

the district level within the decentralization process does not seem to work properly yet in the

study region. Restructuring the Ministry of Agriculture, Animal Industries and Fisheries

(MAAIF) involved a drastic reduction in the number of staff. Selecting innovative households

and opinion leaders for the CIAT technology trials seems to be a promising strategy for a

widespread diffusion of innovations (Esilaba et al., 2001). The theory on diffusion of

innovations emphasises the importance of social networks for the adoption and diffusion of

innovations (Rogers 1995, Valente 1995, Berger 2001).

CIAT/A2N

DFI

Peers/Fr.

SG RelativesExtension

Other

0

20

40

60

Frequency

Figure 4: Major information sources for applied technologies

Abbreviations: DFI=District Farm Institute; SG=Sasakawa Global 2000


Rogers (1995) pointed out that typically the cumulative S-shaped adopter distribution closely

approaches normality. The normal frequency distribution has several characteristics that are

useful in classifying adopters. Mean values and standard deviations are used to classify

adopters in the following four categories: early adopters, early majority, late majority and

laggards. Valente (1995 and 1996) introduced a method to estimate thresholds of adoption,

empirically based on the concept of social networks. Empirically estimated thresholds of

individual farm households to adopt new technologies will help to illustrate the diffusion

patterns of promoted technologies. Figure 5 confirms Rogers’ classification of adopter

categories for a mosaic resistant cassava variety in the study area.

The African Cassava Mosaic Virus (ACM-V) has had a devastating effect on cassava, one of

the major food crops in the region. Therefore, in the middle of the 90s the International

Institute of Tropical Agriculture (IITA) and the National Agricultural Research Organization

17

(NARO) initiated a cassava multiplication programme for multiplying ACM-V resistant

cassava varieties in the district (Vredeseilanden-Coopibo-Uganda, 1998).

Insufficient awareness, unavailability, and high input costs are the most important reasons for

the non-adoption of new technologies (Figure 6). These figures are an indicator for the

importance of well-functioning markets and social networks for the adoption and diffusion of

innovation.

time of adoption compared to social system

5,04,03,02,01,00,0-1,0-2,0-3,0

Figure 5: Time of Adoption (Cassava)

Freq

uenc

y

20

10

0


new

insuf.aware

not available

input costs

other

0

50

100

150

Frq.

Figure 6: Reasons for non-adoption of technologies


3.4 Bio-Economic Household Model (Computational Model)

The main objective of the second household/plot-level survey was to provide useful input for

the calibration of the programming matrices, including data for the estimation of input-output

coefficients and farm income analysis. Altogether 20 households were interviewed in-depth.

18

Out of each cluster at least three households were chosen, which are most closely to the

cluster center. Additionally, 7 farm households out of the 44 trial farmers were captured with

respect to the type of trials they are conducting in order to collect sufficient data for analyzing

the effects of promoted technologies. Besides, CIAT provided the trial data of 4 seasons in

2000 and 2001, and soil data of their test farmers used for yield estimation.

Furthermore, 110 composite soil samples for 0-20 cm and 20-40 cm were taken, and 61 plots

were measured to provide more specific data for the subsequent modelling work. Compasses

were used for measuring plot angles, the distances were measured by tapes. Supplementary,

intensive cooperation with local technical experts and collection of secondary data allow a

modelling work reflecting the “real world situation”.

Bio-economic models combine socio-economic factors influencing farmers’ objectives and

constraints with biophysical factors affecting production possibilities and the impacts of land

management practices (Barbier 1996 and 1997; Kruseman et al. 1997). These models may

identify the optimal level of technology adoption and the impact on incomes and natural

resource conditions under a changing socio-economic environment. Implemented as multi-

agent systems (Berger 2001) outcomes from agent-agent and agent-environment interactions

could be captured as well, e.g. diffusion of innovations together with the evolution of farm

incomes and natural resource conditions over time. The bio-economic modelling approach

developed for this study consists of three major components: mathematical programming

model to reflect the farm household decision-making process under certain constraints,

artificial neural networks as a yield estimator, and nutrient balances as a sustainability

indicator. The results of the yield estimator and the calculation of nutrient balances are then

incorporated into the programming model. In the following each component and the

integration into one modelling approach will be described.

3.4.1 Artificial Neural Networks as a Yield Estimator

Artificial Neural Networks (ANN) belong to a new research area called artificial intelligence.

ANN attempts to mimic the human brain and its structure to develop a processing strategy. As

in the human brain, multiple parallel processing units are engaged in pattern recognition.

There are many different types of ANN with different applications available, which are

described for example in Bishop (1998). As far as the author knows neural networks were not

used for yield estimation purposes before. Park and Vlek (2002) examined the possibility of

predicting soil property distribution with ANN. The main reasons why this type of model was

preferred for yield estimation instead of a deterministic simulation model were:

19

• their enormous data requirements, which cannot be satisfied in the study region

• it is frequently quoted that many deterministic models are just working at the calibrated

sites (Beven et al., 1989 and Hoosbeek et al., 1992)

• their limited capacity of estimating the impact of many different technologies.

Traditional statistical approaches, e.g. General Linear Model (GLM) and multiple regressions,

have several limitations to model complex and non-linear dynamics in crop yield. The main

differences between GLM and multiple regression analysis are that for the GLM the

distribution of the dependent variable does not have to be continuous, and GLM allows linear

combinations of multiple dependent variables (see Park and Vlek, 2002). The modeling

accuracy (R2=0.75) of the ANN yield estimator is promising.5 Comparing the results for yield

estimations of the ANN with GLM (Figure 7) led to the conclusion that non-linearity should

be taken into account while estimating the impact of soil characteristics and different

technology options on yields.

Figure 7: Comparison of Modelling Accuracy ANN - GLM

0,00,10,2

0,30,40,50,6

0,70,80,9

1,0

SY ASY CY ACY GY AGY TB ATB

Yield Index

R2

ANN

GLM

Abbreviations: SY=stover yield, ASY=adjusted stover yield; CY=cob yield; ACY=adjusted cob yield;

GY=grain yield; AGY=adjusted grain yield; TB=total biomass; ATB=adjusted total biomass

Figure 7 illustrates that ANN achieves a higher modeling accuracy not only for grain yield,

but for all other yield indices considered as well, e.g. stover yield and cob yield. The main

limitations of ANN are its data dependency and the absence of any statistical inference tests

for model weights of overall model fit.

5 For ANN output reports see appendix A 8.

20

1124 CIAT farm trial data, collected during 4 seasons in 2000 and 2001 in Magada and

Buyemba were used for the yield estimation. 25 farm households participated in these maize-

trials. Additionally, eleven basic soil attributes were measured for one sample per farmer.

Therefore, the following yield function was estimated with the ANN:

Y = f (PH, OM, N, P, K, Na, Mg, Sand, Clay, Silt, FYM, GM, BRP, MRP, BB, Pre-Pac, TSP,

Nfert, NP, NPK),

where

Y= maize yield level

PH=soil-ph, OM=organic matter, N=nitrogen content of soil, P=phosphorus content of soil,

K=potassium content of soil, Na=sodium content of soil, Mg=magnesium content of soil,

Sand=sand content of soil, Clay=clay content of soil, Silt=silt content of soil,

FYM=farmyard manure trial, GM=green manure trial, BRP=Busumbu rock phosphate trial,

MRP= Minjingu rock-phosphate trial, BB= Busumbu Blend trial (90% Busumbu rock

phosphate, 10% TSP), Pre-Pac= Pre-Pac (rock phosphate, urea and rhizobia), TSP=Triple-

Super-Phosphate trial, Nfert=Urea-trial, NP=Urea+Triple-Super-Phosphate trial,

NPK=Urea+Triple-Super-Phosphate+Muriate of Potash trial

During the process of building a network, the type of neural network model, the number of

nodes, the type of activation function, number of hidden layers, and the learning rule have to

be determined. The neural network type applied in this study is the most commonly used

multiplayer perceptron model. The ANN is a sequential arrangement of three basic types of

layers: input layer, output layer, and hidden layer. These layers consist of small processing

units called nodes, which are analogous to the neuron of a human brain. Nodes are receiving

input values and creating output values. The incoming data is processed by creating a

summated value in which each input value is multiplied by its respective weight. The summed

value is processed by an activation function to create output, which is sent to the next node in

the system. The activation function chosen here is the commonly used hyperbolic tangent

function in order to introduce non-linearity. An output node receives an input value and

calculates an output value. This value is the final one, it is not sent to another node. In order to

represent more complex relationships than just a one-to-one relationship from input to output

a third type of layer (hidden layer) is introduced, which is located in between the input and

output layer.

21

The main difference in comparison to other multivariate techniques is the ability of the neural

network to “learn”. The created output values are compared with the actual values. If there is

a certain difference, the error in the output value is calculated and then distributed backward

through the system. This common form of training is called “backpropagation”. During the

modeling process the number of nodes and hidden layers were changed. The best performance

(lowest Mean Square Error and highest R2 (0,75)) was achieved with 21 input nodes, 8 output

nodes and one hidden layer with 25 nodes. ANN analyses were conducted using the software

NeuroDimension 3.0 (Principe et al., 2000).

Figure 8: ANN Model Output -Impact of technology options on maize yield

-60%-40%-20%

0%20%40%60%80%

100%120%140%

FYM MRP BRP TSP Blend Pre GM Nfert NP NPK

chan

ge

of

yiel

d soil1soil2soil3soil4soil5soil6

Source: ANN model output (based on CIAT trial data 2000 and 2001)

The ANN output desired for the programming matrix is to quantify the impact of different

land management practices and soil classes on the yield. Therefore, the CIAT soil data and

the soil data collected during the second household/plot level survey were used for clustering.

Finally, six different soil classes were identified (see A 6)6. The soil characteristics of the

different classes and the technology options mentioned above were tested against the trained

network to receive the desired output. Figure 8 illustrates the impact of different technology

options on maize yield in comparison to the control plots without any treatment as received

from the ANN model. The ANN estimates for yield response under BRP, Blend, NP and NPK

treatments are quite positive with an average yield increase of around 40%. For TSP the

average increase was 26% and for MRP 16%. The other trials (FYM, Pre-Pac, GM and NP)

had on average nearly no impact on the yield in comparison to the control plots.

6 Hierarchical clustering with Ward’s Method as cluster method and Squared Euclidean Distance for interval measurements was chosen.

22

Figure 9: Experimental data: Impact of technologies on maize yield

-10%

0%

10%

20%

30%

40%

FYM MRP BRP TSP Blend Pre-Pac GM Nfert NP NPK

ch

ang

e o

f yi

eld

Source: own calculations (based on CIAT trial data 2000 and 2001)

ANN model results and descriptive analysis of experimental data received from CIAT (Figure

9) indicate similar effect of the promoted technologies on yield. The only major difference

can be observed for the impact of Blend. As mentioned above, the ANN model indicates a

significant positive impact of this technology on maize yield, whereas, the average impact

received from the experimental data is very low (2 %). This is a surprising result since the

experimental data show a positive impact of BRP as well (nearly 40 %), and Blend consists of

90 % BRP and 10 % TSP. One possible explanation for the differences of the results for

Blend received from the model and from the experimental data is, that for the latter data

source the impact of different soil classes is neglected. As Figure 8 illustrates the ANN model

indicates a low impact of Blend on soil class 2 (10 %) as well.

Apparently, the impacts of the selected technology options on maize yield are not as high as

one might expect from trials conducted at research stations under the control of scientists.

Therefore, the method of on-farm trials allows to estimate the incentives for the farmers to

adopt a specific technology more realistically.

3.4.2 Nutrient Balances as Sustainability Indicator

Nutrient balances for nitrogen (N), phosphorus (P) and potassium (K) were selected as a

sustainability indicator. The appropriateness of nutrient balances as an indicator for soil

productivity and sustainability assessment was discussed in the literature frequently (Lynam

et al., 1998). Among other aspects Lynam et al. (1998) criticized that farm household

decision-making exclusively based on nutrient balances leaves out any economic

consideration. There are cases where farmers consciously exploit nutrient stocks to invest in

23

capital that leads to sustainable development in the long run. The integration of nutrient

balances and economic considerations in a mathematical programming approach can help to

overcome this problem. Additionally, it is very difficult to quantify the feedback effects of

nutrient losses on the development of the yield level, taking into account the complexity of

nutrient dynamics in the soil. But nutrient balances can at least serve as an indicator of soil

productivity in the future considering the total nutrient stock in the soil.

The calculations are based on the concept of the study “Nutrient Balances and Expected

Effects of Alternative Practices in Farming Systems of Uganda” (Wortmann and Kaizzi,

1998). This study was carried out in Imanyiro Sub-county, to which also Magada and

Buyemba belong. Own data collection (especially the soil samples) helped to adjust the

nutrient balances for the households under consideration. The nutrients-removing factors

captured in the programming matrix are: erosion, harvest of the main product, harvest of the

crop residues, leaching, and denitrification. The nutrients-adding factors are: inorganic

fertilizers, farmyard manure, green manure, biological nitrogen fixation, and atmospheric

deposition. Nutrient losses through erosion are based on the Universal Soil Loss Equation

(USLE). The rainfall factor R was set at 4007, the topographic factor SL was estimated

according to Renard et al. (1994) for cropped land of moderate ratio of rill to inter-rill erosion,

and for fallow using scales for rangeland. The slope gradient was assumed to be 4 %,

according to a study carried out in Magada by the Center for Development Research. Since

the input-output coefficients of the programming matrix were calculated on the basis of one

hectar, assuming a quadratic plot shape, slope length was set at 100 meters. According to

Wortmann and Kaizzi (1998) the crop factor C ranged from 0.2 to 0.4 (0.2 for banana and

coffee intercrop associations, 0.3 for annual intercrop associations and sole crop bananas and

coffee, 0.4 for annual sole crops) and the erosion management factor P was set at 1 unless the

farmers were attempting to control erosion. The soil factor K was set at 0.04 (Barber et al.,

1979) and the nutrient enrichment factor of the runoff was assumed to be 1.5. The calculation

of N-losses through erosion is based on the organic matter content received from the analysed

soil samples. Extractable P was received from the soil analysis, total P calculation is based on

the method developed by van den Bosch (1997). K nutrient losses through erosion is based on

total K as calculated by Stoorvogel and Smaling (1990). Secondary data were used to quantify

nutrient contents of harvested products, residues, and fertilizers (Stoorvogel and Smaling,

1990; Defoer et al., 2000; Wortmann and Kaizzi, 1998).

7 Compare Wortmann and Kaizzi (1998).

24

3.4.3 Mathematical Programming

Currently available bio-economic models are based on mathematical programming

approaches. A mathematical programming model helps to find the farm plan (defined by a set

of activity levels) that maximizes the objective function, but which does not violate any of the

fixed resource constraints, or involve any negative activity levels.

This model type offers great possibilities to formulate a wide range of actual and potential

activities and to determine their relative attractiveness. Advanced techniques offer the

possibility to reflect farmers ̀ behaviour realistically, e.g. the inclusion of risk aversion and

household food requirements in the objective function and constraints (Hazell and Norton,

1985).

Two extreme prototypes of agricultural programming models were defined by Hanf (1989).

The “simultaneous equilibrium approach” maximizes a common sectoral utility function and

assumes a perfect market mechanism. Secondly, the “representative independent farm model”

that can be defined as an independently computed farm model representing a certain farm

type. Computational results are added up to regional results. Hanf (1989) concludes that the

latter model should be chosen, if the sector development is characterized by 1) imperfect

markets, 2) behavior other than pure profit maximization and 3) adjustment processes.

Therefore, a model approach similar to this type seems to be an appropriate choice taking into

account the conditions in the study area. The approach offers the possibility to analyse the

behaviour of the individual farmer. Programming models are able to simulate adjustment of

land use under changing conditions. Therefore, they are an appropriate approach to analyze

the choice among alternative activities and technologies and to assess the impacts of

alternative policies in the short and long run.

Brandes (1985) criticized linear programming methods in the sense that as a consequence of

compensating errors and due to the temptation of manipulation, the model builder could give

the impression that his model reflects reality. An important weakness of these conventional

simulation models, apart from the aggregation error, is that they do not explicitly capture the

interactions between the farm households and therefore neglect transaction and information

costs.

The mathematical programming model computes the optimal production and consumption

plans based on a lexicographic utility concept: the households first satisfy their nutrition goals

before maximizing the household income, subject to financial, technical and sustainability

constraints. Nutrient requirements and consumption preferences that the households

articulated during the in-depth interviews are included. The farm household decision-making

25

problem is captured through mixed-integer programming consisting of 507 variables and 201

constraints. The inclusion of risk aversion in the objective function (e.g. by MOTAD or

quadratic risk programming, Hazell and Norton 1980) was neglected since district agricultural

annual reports did not indicate a high variability of yields of main crops during the last 10

years (e.g. mean yield value for maize: 1467 kg/ha, standard dev.: 130 kg/ha). Of course the

yield variability at the disaggregated farm household level could be different, but

corresponding time series data were not available. Discussions with extension officers

confirmed that the variation of yields does not have a significant impact on the planning

process of farm households. Data from the second household/plot-level survey were used to

calculate the variable costs per hectare of land for crop activities and per livestock unit. Gross

returns of the production activities were not captured in the objective function directly, in

order to give the household the option either to sell or to consume the produced commodities.

The market or farm gate price of the products appears as the objective function value in the

respective selling activity. Production expenses, which were not included in the variable

costs, such as hiring in labor, hiring in tractor, borrowing credits etc. were accounted through

the objective function values of the respective activities.

Activities included in the applied programming approach

The main activities captured are: crop production, livestock production, consumption and

selling of agricultural products, permanent off-farm employment, hiring in/hiring out

temporary labor, labor exchange, labor transfer, hiring tractor, investment activities,

borrowing credit, and technology options based on CIAT farm trials.

Regarding crop production activities, two growing seasons, six different soil types, different

cropping methods (mono-cropping, inter-cropping), and different technology levels

(improved seeds, local seeds, different seed rates etc.) are taken into account. Crops under

consideration are: maize, beans, sweet potatoes, cassava, groundnuts, millet, sorghum, coffee,

bananas, tomatoes, onions and passion fruits. For livestock production activities the focus is

on local and exotic cows. Small stock production activities (e.g. chicken, porkers and goats)

were neglected due to lack of reliable data at the farm household level.

The model allows for the option to sell the whole amount produced of a specific crop, to use it

for home consumption or to choose a combination of both. The prices for agricultural inputs

and outputs were taken directly from the household survey. Besides, if necessary, secondary

data (APSEC, 2000) were used.

26

Households are offered to opt for permanent off-farm activities as a binary activity, if the

interview indicated relevance. Furthermore, temporary off-farm activities are divided into 5

different time periods throughout one year. For the same time periods casual labor can be

hired in. The division into 5 periods guarantees that enough labor is available for farming

activities which have to be carried out at fixed points of time. A price difference between the

activities of hiring in and hiring out labor reflects control costs for the time labor is hired out.

Labor activities were divided into male and female labor, since certain agricultural production

activities are traditionally only carried out by men (e.g. land clearing), others only by women

(e.g. cultivation of sweet potatoes). Transfer activities make sure that labor is provided for

production activities, which are carried out by both, male and female worker, and only adult

labor (older than 16 years) is hired out. Labor exchange was captured as an activity although

its importance is declining. As mentioned above, the labor market changed during the last 10

years from labor exchange oriented to labor hiring oriented. Scenarios with two different

labor sources could help to understand the impact of the changes in the labor market on land

management practices. Constraints guarantee that only as much labor is imported by labor

exchange as labor is exported by the farm household. There appears no objective function

value for the labor exchange activities. Its value for the household is reflected by the shadow

prices, which are calculated within the sensitivity analyses.

Because of their policy and practical relevance in Iganga District, the following investment

activities were chosen: treadle pump for irrigation, walking tractor, and zero-grazing unit.

Data on investment costs, repair and maintenance costs and technical details were received

from AETRI (Agricultural Engineering and Appropriate Technology Research Institute under

NARO). Investment costs and costs for repair and maintenance were annualized.

Financial constraints are often mentioned as one of the major constraints for the adoption of

modern agricultural technologies. Therefore, two different types of credit schemes were

considered explicitly: micro finance schemes operated by commercial banks and micro-

finance institutions (e.g. Finca, Pride Africa). Credits for investments and for annual

production costs were differentiated for each scheme. Different credit conditions, such as

credit limits, gestation periods and interest rates, were accounted for through respective

formulation of activities and constraints of the programming matrix.

Finally, different technology options based on CIAT farm trials are captured as additional

production activities (see 3.4.1). Application rates and input prices were received either from

CIAT/A2N or collected from markets.

27

The FAO is conducting a small scale irrigation project in cooperation with the Ugandan

Ministry of Agriculture, Animal Industry and Fisheries in seven districts, among others in

Iganga’s neighbour district Jinja.8 Gross margin calculations received from project officials

were used to capture the economic impact of irrigation for farm households. Irrigated crops

were limited to high value crops, such as clonal coffee, tomatoes, onions, and passion fruits.

Labor requirements for irrigation with treadle pumps were received by interviewing AETRI

staff. Irrigation is restricted to the dry seasons (December – March and July – September) in

the model. A2N is already promoting small scale irrigation with treadle pumps in neighbour

villages.

Constraints included in the applied approach

The main resources and constraints considered are: total land area, crop rotation, labor,

nutrient requirements of household members, consumption preferences, capital constraints

(including credit limits), and nutrient balances as a sustainable indicator.

Land is the basic resource of production. The land related production activities are expressed

on a per hectare basis. It is taken into account that a certain percentage of the total land area is

needed for fallow. The available land resources are divided into six soil classes as mentioned

above. Discussions with soil scientists from ZEF reveal the importance of considering each

soil parameter for an identified soil class. One parameter might give the impression of a fertile

soil, whereas, other values are very low or even reach toxic levels (compare appendix A 9).

Soil class 1 is characterized by average values for ph (5.7), organic matter content (2.7 %) and

potassium content (22 mg/100 g soil). Outstanding is the low phosphorus content (3.3 ppm).

The ph-value of soil class 2 is a little bit lower (5.2), the phosphorus content is higher (4.3

ppm) and the sand content is quite high as well(71 %). Soil class 3 is characterized by a

relatively low ph-value (4.2), whereas for soil class 4 a relatively high ph-value is indicated.

Furthermore, the calcium content is extremely high on soil 4 (160 mg/100 g soil). A very high

phosphorus content is reported for soil 5. Soil 6 can be characterized by a low ph value of 4.2,

a low organic matter content of 1.4 % and an extreme high sand content of 81 %.

Crop rotation constraints make sure that the cultivation of a specific crop does not exceed a

certain percentage of the total land area. The fact, that continuous cultivation of one crop on

the same plot would lead to nutrient mining, pest and diseases is realized by the farmers. The

crop rotation constraints were formulated with the help of local extension officers.

8 Project title: Small Scale Irrigation Technology Demonstration, Project Symbol: TCP/UGA/8821

28

Labor is a major factor of production in the less mechanized and less developed agriculture in

the study area. Labor requirements by various production activities and labor availability

during an agricultural year jointly determine the resource allocation and productivity of a farm

household. Because of the uneven distribution of land and household members among

different farm households, the labor availability per unit of land varies across farms.

Therefore, theses farms have different degrees of work participation on the farms and in the

rural labor markets. The production period of 12 months was divided into 5 time periods to

make sure that different critical time spans during an agricultural year are captured with

respect to their labor requirements. If seasonality of resource use is ignored, it is likely that

the model solution obtained will be unrealistic by requiring more resources (here labor) in

some periods than are available (Hazell and Norton, 1980). Furthermore, labor was divided

into three age groups to capture their different productivity levels (child labor between 14-16

years is treated as equivalent to 80 % of adult labor, adult labor above 55 years is treated as

equivalent to 50 % of adult labor between 17 and 55 years). It is assumed that men and

women with primary activity farming and no off-farm activity are working 6,5 hours per day

in agricultural production. If they have an off-farm activity as a secondary activity, a

minimum of 4 hours is spent on agricultural production and a maximum of up to 4 hours on

off-farm activities. Children are working 4 hours per day on the farm, their availability is

restricted to the school holidays. Adults above 55 years are working not more than 4 hours per

day on the farm.

The priority of the farm households across all household groups is to satisfy the basic needs of

the household members, especially the food requirements. Most crops are used at least partly

for own consumption, the surplus is sold. Only coffee, fruits and vegetables can be considered

as traditional cash crops. The seasonal nutrient requirements of the household members are

introduced as a constraint in the programming matrix. Average estimates for each type of

consumer unit were worked out to obtain seasonal consumption requirements of the

households. The differences due to age, sex , and activity level are also captured (Latham,

1997). To reflect consumption priorities, the amounts of different crops consumed are

considered as well.

Another major constraint of the farm households in the study area is the amount of capital

available for agricultural production. Two types of capital requirements are differentiated:

capital required in the first year for investment activities, and capital required for annual

operational production costs. To calculate the available amount for the first type, investment

costs of investments still used were added up and to the sum 10 % were added as a tolerance

29

level. For the latter type annual production costs were calculated and again 10 % were added

as a tolerance level. Of course, these capital constraints could be relaxed by credit activities in

the scenarios.

Programming models for each identified household type (subsistence farm households, semi-

subsistence farm households, trial farm households and commercial farm households) were

calibrated. Model validation was conducted through measuring the association of model

solutions with observed values as suggested by Mc Carl and Apland (1986). The model

results were regressed on observed values, where a perfect association would be indicated by

an intercept of 0 and a slope of 1. The received values for R2 are 0.95. 0.99, 0.89, 0.94, the

received values for the intercept are: 0.04, -0.01, -0.03 and 0.06, the received slope values are:

0.96, 1.01, 1.02 and 0.83. Therefore, the “goodness-of-fit” test is indicating sufficient

association between real world data and model results. The programming models were solved

using the software Premium Solver Platform V3.5 for Excel (Frontline Systems, Inc.).

In the following first simulation results on the impact of different soil classes on land

productivity and labor intensity will be discussed. The four households selected for

representing the four identified clusters are having their plots on soil class 1, where nearly

50% of the samples included in the soil clustering belong to. The following two diagrams

(Figure 10 and 11) reflect the simulation results for land productivity and labor intensity

under the current land use practices for the four different household types.

The highest productivity is achieved on soil class 2 by all household types, followed by soil

classes1 and 3. The productivities on soil classes 4 to 6 are comparatively low. On soil 6 the

semi-subsistence farm household for instance just achieved 25 % of the productivity achieved

on soil 1. The highest productivity on each soil class is indicated for the commercial

household, followed by the trial farm household. The major factor contributing to higher

productivities of these farm household types is the cultivation of improved seeds and the use

of agrochemicals. On the first three soil classes the semi-subsistence household achieved

higher productivities than the subsistence farm households. On soil classes 3-6 both

household types had to reduce their consumption preferences, the subsistence household more

drastically than the semi-subsistence. On soil class 6 the subsistence farm household is no

longer able to satisfy the calorie requirements of its members. The recommended amount has

to be reduced to 70 %, which has serious consequences for food security.

30

Figure 10: Land Productivity (Gross Output/ ha/ year)

0

200

400

600

800

1000

1200

1400

1600

soil1 soil2 soil3 soil4 soil5 soil6

productiv 10^3 Ush

SubSemiTrialCom

Source: Bio-economic model output

Figure 11: Labor Intensity

0

500

1000

1500

2000

2500

Sub Semi Trial Com

h/h

a an

d y

ear

soil1

soil2

soil3

soil4

soil5

soil6

Source: Bio-economic model output

Figure 11 reveals the labor intensity for the different household types on different soil classes.

The labor intensity of the subsistence household is increasing on soils 4-6. These soil classes

are less productive leading to lower crop yields. To satisfy the household consumption

preferences a certain amount of specific crops have to be produced. To achieve the desired

output level of one crop, the area under this crop has to be expanded on less productive soils.

This leads to changing production patterns, since consequently the area under other crops will

have to be reduced. If the production activities, which are expanded, are more labor intensive

than the ones which are reduced, consequently overall labour intensity increases. In the case

of the subsistence farm household the food crop production activities are comparatively more

labour intensive. Therefore, the necessary expansion of their production on less productive

31

soils contributes to increasing labor intensities on soils 4-6. Similar conclusions can be drawn

for the semi-subsistence household, though labor intensity is not increasing as significant as

for the subsistence farm household. For the trial farm household and commercial farm

household the constraints of satisfying consumption preferences are not as binding on most

soil classes as for the household types discussed before due to higher overall productivity.

These household types sell a larger proportion of their produced goods. The development of

labor intensity on soils 1-4 follows standard economic theory of factor allocation. Marginal

productivity is increasing and consequently labor intensity is increasing as well and vice

versa. On soils 5 and 6 land productivity is so low that the goal of farm households of

satisfying basic needs is determining the production structure and therefore labor allocation.

Only a minor proportion of produced goods is sold. The consequences are increasing labor

intensities although land productivities are decreasing.

4. Agricultural Policy Scenarios

The selection of relevant policy scenarios is based on the “Priority Areas for Action” defined

in the Plan for Modernization of Agriculture (PMA) published by the Ugandan Government

in 2000. The priority areas for action are: improving access to rural finance, improving access

to markets, research and technology development, sustainable natural resource utilization, and

management and education for agriculture. The PMA was composed in line with the economy

wide decentralization, privatization and liberalization policy in Uganda by redefining the roles

of the private sector and the central and local governments. The role of the public sector in

agriculture should be reduced substantially, focusing mainly on research, extension, and

regulatory functions. The Government is committed not to be involved in the import or

distribution of agricultural inputs, nor to subsidize inputs or to be involved in direct provision

of rural financial services. The Government “would like to see all activities connected with

agricultural production, processing, trading, supply of inputs, exports and imports be carried

out entirely by the private sector” (FAO, 1999). Therefore, the selected policy scenarios

neglect direct market interventions concerning output price policies (e.g. taxes, subsidies,

fixed prices), input policies (e.g. subsidies, input delivery systems), marketing policies (e.g.

monopoly parastatals, trader licensing) and credit policies (e.g. credit provision by the state).

The focal point of the model scenarios is rather to which extent and how the market

environment could be improved to provide sufficient incentives to reach growth and

sustainability goals simultaneously.

32

The objectives of the scenarios, run with the developed integrated bio-economic model

approach, can be divided into four steps:

1. Explore the feasibility of reaching socio-economic goals of farm households

(satisfying basic needs of its members and income maximization) and sustainability

goals (non-negative nutrient balances) simultaneously under the current constraints.

2. Explore whether the relaxation of technical constraints (introduction of new

technologies) and capital constraints (provision of credit) could contribute to achieve

the goals defined above.

3. Illustrate to which extent improvements of the market environment (reflected by

reduction of distortion of input and output prices) in combination with provision of

credit and promotion of alternative forms of labor acquisition could contribute to a

significant increase of household welfare and nutrient balances. Furthermore, the

potential means leading to a reduction of market imperfections will be discussed.

4. Illustrate the economic and ecological impacts of the intensive production of high

value crops, since this pathway could be identified as a potential profitable pathway in

a development domain with high agricultural potential, high market access and high

population density. Furthermore, the transformation of subsistence to commercial

farming was mentioned as one major objective in the PMA.

Each single objective has to be seen in the context of the research question, why agricultural

production in a region belonging to a development pathway of high population density, high

market access and high agricultural potential is predominantly characterized by a low

intensity system.

Scenarios #1:Binding constraints and feasibility of non-negative nutrient balances

In the following it is discussed for the identified household types whether, under the current

constraints, the socio-economic goals of farm households to satisfy basic needs of its

members and to maximize the income and the sustainability goals of non-negative nutrient

balances could be reached simultaneously. Additionally, it is explored whether the relaxation

of technical constraints (by introduction of new technologies promoted by CIAT) and capital

constraints (by provision of credit) could contribute to reach this goal. For this purpose the

developed bio-economic model is used to run different scenarios for all household types (the

scenarios a - f are defined in tables 3 to 6). The technologies included are the ones captured in

the ANN yield estimator. The objective function value of the model (Total Gross Margin,

33

TGM) is used as an indicator for household welfare and the nutrient balances as sustainability

indicator.

Under current constraints the subsistence farm household type can satisfy the basic household

needs on the first three soil classes associated with high negative nutrient balances (scenario

a). As expected from previous results the highest Total Gross Margin (TGM) is achieved on

soil class 2. On soil classes 4 – 6 no feasible solution can be attained since the consumption

requirements cannot be fulfilled on the less productive soils. On soil classes 4 and 5 the

consumption preferences have to be reduced partially and on soil class 6 the recommended

nutrient requirements of the household members have to be reduced to 70 % of the original

value. The introduction of the sustainability constraint of non-negative nutrient balances

(scenario b) leads to no feasible solutions indicating that this sustainability goal cannot be

achieved with the current land management practices.

Tab. 3: Subsistence Farm Household Type: Feasibility of socio-economic and sustainability goals

Scenario Soil 1

Soil 2 Soil 3 Soil 4 Soil 5 Soil 6

a. current constraints TGM (103 USh) N-Balance (kg/ha) P-Balance (kg/ha) K-Balance (kg/ha)

1299 -28 -8 -39

1742 -37 -9 -54

1193 -28 -9 -40

# not feasible

# not feasible

# not feasible

b. + sustainability constraint TGM (103 USh) N-Balance (kg/ha) P-Balance (kg/ha) K-Balance (kg/ha)

# not feasible

# not feasible

# not feasible

# not feasible

# not feasible

# not feasible

c. + new technologies TGM (103 USh) N-Balance (kg/ha) P-Balance (kg/ha) K-Balance (kg/ha)

1310 -29 -6 -40

1776 -40 -2 -57

1194 -28 -8 -40

# not feasible

# not feasible

# not feasible

d. +sustainability constraint +new technologies

TGM (103 USh) N-Balance (kg/ha) P-Balance (kg/ha) K-Balance (kg/ha)

# not feasible

# not feasible

# not feasible

# not feasible

# not feasible

# not feasible

e. +sustainability constraint +new technologies +credit


# not feasible

# not feasible

# not feasible

# not feasible

# not feasible

# not feasible

f. +new technologies +credit


1356 -39 +12 -51

1822 -60 +35 -77

1195 -29 -8 -41

# not feasible

# not feasible

# not feasible

Source: own calculations (model results)

34

The introduction of new technologies promoted by CIAT (scenario c) leads just to a slight

increase of the TGM (e.g. + 0.8 % on soil 1) with nearly no improvement of the nutrient

balances, since the adoption of these technologies is not profitable under current market

conditions. Therefore, the introduction of new technologies alone cannot lead to the

achievement of the sustainability goal (scenario d). Even with the additional provision of

credit (scenario e), the subsistence farm household type is not able to achieve non-negative

nutrient balances. The last scenario type is neglecting the sustainability constraint and is

focusing on the introduction of new technologies coupled with the provision of credit.

Comparing the results with the baseline scenario a, TGM is increasing by 4 % on soil 1, but

the nutrient balances for nitrogen and potassium are becoming more negative (-39 % and –31

% respectively). Profitable adoption of rock phosphate contributes to increasing P-balances.

Tab. 4: Semi-Subsistence Farm Household Type: Feasibility of socio-economic and sustainability goals


Scenario Soil 1



1490 -52 -12 -62

1886 -65 -14 -82

1405 -51 -12 -63

# not feasible

# not feasible

# not feasible


# not feasible

# not feasible

# not feasible

# not feasible

# not feasible

# not feasible


1524 -66 +24 -83

1886 -65 -14 -82

1405 -51 -12 -63

# not feasible

# not feasible

# not feasible



# not feasible

# not feasible

# not feasible

# not feasible

# not feasible

# not feasible



# not feasible

1730 +8 +69 0

1130 +6 +56 0

# not feasible

# not feasible

# not feasible



1633 -79 +59 -94

1945 -64 -8 -78

1413 -48 -3 -59

# not feasible

# not feasible

# not feasible

35

The scenario results for the semi-subsistence farm household type are in some aspects

different to the ones for the subsistence farm household type. Under the current constraints

(scenario a) the nutrient requirements of the household members can be satisfied on each soil

class, although the consumption preferences have to be reduced partially on soils 4 – 6.

Again, high negative nutrient balances on each soil class can be observed. With current land

management practices non-negative nutrient balances cannot be achieved (scenario b). The

introduction of new technologies (scenario c) is leading to a slight increase of the TGM (2 %)

on soil 1 associated with higher negative nutrient balances (nutrient balance for nitrogen is

decreasing by 27 %). The new technologies alone cannot contribute to the achievement of the

sustainability goal (scenario d), whereas the combined introduction of new technologies and

provision of credit (scenario e) can reach this goal on soil 2 and 3. The introduction of new

technologies and provision of credit while neglecting the sustainability constraint (scenario f)

is leading to an increase of the TGM by 12 % on soil 2 and 25 % on soil 3, while the nitrogen

balance is decreasing from + 8 kg/ha to –64 kg/ha on soil 2 and from +6 kg/ha to –48 kg/ha.

These adverse effects can be interpreted as an indicator for a trade-off between economic and

ecological goals.

Just as the semi-subsistence farm household type, the trial farm household type can fulfil the

nutrient requirements of its members on each soil class. Table 5 indicates very high negative

nutrient balances in the baseline scenario a. The consumption preferences cannot be fulfilled

on soils 4 – 6, but a minor reduction of preferences could contribute to a feasible solution

(scenario a). As discussed for the two household types before, the defined sustainability goal

of non-negative nutrient balances cannot be achieved with current land management practices

and the introduction of new technologies alone (scenarios b – d). Scenario c indicates that the

introduction of new technologies alone cannot increase TGM significantly. Only on soil 1 the

phosphorus balance is improving due to the adoption of rock phosphate. At the same time the

balances for N and K are declining. With the simultaneous introduction of new technologies,

provision of credit, and sustainability constraints, the only feasible solution is achieved on soil

3 (scenario e). Neglecting the sustainability constraint (scenario f) is leading to an increase of

TGM on soil 3 by 39 %, but the nutrient balances are decreasing again (for example nitrogen

from +8 kg/ha to –41 kg/ha).

36

Tab. 5: Trial Farm Household Type: Feasibility of socio-economic and sustainability goals


The commercial farm household can fulfil the nutrient requirements and consumption

preferences of its members on each soil type in scenario a. The nutrient balances indicate very

high negative values. As for the other three household types, the sustainability goal cannot be

achieved with current land management practices (scenario b). The introduction of new

technologies (scenario c) is contributing to a slight increase of the TGM on soils 1, 2, 5 and 6

(e.g. on soil 1 by 8 %), whereas, the nutrient balances for nitrogen and potassium are

decreasing due higher nutrient losses through harvested products. The adoption of rock

phosphate is leading to a significant improvement of the phosphorus balance (on soil 2 for

example from – 20 kg/ha to +18 kg/ha).9 The sustainability goal can be reached on soil

9 Negative environmental impacts of high positive nutrient balances (eutrophication) have be taken into account.

Scenario Soil 1



2395 -43 -11 -47

2907 -71 -17 -79

2306 -41 -11 -47

# not feasible

# not feasible

# not feasible


# not feasible

# not feasible

# not feasible

# not feasible

# not feasible

# not feasible

c. +new technologies TGM (103 USh) N-Balance (kg/ha) P-Balance (kg/ha) K-Balance (kg/ha)

2452 -60 +21 -69

2915 -75 -8 -84

2306 -41 -11 -47

# not feasible

# not feasible

# not feasible



# not feasible

# not feasible

# not feasible

# not feasible

# not feasible

# not feasible



# not feasible

# not feasible

1735 +8 +50 0

# not feasible

# not feasible

# not feasible



2575 -69 +35 -81

3024 -77 -5 -86

2418 -41 -10 -47

# not feasible

# not feasible

# not feasible

37

classes 2 –6 by introducing the new technologies (scenario d). Comparing the results of the

technology scenarios with and without the sustainability constraint (scenarios d + c) reveals

again a trade-off between economic and ecological goals. Neglecting the sustainability

constraint is leading to an increase of the TGM by 26 % on soil 2 and an increase of the

nitrogen balance from –83 kg/ha to 0 kg/ha. Similar trade-offs could be observed when

comparing the scenarios where new technologies are introduced and credits are provided; in

one case with, in the other without the sustainability constraint (scenarios e + f).

Tab. 6: Commercial Farm Household Type: Feasibility of socio-economic and sustainability goals

Source: own calculation (model results)

To sum up, the results of the scenarios reveal difficulties in achieving the goal of non-

negative nutrient balances, especially for the subsistence, semi-subsistence, and trial farm

household types. Current land management practices are leading to high negative nutrient

Scenario Soil 1



4800 -77 -15 -71

6108 -110 -20 -111

4555 -74 -14 -72

3387 -27 -6 -20

3298 -26 -6 -20

2393 -26 -5 -18


# not feasible

# not feasible

# not feasible

# not feasible

# not feasible

# not feasible


5204 -96 +58 -100

6157 -83 +18 -81

4555 -75 -14 -72

3387 -26 -2 -22

3307 -25 -1 -18

2402 -22 -4 -15



# not feasible

4882 0 +54 0

3739 0 +26 0

2773 0 +17 0

2887 0 +15 0

1079 0 +16 0



3373 0 +47 0

5316 0 +65 0

3794 0 +53 0

2885 0 +17 0

2957 0 +16 0

1764 0 +16 0



5223 -104 +56 -106

6199 -67 +51 -58

4555 -74 -14 -73

3387 -27 -6 -20

3307 -25 -1 -18

2402 -22 -4 -15

38

balances. Due to higher yields and consequently higher nutrient losses through harvested

products and stover, the value of the nutrient balances for the commercial farm household

type is higher negative than for example the nutrient balance value for the subsistence farm

household type. In most cases, neither the relaxation of technical constraints (by introduction

of new technologies), nor the relaxation of capital constraints (by the provision of credit)

could contribute to reach this sustainability goal. The main reason for this result is the non-

profitability of the promoted technologies under current market conditions resulting in low

adoption. Consumption preferences have to be reduced partially for the first three household

types on less productive soils. The subsistence farm households could not even satisfy the

recommended nutrient requirements of its household members on one soil class. For the

commercial farm household it is feasible to fulfil non-negative nutrient balances with the

introduction of new technologies in most cases, and with the additional provision of credit in

any case. This household type faced no problems to satisfy the basic needs of its household

members. Moreover, the scenarios illustrated a trade-off between the goals of improving

household welfare and of achieving ecological goals under the current market environment.

Since the sustainability goal of non-negative nutrient balances could not be reached in many

cases and direct market intervention were excluded as suitable policy measures for Uganda,

the subsequent scenarios deal with the potential of market improvements to contribute to a

significant increase of household welfare and nutrient balances simultaneously.

Scenarios #2: Economic and ecological impacts of promoted technologies under market

improvements

Before the scenario results will be discussed, the market environment in Uganda and in the

study region will be described, the potential extent to which input and output prices could

change will be examined, and potential policy measures to reach these changes will be

discussed.

As already indicated before the market environment in Iganga District, as in most parts of

Uganda, is far from being perfect. High transaction costs, high transportation costs, and

imperfect competition are leading to a low level of output prices and unreasonably high input

prices. The marketing chain for agricultural products in Iganga District involves middle men

in the villages, local buyers in trading centers and Iganga town, and traders from Kenya and

Mbale, Busia and Kampala (see figure 12). The price offered to the farm households by the

middle men depends on the price set by the local buyers in the town or trading centers, which

in turn is determined by the price offered by the foreign buyers. A study carried out by

39

Vredeseilanden-Coopibo-Uganda (1998) indicated a mark-up of 60 % between the price

received by the farmer and the price retailers were offering for Iganga District. Own survey

data indicated even higher price differences between farmer, wholesaler and retailer. In 2001

the price mark-up of maize between farmer and wholesaler was 62 %, and between farmer

and retailer 212 % (see figure 13). Farmers do not have the bargaining power when selling

their products at the farm gate, because in many cases they do not know the price offered at

other levels of the marketing chain.

Figure 12: Marketing chain for agricultural produce in Iganga District

Source: Vredeseilanden-Coopibo-Uganda (1998)

LOCAL BUYERS IN TRADING

CENTRES AND IGANGA TOWN

TRADERS IN

KENYA

TRADERS IN

MBALE; BUSIA;

KAMPALA

FARMERS

MIDDLE MEN IN VILLAGES

40

Figure 13: Price mark-ups of maize in the marketing chain 1998-2002

0%50%

100%150%200%

250%300%350%400%450%

1998 1999 2000 2001 2002

price mark-up farmer -wholesaler [%]

price mark-up farmer -retailer [%]

Source: own survey data (interview with Iganga District Cooperative and Marketing Officer)

These illustrations reveal the impact reduced transaction costs could have on the increase of

agricultural product prices received by the farmers. An essential step would be to improve

market transparency by providing farm households relevant agricultural information. One

option would be to implement a market information system (MIS) as suggested by IFDC for

the input markets (IFDC 1999). This MIS could collect, analyze, and publish information on

the development of output prices, input prices and other relevant information. The next step

would be the identification of appropriate channels to spread this information to the farm

households (see figure 14). Two possible channels could be taken into account: 1) mass media

(especially radio) and 2) extension workers via opinion leaders within the villages acting as

channel agents. Mass media (radio) could either spread the information direct to the

households or via opinion leader. 66 % of the surveyed households reported that they own a

radio, whereas other potential information sources like newspapers or telephone are only used

by 15 % and 6 % respectively. Of course these information sources could be important in the

medium to long run, but in the short run radio seems to be an appropriate mean to reach rural

households. On the other hand the issue of maintaining the radios (e.g. batteries), and the

perception and usage of the radio as an important information source have to be taken into

account as well. The latter aspect could be subject to public campaigns and part of the work of

extension services.

41

Figure 14: Agricultural Information Network

The second potential channel is based on social interactions between farm households. Above

it was mentioned that relatives and friends are very important information sources for the

adoption of new technologies. Analyses carried out by CIAT in the study region confirmed

that social interactions played a key role for the diffusion of innovation. Based on these

findings, extension agents could spread relevant agricultural information via identified

opinion leaders. Certain opinion leaders are of special importance, taking into account that 50

% of the surveyed households had at least one person belonging to the group of trial farm

households in their communication network with whom they discuss issues related to

agriculture.10 The identification of opinion leaders could be a task carried out by extension

agents as well.11 Since restructuring the MAAIF involved a drastic reduction in the number of

staff, the selection of opinion leaders for spreading information to individual farm households

seems to be a cost-effective and promising strategy. One critical aspect following this strategy

is certainly the identification of appropriate opinion leaders providing information for all

social and religious groups.

Inefficiency in procurement, high transportation costs, and absence of competitive pressure

are leading to unreasonably high input prices, especially fertilizer prices (IFDC, 1999). The

fertilizer market in Uganda is in a very early stage of development. The total fertilizer use is

estimated to be approximately 12000 tons in 1999. There are only four fertilizer

importers/wholesalers. The linkages with traders in Kenya, a natural place for importing

10 Which is an enormous proportion if one takes into account that trial farm households account only for 7 % of all households in both villages.

Market

Information

System

Mass Media

(Radio)

Opinion Leader

(Change Agent)

Household

Household

Household Extension

Agent

42

fertilizers in Uganda, are surprisingly few. Since the market liberalization policy was

implemented, government policy is to remain the import of fertilizer entirely in the private

sector. Figure 15 illustrates the potential levels to which fertilizer prices could be reduced by

improving the market environment and marketing chain. In the Soil Fertility Initiative

Concept Paper by FAO (1999), it was reported that at the end of 1998 the average price for

fertilizer landed in Mombasa (Kenya) was US$ 250 per ton and freight to Kampala was about

100 US$ per ton. Further US$ 50 were added due to clearance at the border, trans-loading,

storage and import charges. Therefore, the total cost c.i.f. Kampala was about 400 US$, which

is very high in comparison to Kenya and other neighbouring countries. It is estimated that

c.i.f. price Kampala would fall by a quarter only by rising the volumes to levels, which would

justify shiploads and trainloads (FAO, 1999). The majority of the fertilizer is delivered to

stockists in 50 kg bags. The fertilizers are repacked into smaller units of 5 kg and 1 kg leading

to a price increase of 100 %. Combining both effects (economies of scale in transportation

and neglecting the costs of repacking the fertilizers) could optimally result in a fertilizer price,

which is 37.5 % of the current prices. A further input price decrease could be attained through

increased competition on the fertilizer market. The FAO (1999) compared the price build-up

for small fertilizer retailers with the price build-up for small sugar or soap powder retailers,

both supplied by stockists. The market for sugar or washing powder is a more mature and

higher volume market in Uganda. A comparison of the price build-ups shows that fertilizer

prices could additionally decrease when the competition and the volumes on the fertilizer

market would increase.12 Taking into account the high transportation costs and the high mark-

up of retailers, there is a huge potential to reduce the fertilizer price substantially. An

alternative P-fertilizer is Busumbu Rock-Phosphate with deposits near Tororo. The P-content

is low in comparison to TSP, but the price should also be far lower. The “International Mining

and Development Ltd.”, a Canadian based company, is at the exploratory stage of a

commercial investment in Busumbu.13

11 For the identification of opinion leaders extension agents could collaborate with experienced NGO staff or researchers, who applied similar approaches (as for example CIAT/A2N in the study region). 12 The impact of increased competition on the fertilizer price cannot be quantified exactly. 13 Rock phosphate is not available on the fertilizer market in Uganda yet. Due to uncertainty about its availability in the future rock phosphate is not included in the scenarios if not explicitly mentioned.

43

100% 75

% 50% 37,5

% ??%

0%

25%

50%

75%

100%

% of current input price

Scenario1 Scenario2 Scenario3 Scenario4 Scenario5

Figure 15: Potential levels of fertilizer price reduction

minimum input price achievable [%] maximum input price reduction achievable [%]

Source: based on FAO (1999)

Scenarios #2.1: Economic and ecological impact of decreasing fertilizer prices

This type of scenarios focuses on the economic and ecological impact of the stepwise

reduction of fertilizer prices. Sensitivity analysis was chosen to reflect at which critical

parameter values (fertilizer prices) a significant improvement of nutrient balances could be

achieved for the different household types. The price changes assumed in the scenarios should

be compared with the potential price reductions discussed above. It should be emphasized

here that the following analyses refer to soil class 1 only and that the new technology option

are just available for maize. The plots of the four representative households belong to this soil

cluster, as well as nearly 50 % of the soil samples included in the soil clustering. Figures 8

and 10 reveal that the soil productivity and impact of the technologies on yield are different

for the other soil classes.

Scenario 1 Scenario 2 Scenario 3 Scenario 4 Scenario 5

Potentials for input price reduction

• none (current situation)

• economies of scale transport

• marketing efficiency





• increased competition

Min. input price achievable (% of current price)

100 %

75 %

50 %

37,5 %

??

44

Figure 16: Subsistence Farm Household: Sensitivity Analysis Fertilizer Price Reduction

-50

0

50

100

150

100 75 50 40 37,5 30 20 10 5 0

% of current price

TGM 10^4 Ush

kg/ha NPK

TGM NPK

Source: model results

The diagrams for each household type (Figures 16 –19) illustrate the development of the

TGM as an economic indicator and the nutrient balances as a sustainability indicator with

stepwise reduced fertilizer prices. The tables below the diagrams indicate the development of

reduced costs for selected farming activities, and the area on which certain fertilizers were

adopted. The reduced costs are an indicator for the profitability of farming activities. They

reveal by which sum the costs of an economic activity have to be reduced before it enters the

optimal solution, meaning before the household adopts this activity. Figure 16 illustrates that

the subsistence farm household is far from adopting fertilizers at the current prices, e.g. the

reduced costs for maize with NPK-fertilizer application is 313*103 USh (season 1). In the

following the reduced costs are further decreasing with reduced fertilizer prices. When a price

reduction of 37.5 % of the current price is reached, the subsistence farm household starts to

adopt NP fertilizers on 0.02 ha. The increase of the TGM and the improvement of nutrient

balances are negligible. The fertilizer prices have to be reduced up to 10 % of the current

prices before NPK is adopted on 0.07 ha. Assuming that fertilizers would be available for

14 Abbreviations: Maize+N1=nitrogen fertilizer application on maize in season 1 etc.

% of current price

100 75 50 40 37.5 30 20 10 5 0

Reduced costs (103 USh)14 Maize+N1 Maize+NP1 Maize+NPK1 Maize+N2 Maize+NP2 Maize+NPK2

182 211 313 188 252 364

139 124 193 144 165 244

95 37 73

101 79

123

78 2

25 83 44 75

78 0

19 79 35 63

84 0

11 66 9

27

85 0 2

74 0 5

89 4 0

89 4 0

97 7 0

97 7 0

131 22 0

131 22 0

Area adopted (ha)

0 0 0 0 NP1 0.02

NP1 0.02

NP1 0.02 NP2 0.03

NPK1 0.03

NPK2 0.04

NPK1 0.05

NPK2 0.06

NPK1 0.14

NPK2 0.14

45

free, the subsistence household could adopt NPK on 0.28 ha, leading to a positive nutrient

balance for phosphorus (+2 kg/ha). The balances for nitrogen and potassium are just

improving slightly in comparison to the current price situation, from –29 kg/ha to –22 kg/ha

and from –40 kg/ha to –33 kg/ha respectively. Figure 16 shows that the reduction of fertilizer

prices has nearly no impact on the TGM. Free available fertilizers would increase this

household welfare indicator by only 2 % in comparison to a situation with current prices.

Figure 17: Semi-Subsistence Farm Household: Sensitivity Analysis Fertilizer Price Reduction

-100

-50

0

50

100

150

200

100 75 50 40 37,5 30 20 10 5 0

% of current price

TGM 10^4 Ush

kg/ha NPK

TGMNPK


The semi-subsistence farm household starts to adopt NP (0.08 ha) not until 10 % of current

fertilizer prices are reached (see figure 17). The reduced costs indicate that the

competitiveness of farming activities, including the adoption of fertilizers, is very low due to

non-profitability until this tremendous price reduction is attained. When the fertilizer price

amounts to 5% of the current price, positive nutrient balances for nitrogen (+16 kg/ha) and

phosphorus (+76 kg/ha) could be achieved. The balance for potassium would still be negative,

but the high negative balance of –63 kg/ha could be reduced to –8 kg/ha. Under this price

scenario NPK-fertilizer could be adopted on 1.59 ha and NP on 0.13 ha. The TGM increased

% of current price

100 75 50 40 37.5 30 20 10 5 0

Reduced costs (103 USh) Maize+N1 Maize+NP1 Maize+NPK1 Maize+N2 Maize+NP2 Maize+NPK2

505 833

1154 464 792

1114

392 606 841 351 566 801

279 380 528 238 340 488

234 290 403 193 249 363

222 267 372 182 227 331

189 200 288 148 159 237

143 109 153 103 69

112

72 3

20 69 0

17

118 12 0

94 0 0

151 26 0

127 14 0

Area adopted (ha)

0 0 0 0 0 0 0 NP2 0.08

NPK1 0.76 NP2 0.13

NPK2 0.83

NPK1 0.76

NPK2 0.96

46

by 5.5 % in comparison to the baseline scenario. A fertilizer price of 0 would lead to a further

slight increase of the TGM, the nutrient balances for nitrogen and phosphorus would be

positive, the nutrient balances for potassium would be slightly negative (-2 kg/ha). The

overall impact of fertilizer price reduction on TGM of the semi-subsistence farm household is

very modest again.

Figure 18: Trial Farm Household: Sensitivity Analysis Fertilizer Price Reduction

-60

-40

-20

0

20

40

60

100 75 50 40 37,5 30 20 15 10 5 0

% of current price

TGM 10^5 Ush

kg/ha NPK

TGMNPK


The fertilizer prices have to decrease to 15 % of the original value before the trial farm

household starts to adopt NPK-fertilizer on 0.13 ha (see figure 18). The increase of TGM and

the improvements of the nutrient balances are negligible. With 10 % and 5 % of the original

prices the adoption of NPK-fertilizer could increase to 0.25 ha and 1.21 ha respectively. The

latter price decrease could lead to substantial improvements of nutrient balances. In

comparison to the baseline scenario with current prices, the balance for nitrogen could

increase from –43 kg/ha to –7 kg/ha, the balances for phosphorus could increase from –11

kg/ha to +46 kg/ha and for potassium from –47 kg/ha to –18 kg/ha. The increase of the TGM

would be modest (1 %). Free available fertilizers would not provide sufficient incentives to

expand the area on which fertilizer are applied in comparison to a price decrease to 5 % of the

original price. Only the TGM would increase from 2420*103 USh to 2449*103 USh.

15 Reduced costs are not presented since the applied software does not provide a sensitivity report when mixed integer programming is used.

% of current price15

100 75 50 40 37.5 30 20 15 10 5 0

Area adopted(ha)

0 0 0 0 0 0 0 NPK1 0.13

NKP1 0.13

NPK2 0.12

NPK1 0.61

NPK2 0.6

NPK1 0.61

NPK2 0.6

47

Figure 19: Commercial Farm Household: Sensitivity Analysis Fertilizer Price Reduction

-100-80-60-40-20

020406080

100

100 75 50 40 37,5 30 20 15 10 5 0

% of current price

TGM 10^5 Ush

kg/ha NPK

TGMNPK


The commercial farm household type adopts NP-fertilizer on 0.62 ha when the input prices

decrease to 30 % of the current value (see figure 19). The phosphorus balance would increase

from –15 kg/ha to 0 kg/ha, the nitrogen balance would increase slightly, whereas the

potassium balance would decrease slightly. The next significant change could occur with a

fertilizer price decrease to 20 % of the current price. The adoption of NP-fertilizer could

increase to 2.68 ha, leading to a highly positive phosphorus balance (47 kg/ha) and an

improvement of the nitrogen balance from –77 kg/ha in the baseline scenario to –21 kg/ha.

For the potassium balance a modest decrease could be observed from –71 kg/ha to –81 kg/ha.

A further fertilizer price decrease to 10 % of the current value would lead to an adoption pf

NPK-fertilizer on 1.84 ha and NP-fertilizer on 1.66 ha. This scenario would result in an

increase of the TGM of 3.6 % and N-, P-, K-balances of –11 kg/ha, +65 kg/ha and –48 kg/ha

respectively. NPK-fertilizer adoption would jump up to 4.03 ha with a price decrease to 5 %

of the original value. The result of this adoption would be a significant improvement of the

% of current price

100 75 50 40 37.5 30 20 15 10 5 0


213 255 368 189 238 346

170 168 248 143 149 224

127 81

128 100 62

103

109 47 80 83 28 55

105 38 68 78 19 43

95 16 37 71 0

16

95 0 9

79 0

11

105 0 1

87 0 6

116 4 0

94 0 1

133 11 0

110 7 0

149 17 0

127 14 0

Area adopted (ha)

0 0 0 0 0 NP1 0.62

NP1 1.58 NP2

1.1

NP1 1.76 NP2 1.36

NPK1 1.84 NP2 1.66

NPK1 1.84

NPK2 2.19

NPK1 1.84

NPK2 2.19

48

potassium balance (-8 kg/ha). Assuming that the fertilizers would be available for free, would

not change the fertilizer adoption and nutrient balances in comparison to the latest scenario.

The TGM would increase by 7.7 % compared to the baseline scenario, the increase of the

TGM with 5 % of the current fertilizer prices would be 5.7 %.

To summarize, the sensitivity analysis of fertilizer price reduction illustrated that a significant

adoption of fertilizers contributing to a substantial increase of the nutrient balances could be

achieved only with a tremendous reduction of the current fertilizer prices. High reduced costs

in the baseline scenarios revealed that each household type is far from adopting new

technologies, because of their non-profitability in the current situation. The reduced costs are

too high and thus only slight decreases of current fertilizer prices would probably not lead to

an adoption of improved practices. The prices have to be reduced at least to 37.5 % of the

current value before adoption starts. A significant adoption cannot be expected before a

decrease of 10 % - 5 % of the current price is reached. This is very difficult to achieve with

policy measures aiming at reduction of transaction costs and improvements of the market

environment only (as discussed above). At the same only a very modest increase of the TGM

could be observed for each household type. Therefore, policy options focusing only on the

reduction of fertilizer prices would probably not be a promising strategy targeted on a

sustainable intensification of agriculture.

Scenarios #2.2: Economic and ecological impact of increasing agricultural output prices

The next type of scenarios focuses on the impact of stepwise increased agricultural product

prices on household welfare (TGM), production structure, and sustainability indicator

(nutrient balances). The price scenarios used in this sensitivity analysis should be compared

with the results of the discussion about the potentials for an increase of output prices. Figure

20 shows that for the subsistence farm household type nutrient balances for nitrogen,

phosphorus and potassium are nearly constant with increasing prices, even up to 100 % of the

current values. The main reasons are that there is no adoption of new fertilizer technologies

and that the production structure is stable due to the low market orientation of the household

type under consideration. Although the output prices are increasing, the reduced costs (see

appendix A 10) indicate that the subsistence household is far from adopting farming activities

involving the application of new fertilizer. The reduced costs for NPK-fertilizer for example

are decreasing from 313*103 USh in the baseline scenario to 246*103 USh with an output

price increase of 70 %, before the costs increase again. Changing output prices are leading to

a change of the relative competitiveness of farming activities. Therefore, the production

49

structure is changing with a price increase of 80 %, where intercropped coffee and bananas

are reduced from 0.76 ha to 0.65 ha and the area under improved maize increases from 0.07

ha to 0.28 ha. The cultivation of improved maize is of special interest in this context.

Regarding the adoption of this improved variety the model results are different from the

observed values. The households reported that only local maize is cultivated, whereas the

model indicates an adoption of an improved maize variety on 0.07 ha in two seasons.

Anyway, the model indicates that the adoption of the improved variety would become more

profitable with an output price increase of 80 %.

Figure 20: Subsistence Farm Household: Sensitivity Analysis Output Prices

-100

-50

0

50

100

150

200

0 10 20 30 40 50 60 70 80 90 100

output price increase %

TGM 10^4 Ush

kg/ha NPK

TGMNPK

Abbreviations: Maize l/Beans l=local maize variety intercropped with local beans variety; Maize i: improved maize variety; Sweet Pot Bu=sweet potatoes (variety Bunduguza); Maize l/Gnuts=local maize variety intercropped with groundnuts; Coffee l/Ban=local coffee variety intercropped with cooking bananas. % of outprice increase

0 10 20 30 40 50 60 70 80 90 100

Production structure (ha) Maize l/Beans Maize i Sweet Pot Bu Cassava Maize l/Gnuts Coffee l/Ban

1.2 0.07 0.85 0.1 0.05 0.76

1.2 0.07 0.85 0.1 0.05 0.76

1.2 0.07 0.85 0.1 0.05 0.76

1.2 0.07 0.85 0.1 0.05 0.76

1.2 0.07 0.85 0.1 0.05 0.76

1.2 0.07 0.85 0.1 0.05 0.76

1.2 0.07 0.85 0.1 0.05 0.76

1.2 0.07 0.85 0.1 0.05 0.76

1.2 0.28 0.85 0.1 0.05 0.65

1.2 0.28 0.85 0.1 0.05 0.65

1.2 0.28 0.85 0.1 0.05 0.65


Increasing the agricultural output prices by 50 % and 100 % could lead to a raise of TGM by

11.5 % and 23 % respectively. The inclusion of rock phosphate would lead to an adoption of

0.07 ha Busumbu Rock Phosphate and therefore to a slight improvement of the phosphorus

balance (from –8kg/ha to –6 kg/ha). A further output price increase would not provide

sufficient incentives for expanding the area under rock phosphate. As already pointed out for

the subsistence household, the semi-subsistence farm household cannot profitably adopt the

promoted fertilizer technologies just by increasing the agricultural output prices (see figure

50

21). Since sustainable land management practices are not adopted, the nutrient balances are

not increasing either. A change of the production structure is observed when the output price

is increased by 70 %. The area under improved maize would jump from 0.21 ha up to 0.37 ha,

simultaneously the area under intercropped improved maize and cassava and the area under

intercropped coffee and bananas decreases. This changing production structure leads to a

slight deterioration of nutrient balances.

Figure 21: Semi-Subsistence Farm Household: Sensitivity Analysis Output Prices

-100

-50

0

50

100

150

200

250

0 10 20 30 40 50 60 70 80 90 100


TGM 10^4 Ush

kg/ha NPK

TGMNPK


In comparison to the baseline scenario TGM would increase by 23 % with an output price

increase of 50 %, and by 47 % with an output price increase of 100 %. If rock phosphate

would be included, it could be profitably adopted on 0.89 ha under current price conditions,

which would lead to an improvement of the phosphorus balance from –12 kg/ha to +24 kg/ha.

Another increase of the area under rock phosphate (1.08 ha) would be achieved by raising the

output price up to 80 % of the current values.

The production structure of the trial farm household type is less stable compared to the

household types already discussed, when output prices are increased stepwise. As we have

seen for the other two household types already, improved maize can be adopted profitably

with increasing agricultural product prices. Figure 22 shows that the relative competitiveness

% of outprice increase

0 10 20 30 40 50 60 70 80 90 100

Production structure (ha) Maize I Maize i/Cassava Sweet Pot Bu Millet Sorghum Coffee l/Ban

0.21 0.75 0.33 0.21 0.11 0.25

0.21 0.75 0.33 0.21 0.11 0.25

0.21 0.75 0.33 0.21 0.11 0.25

0.21 0.75 0.33 0.21 0.11 0.25

0.21 0.75 0.33 0.21 0.11 0.25

0.21 0.75 0.33 0.21 0.11 0.25

0.21 0.75 0.33 0.21 0.11 0.25

0.37 0.73 0.33 0.21 0.11 0.2

0.37 0.73 0.33 0.21 0.11 0.2

0.37 0.73 0.33 0.21 0.11 0.2

0.37 0.73 0.33 0.21 0.11 0.2

51

of the sweet potato variety Silk is rising, leading to an increase of the cultivated area. In the

baseline scenario only 0.19 ha are under Silk, whereas a price increase of 30 % to 100 %

could result in a cultivation of 1.12 ha. Another interesting aspect is the decreasing area under

local coffee and the simultaneous increase of the area under bananas.

Figure 22: Trial Farm Household: Sensitivity Analysis Output Prices

-80

-60

-40

-20

0

20

40

0 10 20 30 40 50 60 70 80 90 100

outprice increase %

TGM 10^5 Ush

kg/ha NPK

TGMNPK


Under current price conditions local coffee is planted on 0.55 ha. An output price increase of

10 % and more would affect the relative profitability of coffee cultivation negatively,

resulting in an exclusion of this farming activity. Instead, banana cultivation becomes

economically more attractive and can be adopted profitably on 0.10 ha when the agricultural

product prices increase by 10 %. Again, the relative competitiveness of farming activities

involving the application of fertilizers is not increasing sufficiently for an adoption of these

technologies. Therefore, no improvement of the nutrient balances can be expected from rising

the level of output prices only. Rather the contrary is the case: the changing production

structure affects the nutrient balances negatively, e.g. the balance for nitrogen in the baseline

scenario is –43 kg/ha and is decreasing with a price increase of 30 % to –57 kg/ha. One factor


0 10 20 30 40 50 60 70 80 90 100

Production structure (ha) Maize I Maize i/Beans Sweet Pot Bu Sweet Pot S Cassava Maize i/Cassava Gnuts I Coffee l Ban Cow local

0 0.54 0.30 0.19 0.32 0 0.17 0.55 0 1

0 0.54 0.30 0.17 0.60 0 0.17 0.17 0.10 1

0 0.54 0.30 0.82 0.36 0 0.17 0.0 0.13 1

0.09 0.54 0.30 1.12 0.19 0.02 0.17 0 0.15 1

0.09 0.54 0.30 1.12 0.19 0.02 0.17 0 0.15 1

0.09 0.54 0.30 1.12 0.19 0.02 0.17 0 0.15 1

0.09 0.54 0.30 1.12 0.19 0.02 0.17 0 0.15 1

0.11 0.54 0.30 1.12 0.21 0 0.17 0 0.15 1

0.11 0.54 0.30 1.12 0.21 0 0.17 0 0.15 1

0.11 0.54 0.30 1.12 0.21 0 0.17 0 0.15 1

0.11 0.54 0.30 1.12 0.21 0 0.17 0 0.15 1

52

contributing to this negative impact is the relatively high nutrient extraction caused by sweet

potatoes. It is obvious that TGM is rising with increasing product prices, e.g. 50 % output

price increase results in a TGM increase of 17 %. Including rock phosphate in the selectable

technology options would lead to a significant improvement of the phosphorus balance. The

adoption would increase from 0.85 ha under current prices to 1.3 ha with a price increase of

50 %, and to 1.45 ha with a price increase of 100 %. At the same time the phosphorus balance

improves from –11 kg/ha under current prices when no rock phosphate is provided, to +21

kg/ha when it is selectable, to 37 kg/ha and 42 kg/ha with a 50 % and 100 % price increase

respectively.

Figure 23: Commercial Farm Household:Sensitivity Analysis Output Prices

-100-80-60-40-20

020406080

100

0 10 20 30 40 50 60 70 80 90 100


TGM 10^5 Ush

kg/ha NPK

TGMNPK


0 10 20 30 40 50 60 70 80 90 100

Production structure (ha) Maize I Maize I/Beans Maize I/Cassava Sweet Pot Bu Sweet Pot S Gnuts I Coffee Clonal Cow exotic

1.90 0.43 0.77 0.56 0.43 0.24 0.22 1

3.93 0.43 0.02 0.56 0.44 0.24 0.22 1

3.39 0.43 0.55 0.56 0.98 0.24 0.22 1

3.92 0.43 0.02 0.56 1.51 0.24 0.22 1

3.04 0.43 0.20 0.56 2.05 0.24 0.22 1

3.04 0.43 0.20 0.56 2.05 0.24 0.22 1

3.04 0.43 0.20 0.56 2.05 0.24 0.22 1

3.09 0.43 0.31 0.56 2.32 0.24 0.22 1

3.09 0.43 0.31 0.56 2.32 0.24 0.22 1

3.47 0.43 0.12 0.56 2.32 0.24 0.22 1

3.47 0.43 0.12 0.56 2.32 0.24 0.22 1


The production structure of the commercial farm household reveals its relatively high degree

of market orientation (see figure 23). A quite big proportion of the cultivated land is under

improved maize, the sweet potato variety Silk, and clonal coffee. Furthermore, the household

already adopted an exotic cow for milk production. The production structure is subject to

several changes when output prices are increased stepwise. Especially, the cultivated area

under improved maize and sweet potato Silk is significantly increasing. The area under clonal

coffee and the numbers of exotic cows is constant. Again, the reduced costs of farming

53

activities involving fertilizer application are not decreasing sufficiently by changing output

prices only. Thus, their adoption is not profitable yet. Changing production patterns are

leading to a slight decrease of the nutrient balances for nitrogen, phosphorus and potassium.

An increase of output prices of 50 % rises TGM by 26 %. The inclusion of rock phosphate as

an additional technology option is leading to an adoption of Busumbu Rock Phosphate on

4.03 ha in the baseline scenario. Consequently, the nutrient balance for phosphorus is

increasing from –16 kg/ha to +59 kg/ha. A price increase for agricultural products does not

result in a further increase of the adoption of rock phosphate.

This sensitivity analysis reveals that increased agricultural product prices alone do not lead to

a profitable adoption of new fertilizer technologies. Therefore, the nutrient balances remain

highly negative or even decrease further when output prices are increasing due to changes of

the production structure. The production structure of the subsistence and semi-subsistence

household type is relatively stable due to a low degree of market orientation, whereas the

changes for the trial farm household and commercial farm household are significant. For each

household type the cultivation of improved maize becomes more profitable, and the trial farm

and the commercial farm household type are expanding the cultivated area under the sweet

potato variety Silk. As expected, TGM increases with increasing output prices, e.g. with a

price increase of 50 % TGM increases 11.5 % – 26 %.

Scenarios #2.3: Introduction of credit, improvement of price relations, promotion of labor

exchange

The sensitivity analyses discussed above indicate that neither a sole decrease of fertilizer

prices to a realistic degree, nor the sole increase of agricultural output prices could probably

contribute to the simultaneous achievement of sustainability and economic goals. Therefore,

in the following it is examined whether combining both price effects and additional measures

like provision of credit or the promotion of alternative forms of labor acquisition could

potentially improve the current situation which is characterized by serious nutrient depletion.

For this purpose, sensitivity analyses of simultaneously decreasing fertilizer prices and

increasing agricultural product prices are conducted to identify promising price relations with

low reduced costs for farming activities involving the application of fertilizers. Shadow prices

of resource constraints help to identify most binding factors affecting the adoption of new

technologies. The following figures 24 – 27 illustrate the potential consequences of promoted

agricultural technologies and reduction of market failures on household welfare and

sustainability criteria for the different household types. The reduction of market failures is

54

reflected by simultaneously decreasing input prices and increasing output prices. Provision of

credit is included for assessing the impact of capital constraints on the adoption of new

technologies. Another constraint frequently quoted as a major factor prohibiting the adoption

of new technologies is the labor constraint. Labor exchange, a traditional form of labor

acquisition in the study region, was included in the scenarios to investigate whether it would

be an appropriate option to overcome the problem of labor shortages. The tables below the

diagrams define the scenarios considered and indicate the area on which a certain technology

is adopted.

The constraints for credit provision differ among the farm household types. Collaterals are the

fundamental prerequisite for getting a loan. The access to credit from microfinance institutes

like Pride Africa is restricted to households with regular off-farm income, since the repayment

of the loan is split into successive periods. The model divides the repayment of loans received

from microfinance institutes into periods of 4 months. The duration of commercial loans is 12

months. The maximum loan a household can receive is determined mainly by the level of off-

farm income and other collaterals. This maximum amount differs among the different

household types.16 The interest rate for a loan from a microfinance institute is set at 20 %. The

interest rate for a loan from a commercial bank is set at 22 %. Households with a low wealth

status – like subsistence farm households – would probably face serious problems of applying

successfully for loans from microfinance institutes as well as from commercial banks. Despite

these problems, this household type is included in the credit scenarios to illustrate the

potential benefits of a widespread provision of credit, including poor farmer.

Scenario 1 in figure 24 illustrates the situation for the subsistence farm household type under

current price relations. Under these conditions there is no adoption of fertilizer technologies

and the nutrient balances are highly negative (nitrogen: -29 kg/ha, phosphorus: -8 kg/ha,

potassium: -40 kg/ha). A price relation that allows the subsistence farm household to

profitably adopt NP-fertilizer (0.02 ha) is given when the input price is reduced to 40 % of the

current value and the output prices are increased by 50 %. The TGM increases by 12 % in

comparison to the scenario with current price relations (scenario 1), but the increase of the

nutrient balances is not significant. Providing credit would increase the area on which NP-

fertilizer is applied slightly (to 0.05 ha), contributing to a further modest improvement of the

nitrogen and phosphorus balances. Scenarios 4 and 5 compare the impact of credit provision

when the fertilizer prices are reduced to 37.5 % of the current prices and the agricultural

product prices are increased by 50 %. Without credit NP-fertilizer are profitable adopted only

16 Determination of maximum loan and down payment is based on expert knowledge.

55

on 0.03 ha. With the provision of credit the subsistence farm household type might adopt this

technology on 0.73 ha, leading to a positive phosphorus balance (+17 kg/ha) and a significant

improvement of the nitrogen balance (-11 kg/ha). The balance of potassium is decreasing

further though from –40 kg/ha to –49 kg/ha. Regarding the goal of reducing the high negative

values of nutrient balances, the most favourable situation is achieved when the fertilizer prices

are reduced at least to 20 % of the current prices, output prices are increased by 50 % and

credit is provided. The household can adopt NPK-fertilizer on 0.73 ha. In comparison to

scenario 1 TGM increases by 16 % and the N-, P-, K-balances are increasing from –29 kg/ha

to –12 kg/ha, -8 kg/ha to +17 kg/ha and –40 kg/ha to –23 kg/ha respectively.

Scenario 1 Scenario 2 Scenario 3 Scenario 4 Scenario 5 Scenario 6 Scenario 7 Characteristics • current

conditions • trial tech

• trial tech • input

price: 40%

• output price: +50%


price: 40%


• credit


price: 37.5%



price: 37.5%


• credit


price: 20%



price: 20%


• credit Area adopted (ha) 0 NP

0.02 NP 0.05

NP 0.03

NP 0.73

NP 0.05

NPK 0.73


Figure 24: Subsistence Farm Household: Scenar ios on combined ef fects

-90

-40

10

60

110

160

Scen1 Scen2 Scen3 Scen4 Scen5 Scen6 Scen7

TGM 10^4 Ush

kg/ha N P K

TGMNPK

56

A price relation which in combination with the provision of credit leads to a profitable

fertilizer adoption by the semi-subsistence farm household type is given when the input price

is reduced to 30 % of the current value and 10 % are added to the current product price

(scenario 3). NP-fertilizer can be adopted on 0.21 ha contributing to improvements of the

nitrogen and phosphorus balance (for N from –52 kg/ha to –44 kg/ha and for P from –12

kg/ha to –2 kg/ha). TGM rises by 5 %. Without the provision of credit NP-fertilizer adoption

is not profitable. NPK-fertilizer adoption becomes profitable when the input prices are

decreased to 25 % of the current price, the ouput prices are increased by 40 %, and credit is

provided.




price: 30%



price: 30%


• credit


price: 25%



price: 25%


• credit


price: 25%


• credit


price: 25%


• credit • labour

exchange Area adopted (ha) 0 0 NP

0.21 0 NPK

1.28 NPK 1.28

NPK 1.72


The application of NPK-fertilizer on 1.28 ha is significantly increasing the nutrient balances

(nitrogen: -2 kg/ha, phosphorus: +53 kg/ha, potassium –17 kg/ha.). Again, scenario 4

illustrates that fertilizer adoption is not profitable with same price relations but without the

provision of credit. Hence, a high shadow price for capital is indicating the profitability of

improved access to credit provision. Scenarios 6 and 7 explore the effects of a further input

price reduction (25 % of current price), a further output price increase (50 %), credit

F i g u r e 2 5 : S e m i - S u b s i s t e n c e F a r m H o u s e h o l d : S c e n a r i o s o n c o m b i n e d e f f e c t s

- 1 0 0

- 5 0

0

5 0

1 0 0

1 5 0

2 0 0

S c e n 1 S c e n 2 S c e n 3 S c e n 4 S c e n 5 S c e n 6 S c e n 7

T G M 1 0 ^ 4 U s h

k g / h a

N P K

T G MNPK

57

provision, and an additional option for labor exchange in scenario 7. The results of scenario 6

are nearly the same as for scenario 5. The only change is an additional increase of TGM by 5

%. Labor exchange has the potential to increase the area under NPK-fertilizer application

from 1.28 ha to 1.72 ha. This can lead to positive balances for N (+18 kg/ha) and P (+85

kg/ha), and a significant improvement of the K-balance (-2 kg/ha). Simultaneously, in

scenarios 6 and 7 TGM would rise by 26 % and 27 % respectively.




price: 37.5%



price: 37.5%


• credit


price: 25%



price: 25%


• credit


price: 25%


• credit • labour

exchange


price: 10%


• credit

Area adopted (ha) 0 NP 0.02

NP 0.11

NP 0.13

NPK 0.27

NPK 1.3

NPK 1.3


Focusing on the trial farm household type a price relation which can lead to a first adoption of

NP-fertilizers on 0.02 ha is given when the output price is reduced to 37.5 % and the

agricultural product price rises by 50 %, but only if credit is provided additionally. Because of

the small area on which NP-fertilizer is applied the changes of nutrient balances are

negligible, whereas TGM increases by 17 %. Additional provision of credit could lead to an

application of NP-fertilizer on 0.11 ha and an investment into another local cow. The

following scenarios 4 – 6 are based on an input price decrease to 25 % of the original value

Figure 26 : Tr ia l Farm Household: Scenar ios on combined e f fec ts

-70

-50

-30

-10

10

30

50

S c e n 1 S c e n 2 S c e n 3 S c e n 4 S c e n 5 S c e n 6 S c e n 7

T G M 10^5 Ush

kg/ha N P K

T G MNPK

58

and an ouput price increase by 60 %. Credit is provided in scenarios 5 and 6. In scenario 6 the

option for labor exchange is added. In Scenario 4 (without credit and labor exchange) NP-

fertilizer can be adopted profitably on 0.13 ha. The changes of nutrient balances are

negligible, TGM rises by 21 %. When credit is provided additionally, NPK-fertilizer can be

applied on 0.27 ha, which leads to an increased nutrient balance for phosphorus. The balances

for nitrogen and potassium are decreasing in comparison to scenario 1 from –43 kg/ha to –48

kg/ha and –47 kg/ha to –52 kg/ha respectively, due to changing production patterns. A

significant increase for all nutrient balances can be observed when the trial farm household

type is given the option for labor exchange. The shadow price in the time periods where the

trial farm household is acquiring labor by exchange is 500 USh, which is quite high in

comparison to other periods where the shadow prices are around 200 USh. The nitrogen

balance jumps up to –6 kg/ha, the phosphorus balances reaches a value of +50 kg/ha and the

potassium balances increases to –18 kg/ha. Simultaneously, TGM rises by 29 %. To achieve

the same values for the nutrient balances without labor exchange, the fertilizer prices have to

be reduced to 10 % of the current price, and the output prices have to increase at least by 10 %

(scenario 7). A fertilizer price decrease to 40 % and a simultaneous rise of the product price

by 50 % leads to a profitable adoption of NP-fertilizer by the commercial farm household

type only if credit is provided additionally. The adoption of this fertilizer type is increasing

the nitrogen and phosphorus balances in comparison to scenario 1 (from –77 kg/ha to –53

kg/ha and from –15 kg/ha to +22 kg/ha respectively). At the same time TGM rises by 27 %.

NP-fertilizer can be adopted profitably on 3.66 ha when input prices are reduced to 30 % of

the current value, output prices are increased by 50 %, and credit is provided. Consequently,

nutrient balances for N and P are increasing significantly (N-balance: -17 kg/ha, P-balance:

+65 kg/ha), whereas the balance for K reaches even higher negative values than in scenario 1

(K-balance: -98 kg/ha). Again, this price relation cannot lead to an adoption if credit is not

provided. High shadow prices for capital indicate that the economic situation of the household

can be improved substantially if access to credit is improved. To achieve a profitable adoption

of NPK-fertilizer (on 3.95 ha), fertilizer prices have to be reduced at least to 20 % of the

current price, output prices have to increase by 50 %, and credit has to be provided. This price

constellation cannot lead to a profitable adoption of NPK-fertilizer without credit. Scenario 6

illustrates that without credit NP-fertilizer can be adopted profitably on 1.37 ha.

59




price: 40%



price: 40%


• credit


price: 30%



price: 30%


• credit


price: 20%



price: 20%


• credit Area adopted (ha) 0 0 NP

1.76 0 NP

3.66 NP 1.37

NPK 3.95


The scenarios revealed that improved input-ouput price relations in combination with

provision of credit and alternative forms of labor acquisition (labor exchange) can contribute

to a significant increase of nutrient balances and TGM simultaneously. The sensitivity

analyses focusing exclusively on input price reduction (scenarios #2.1) illustrate that fertilizer

prices have to be reduced to 10 % - 5 % of the current prices before a major increase of the N-

, P- and K-balances can be achieved. To achieve input price reductions to this extent by

improvements of the market environment alone, seems to be very problematic though.

Combining fertilizer price decrease with ouptut price increase, the provision of credit and

labor exchange can lead to significant improvements of nutrient balances when fertilizer

prices are reduced to at least 25 % - 20 % of the current prices. Referring to the discussion

about the potential level of fertilizer price reduction, even a price reduction to this extent is

difficult to achieve. Economies of scale in transport, improvements in the marketing chain,

and increased competition have to be attained simultaneously to reach input prices of

comparable levels.

Figure 27: Commercia l Farm Household: Scenarios on combined ef fects

-120-100

-80-60-40-20

020406080

Scen1 Scen2 Scen3 Scen4 Scen5 Scen6 Scen7

TGM 10^5 U s h

kg/ha N P K

TGMNPK

60

The scenarios revealed that at the same time the agricultural product prices have to increase

by about 50 %. This level might be realistic to attain by improving the bargaining power of

farm households. Generally, it was illustrated that a substantial improvement of the socio-

economic environment is urgently needed to give farmers sufficient incentives to adopt more

sustainable land management practices.

Scenarios #2.4: Economic and ecological impact of the intensive production of high value

crops

A potential profitable pathway involves intensive production of high value crops and

perennial crops due to the characteristics of high agricultural potential, high market access,

and high population density in Iganga District. Therefore, this type of scenarios deals with the

introduction of vegetables and fruits (onions, tomatoes and passion fruits), and investments in

supplementary technologies. Simultaneously, the introduction of credit for operational and

investment costs will be considered. At the moment, only households that belong to the

commercial and trial farm household groups are cultivating vegetables or fruits. Increased

productivity and the transformation of subsistence or semi-subsistence farming systems to

more commercially oriented systems belong to the major objectives of the Plan for

Modernization of Agriculture (PMA). Additionally, it is proposed in the PMA to minimize

over-reliance on rain-fed agriculture, and to promote irrigation technology for high value

crops for specialized markets. Prolonged dry seasons – mentioned by many farmers as the

major problem of agricultural production in the study area – might prevent farm households

from investing in expensive inputs as long as economic returns are not guaranteed due to

potential droughts. For each household type the potential consequences of the defined

scenarios on economic and ecological indicators will be illustrated. The introduction of

vegetables and fruits contributes to a slight increase of the TGM of the subsistence farm

household type (4 %) due to a change of the production structure (see appendix A 11 a).

Onions can be profitably adopted on 0.04 ha. Other economic indicators like labor intensity,

labor productivity, land productivity, capital intensity and capital productivity, for which very

low values are given in the baseline scenario, are increasing to a modest degree as well.17

Nutrient balances are decreasing slightly with the introduction of vegetables.

61

Tab. 7: Introduction of high value crops: Subsistence Farm Household Type Base Scenario 1

• Veg.+Fruits • Trial tech

Scenario 2 • Veg+Fruits • Credit op.cost • Trial tech

Scenario 3 • Veg+Fruits • Credit op.cost • Credit inv.cost • Trial tech

TGM (103USh) 1299 1356 2166 2403 Labor Intensity (h/ha) 1205 1224 1508 1551 Labor Productivity (103USh/h)18

0.29 0.3 0.5 0.66

Land Productivity (USh/ha)19

349 374 752 1025

Capital Intensity (103USh/ha)

259 265 347 494

Capital Productivity (USh/USh)20

1.34 1.41 2.17 2.07

Nutrient Balances21 N -28 -30 -45 -47 P -8 -9 -14 -13 K -39 -41 -51 -55 Source: own calculations (model results)

In the study region, the adoption of vegetables and fruits is not associated with application of

fertilizer so far. Therefore, lack of data makes it impossible to run scenarios with the

combined introduction of vegetables or fruits and fertilizer. In scenarios with the introduction

of treadle pumps for irrigation, fertilizers are applied on the irrigated area. In scenario 2 credit

for operational costs is provided in addition to the introduction of vegetables and fruits. The

major changes of the production structure are the increased proportion of onions and

decreased proportion of intercropped coffee-bananas. The economic indicators are increasing,

TGM rises by 67 %. At the same time a substantial decline of the nutrient balances is

observed, when vegetables/fruits are introduced without application of fertilizers. The

nitrogen balance decreases by more than 60 %. The additional provision of credit for

investments leads to a further rise of TGM (by 85 % in comparison to the baseline scenario).

This credit type is used for investment in a treadle pump for irrigation. The cultivated area

under onions is declining, whereas tomatoes (irrigated) are introduced. The nutrient balances

of nitrogen, phosphorus and potassium decrease to –47 kg/ha, -13 kg/ha and –55 kg/ha

respectively.

Comparing the production structure of the semi-subsistence farm household type (see figures

28-30) in the baseline scenario and in scenario 1, the proportion of improved maize

17 Factors explaining why farm households do not produce vegetables and fruits in the baseline scenario will be discussed in the conclusions. 18 Labor productivity refers to total gross output per labor hour 19 Land productivity refers to total gross output per ha agricultural land 20 Capital productivity refers to total gross output per USh used

62

intercropped with cassava is decreasing from 53 % to 2 % of the cultivated land, whereas

onions are introduced on 25 % in scenario 1. Improved maize mono-cropping (7 %) is

substituted by local maize mono-cropping (28 %). In the third scenario credit is used for

investment in a treadle pump for irrigation leading to an additional change of the production

structure. Tomatoes (irrigated) are introduced on 17 % of the cultivated land and passion

fruits on 10 %. BRP is applied on improved maize on 7 % of the cultivated land. The

economic indicators TGM, labor productivity, land productivity, and capital productivity are

increasing progressively with the introduction of vegetables and fruits in the first scenario,

additional provision of credits for operational costs in the second plus provision of credits for

investment costs in the third scenario. TGM is increasing by more than 100 % from the

baseline scenario to the third scenario. Land productivity is nearly four times higher.

Tab. 8: Introduction of high value crops: Semi-Subsistence Farm Household Type

Base Scenario 1 • Veg.+Fruits

Scenario 2 • Veg+Fruits • Credit op.cost

Scenario 3 • Veg+Fruits • Credit op.cost • Credit inv.cost

TGM (103USh) 1490 2205 2987 3280 Labor Intensity (h/ha) 1207 1610 1902 1925 Labor Productivity (103USh/h)

0,41 0,6 0,79 0,99

Land Productivity (USh/ha)

498 967 1493 1919


272 369 559 758

Capital Productivity (USh/USh)

1,83 2,62 2,67 2,53

Nutrient Balances N -52 -63 -84 -83 P -12 -14 -2 -4 K -62 -69 -87 -87 Source: own calculations (model results)

The negative values of nutrient balances of N and K are increasing, whereas K is slightly

decreasing when vegetables, fruits and credits were introduced. In the baseline scenario the

balance for N was –52 kg/ha, for P –12 kg/ha and for K –62 kg/ha. In scenario three –83

kg/ha for N, -4 kg/ha for P and –87 kg/ha for K.

21 Nutrient balances are measured in kg per ha and year

63

Figure 28: Production Structure (Baseline Scenario)

Maize i7%

Maize i/Cs53%

Spot12%

Millet+Sorg11%

Coffee/Ban17%

Figure 29: Production Structure (Scenario 1)

Maize l28%

Maize i/Cs2%

Spot12%Millet+Sorg

11%

Coffee/Ban14%

Coffee8%

Onions25%

Figure 30: Production Structure (Scenario 3)

Maize i/Cs2%

Spot11%

Millet+Sorg11%

Coffee/Ban14%

Maize i BRP7%

Onions28%

Passion10%

Tomatoes (irrigated)

17%

64

The production structure of the trial farm household is changing with the introduction of

vegetables/fruits in scenario 1 (see appendix 11 b) as well. Onions are introduced on 1.36 ha

and passion fruits on 0.02 ha. At the same time the cultivation of local coffee is not profitable

any more. With the introduction of treadle pumps in scenario 3, tomatoes (irrigated) are

introduced on 0.86 ha and the production of onions declines to 0.5 ha.

Tab. 9: Introduction of high value crops: Trial Farm Household Type

Base Scenario 1 • Veg.+Fruits

Scenario 2 • Veg+Fruits • Credit op.cost



0,56 0,86 1,02 1,32


662 1528 1817 2542


499 994 1173 1600


1,33 1,54 1,55 1,59

Nutrient Balances N -43 -79 -79 -84 P -11 -25 -25 -25 K -47 -74 -74 -87


The introduction of vegetables and fruits (scenario 1) and the additional provision of credits

for operational and investment costs (scenarios 2 and 3) contribute to significant increases of

all economic indicators. TGM for example increases from scenario 1 to scenario 3 by 54 %,

63 % and 119 % in comparison to the baseline scenario. At the same time labor productivity

increases by 54 %, 82 %, and 136 % respectively. This positive development of the economic

indicators is again associated with substantial decreases of nutrient balances. The balance for

nitrogen for example declines by nearly 100 % in comparison to the baseline scenario.

As discussed above the commercial farm households have already in the baseline scenario

high negative nutrient balances. These balances for nitrogen and potassium become even

more negative with introduction of vegetables/fruits, credits, and irrigation pumps. The

nitrogen balance is decreasing from –77 kg/ha (baseline scenario) to –97 kg/ha (scenario 3).

The balance for phosphorus is increasing from –15 kg/ha to +15 kg/ha (scenario 2) due to the

application of rock phosphate on 2.35 ha. In scenario 3 the area on which rock phosphate is

applied declines to 0.24 ha contributing to a decrease of the P-balance to –12 kg/ha. The

production structure of the commercial farm household is changing with the introduction of

65

vegetables/fruits (see appendix A 11 c). Onions are introduced on 2.88 ha and passion fruit is

profitably adopted on 0.61 ha. The production of clonal coffee is neglected in the farm

management plan in scenario 1, whereas in the baseline scenario it is cultivated on 0.22 ha.

The provision of credit and the following investment in pumps for irrigation leads to the

adoption of tomatoes (irrigated) on 1.5 ha, an increase of the area under passion fruit (1.44 ha)

and a decline of the area under onions to 1.38 ha. As for the other households before, TGM is

increasing (by 83 % in scenario 3) simultaneously with other economic indicators. Land

productivity is increasing by 94 %, 122 % and 179 % from scenarios 1 -– 3 in comparison to

the baseline scenario.

Tab. 10: Introduction of high value crops (Commercial Farm Household Type) Base Scenario 1

• Veg.+Fruits Scenario 2 • Veg+Fruits • Credit op.cost



0,68 1,12 1,09 1,3


925 1799 2057 2578


705 1013 1145 1457


1,3 1,78 1,8 1,77

Nutrient Balances N -77 -86 -116 -97 P -15 -18 +15 -12 K -71 -71 -104 -85

Source: own calculation (model results)

The scenarios on economic and ecological impacts of the production of high value crops –

identified before as a promising strategy for the development pathway with high population

density, high market access and high agricultural potential – reveal to a certain degree trade-

offs between economic and sustainability goals. Economic indicators are increasing

substantially while nutrient balances are decreasing in comparison to the current situation,

which is already characterized by serious nutrient depletion. It should be emphasized again,

that due to the lack of data it was not possible to run scenarios with fertilizer application on

vegetables without irrigation. This option should be taken into account seriously when

promoting sustainable intensification of agricultural. Moreover, the scenarios illustrated that

irrigation in combination with application of fertilizers is highly profitable. Therefore, a

potentially profitable strategy would be not to restrict the promotion of irrigation to high value

66

crops such as vegetables and fruits, but also to extend irrigation on crops like maize. Due to

very high yields on irrigated plots, the nutrient extraction through harvested products cannot

be balanced through the amount of fertilizer applied.

5. Conclusions

Current land management practices in the study region can be characterized by low land,

labor and capital productivity for the majority of the farm households leading to poor

economic performances and food insecurity in some cases. These characteristics of low

intensity agricultural production systems reveal that potentially profitable pathways involving

intensive production of high value crops are not realized yet. In addition, highly negative

nutrient balances raise the concern of declining soil productivity in the future. Assuming that

the mass of the topsoil is 3 * 106 kg/ha and the average N content of identified soil class 1 is

0.12 %, the annual decrease of 52 kg/ha (annual N loss of the semi-subsistence household

type in the baseline scenario) is equivalent to 1.4 % of the N-stock. With a constant relative

loss rate of 1.4 % per year, soil N content will be half the present value after 50 years.

Scenario results illustrated that non-negative nutrient balances are not feasible with current

land management practices. Even with the simultaneous introduction of promoted technology

options and credit provision the achievement of this sustainability goal was not feasible in

many cases – especially for the subsistence, semi-subsistence, and trial farm household type -

due to non-profitability. The reasons for this non-profitability are unreasonably high fertilizer

prices and very low agricultural product prices due to market imperfections. Additionally, the

impacts of the promoted technologies on yield are modest in many cases, as the ANN-model

indicated. Yield data were available only for 4 seasons (2 years). Long-term trials might give

higher average yields for these technologies. Moreover, the fertilizer technologies were tested

as on-farm trials. Trials under the permanent control of researchers at research stations might

have led to more positive results. Further nutrient depletion might lead to increasing yield

impacts as well. Sensitivity analyses on input price reduction illustrated that fertilizer prices

have to decline extremely (to 10 % - 5 % of the current prices) before fertilizer application

becomes profitable to an extent, which can contribute to substantial improvements of nutrient

balances. The discussion about the potential impacts of market improvements on fertilizer

price decreases indicate that it is very problematic to reach price reductions of this level.

Sensitivity analyses show that increasing output prices can contribute to a significant

improvement of the household welfare, but simultaneous improvements of nutrient balances

67

cannot be expected. Increasing output prices might be achieved by strengthening the

bargaining power of farm households by improving the access to relevant information.

Capital and labor constraints prohibit farm household from the adoption of new technologies.

Isolated measures with marginal effects on input and output prices will probably not help to

reach the goals of agricultural growth, poverty alleviation, and sustainability. Rather

combined measures leading to significant improvements of the socio-economic environment

are needed to increase the chances of reaching the desired development goals. The next step

was to explore whether the combined effects of declining fertilizer prices to a realistic extent

and increasing output prices in combination with credit provision and alternative forms of

labor acquisition have the potential to reach economic and sustainability goals

simultaneously. Scenario results demonstrate the important role credit provision can play for

the adoption of new technologies. Labor exchange – as an alternative form of labor

acquisition – should be taken into account as well while promoting sustainable land

management practices. At the same time fertilizer price reduction to 25 % - 20 % of the

current prices and an agricultural product price increase by 50 % are needed to achieve

significant improvements of nutrient balances and household welfare. This extent of

decreasing input prices and increasing output prices might be difficult to achieve as well, but

combined effects of economies of scale and reduction of market failures could lead to

substantial price changes. The production of high value crops - identified as a potentially

profitable strategy in a study region belonging to a development pathway with high

population density, high market access and high agricultural potential – could lead to

substantial improvements of economic indicators. The impacts on nutrient balances need

further research, but cultivation without fertilizer application, as predominantly practiced in

the study region, contributes to trade-offs between economic and sustainability goals, since

the nutrient balances are decreasing to a significant extent.

This study concludes that the main reason for the fact that farm households do not realize the

potentials provided by a development domain with high population density, high market

access and high agricultural potential is the non-profitability of intensive agricultural

production activities under current socio-economic and agro-ecological conditions. The main

reasons for this non-profitability are market imperfections reflected by high transaction costs,

high transportation costs, and insufficient access to credit markets. At the same time the

modest impacts of promoted agricultural technologies on yields cannot cover the high input

prices.

68

The study reveals that further research is needed, both in socio-economic and agro-ecological

perspectives. The applied bio-economic model is an appropriate approach for identification of

the optimal level of technology adoption and the impact on incomes and natural resource

conditions for heterogeneous household agents in a changing socio-economic environment.

As discussed above, model solutions sometimes differ from the observed values, e.g. when

vegetables and fruits are introduced exogenously, farm households adopt the production of

these crops in contrast to the real world situation where these activities are not included in the

farm plan. This leads to the important research question what farm households prevent from

producing vegetables and fruits in the baseline scenario. An appropriate method to answer this

question might be multi-agent systems (MAS). The sampling procedure, the selection of

representative farm household types, and diffusion analyses described in previous sections

fulfill specific data requirements to extend the bio-economic model developed in this study to

a dynamic version implemented as a connected multi-agent system. This model type might be

an appropriate approach to forecast the diffusion of innovations together with the evolution of

farm incomes and natural resource conditions over time (Berger, 2001). Two variables affect

the adoption and diffusion of innovations. Firstly, the net benefit of adoption, which can be

“objectively” measured and is accounted for in the programming approach developed for this

study. Secondly, costs that relate to farmer’s managerial capacity. These costs are usually

referred to as adoption costs and include planning and information costs, socio-psychological

adjustment costs, temporary production losses as well as “subjective” risk premiums and

option values.22 In the MAS-approach developed by Berger (2001) a decision rule is

implemented which accounts for adoption thresholds based on social interactions in addition

to net benefit considerations. This is an interesting approach to consider at least part of the

adoption costs and explain why technologies normally do not diffuse as “smooth” as

predicted with approaches, which focus on net benefits only.

From the natural science perspective further research activities are needed to better

understand the impact of the altitude of negative nutrient balances on yield in the long run.

Scenarios on improved price relations, provision of credit and promotion of alternative forms

of labor acquisition indicated that the negative value of N balance of the subsistence farm

household type was reduced to 43 % of its original value, which is equivalent to a loss of 12

kg per ha and year. Assuming the mass of the topsoil defined above and N content of soil

class 1 with a constant N loss the soil N content will be half in 230 years. It has to be clarified

whether slightly negative nutrient balances really affect crop yields negatively in the long run.

22 Metcalfe (1988) illustrates how these two variables influence the diffusion process in the agricultural sector.

69

Therefore, the understanding of nutrient dynamics in the soil and the feedback effects

between nutrient depletion, soil nutrient content and yield level has to be improved.

70

Appendix A1:

Descriptive Statistics: Household Characteristics

1 17 7,85 3,61

1 11 3,60 2,16

20 80 45,97 17,5218 70 35,14 12,39

0 18 6,07 4,67

0 15 4,68 4,07

0 2 ,36 ,64

0 2 ,22 ,52

number of householdmembers

number ofhh-members involvedin farmingage of household headage of wife

years of schoolinghousehold headyears of schooling wife

number ofhh-membersparticipated in training(since1990)

number ofhh-membersparticipated inextension in 1999/2000

Min Max Mean Std. Dev.

A2:

Primary Activity Household Head

farming72%

wage worker

9%

trading10%

other9%

A3:

Primary Activity Wife

farming94%

other6%

71

A4:

Descriptive Statistics: Household Asset Land

,02 30,00 5,61 5,26

,20 2,25 1,08 ,60

0 100 79 23

0 100 34 45

0 100 11 31

0 100 39 45

0 100 72 43

0 100 20 39

total land size (ownedor operated, in acres)land use intensity a

% of upland soils ontotal land size% of land inherited% of land received asa gift% of land purchased% of land underfreehold status% of land undercustomary status

Min Max Mean Std. Dev.

ratio berween land area cultivated in 12 months and total land sizea.

A5: Test for appropriateness of Principal Component Analysis

KMO and Bartlett's Test

.652

186.87428

.000

Kaiser-Meyer-Olkin Measure of SamplingAdequacy.

Approx. Chi-SquaredfSig.

Bartlett's Test ofSphericity

A 6: Cluster Analysis: Determination of cluster numbers

Number of Clusters Agglomeration Coefficient

Percentage Change in Coefficient to Next Level

10 9

20.5 23.6

15.1 16.9

8 27.6 19.2 7 32.9 27.1 6 41.8 28.5 5 53.7 23.3 4 66.2 44.3 3 95.5 42.6 2 136.2 36.6 1 186.0 -

72

A 7: Cluster Analysis

Total Variance Explained

2,84 35,53 35,531,44 17,99 53,521,06 13,25 66,77,91 11,33 78,10,88 10,96 89,06,46 5,78 94,84,33 4,19 99,03,08 ,97 100,00

Component

12345678

Total% of

VarianceCumulative

%

Initial Eigenvalues

Extraction Method: Principal Component Analysis.

A 8: ANN Output Reports

Season 2000 A: Desired Output and Actual Network Output

0

1000

2000

3000

4000

5000

6000

7000

8000

1 38 75 112 149 186 223 260 297 334 371

Exemplar

Ou

tpu

t

Grain yield (kg/ha) Grain yield (kg/ha) Output

Performance Stover Yield Adjusted

stover yield Cob yield Adjusted cob

yield Grain yield Adjusted

grain yield MSE 1161011 13040654 192293 773572 675463 762806 NMSE 0,31 0,33 0,41 0,35 0,29 0,30 MAE 812 2372 236 569 617 650 Min Abs Error 1,42 15,92 0,36 2,61 4,36 1,99 Max Abs Error 4404 18661 3624 4733 3298 3508 r 0,83 0,82 0,77 0,81 0,84 0,84

73

Abbreviations: MSE=mean squared error; NMSE=normalized mean squared error (MSE/variance of desired output); MAE=mean absolute error; Min Abs Error=minimum absolute error; Max Abs Error=maximum absolute error; r=correlation coefficient

Season 2000 B: Desired Output and Actual Network Output

0100020003000400050006000700080009000

10000

1 35 69 103 137 171 205 239 273 307 341

Exemplar

Ou

tpu

t






Season 2001 A: Desired Output and Actual Network Output

0

2000

4000

6000

8000

10000

12000

1 19 37 55 73 91 109 127 145 163 181

Exemplar

Ou

tpu

t





grain yield MSE 1313062 2300541 838583 1715369 3550533 31707294 NMSE 0,19 0,24 0,13 0,21 0,28 0,70 MAE 807 1073 704 1015 1193 2540

74

Min Abs Error 6,05 8,40 2,83 6,84 31,11 9,68 Max Abs Error 4815 5389 3758 4685 17274 65010 r 0,90 0,88 0,93 0,89 0,85 0,55

Season 2001 B: Desired Output and Actual Network Output

0100020003000400050006000700080009000

1 20 39 58 77 96 115 134 153 172 191

Exemplar

Ou

tpu

t






75

A 9: Soil Classification

Soil Class 1

39 5,10 6,90 5,6641 ,435639 1,70 3,60 2,7231 ,382839 ,07 10,50 3,3146 2,503239 12,90 32,50 21,7821 4,5627

39 ,08 ,18 ,1203 2,585E-0214 4,38 7,58 6,0843 ,921339 37,34 124,30 74,8528 24,737014 17,21 41,57 30,7479 5,642339 47,48 75,84 64,9446 6,9893

39 13,24 43,24 25,3897 6,998539 6,20 18,20 9,6656 2,208314

PHOM (%)P (ppm)K (mg/100g soil)

N (%)Na (mg/100g soil)Ca (mg/100g soil)Mg (mg/100g soil)SAND (%)

CLAY (%)SILT (%)Valid N (listwise)

N Minimum Maximum Mean Std. Deviation

Soil Class 2

18 4,50 5,80 5,3278 ,314018 1,70 2,50 2,0556 ,243118 ,45 11,40 4,2683 2,691418 9,80 20,40 14,3889 2,9478

18 ,06 ,13 9,667E-02 1,766E-024 3,20 4,15 3,7325 ,4055

18 25,90 72,36 49,9639 13,58514 13,30 19,04 16,2925 2,9987

18 49,48 80,56 71,3267 8,5671

18 9,24 39,24 19,5533 8,135818 5,28 15,64 9,1200 2,8892

4





Soil Class 3

4 3,90 4,90 4,2000 ,46904 2,70 5,00 3,5000 1,04244 ,97 7,14 4,4075 2,73034 11,40 18,90 14,2500 3,4723

4 ,13 ,25 ,1713 5,543E-021 4,74 4,74 4,7400 ,4 20,43 79,61 41,6600 28,05931 20,20 20,20 20,2000 ,4 54,92 77,48 65,2500 11,9252

4 13,24 34,52 21,5600 9,66004 9,28 21,64 13,1900 5,69371





76

Soil Class 4

14 5,60 7,00 6,1643 ,399214 2,70 4,70 3,6214 ,615414 ,22 16,70 7,3050 5,342114 25,50 47,50 34,4357 6,6609

14 ,08 ,24 ,1664 4,757E-023 7,23 8,65 8,0567 ,7382

14 65,32 268,02 160,3336 69,20283 25,85 34,50 30,4800 4,3571

14 45,48 81,48 59,2571 11,8709

14 9,24 47,24 29,9029 10,520714 6,20 21,28 10,8400 4,1181

3





Soil Class 5

3 5,60 6,00 5,7667 ,20823 2,60 3,20 2,8667 ,30553 22,20 27,80 24,1333 3,17703 24,10 32,90 29,7333 4,8911

3 ,10 ,16 ,1333 3,055E-021 5,92 5,92 5,9200 ,3 51,80 132,75 98,9233 42,08121 28,51 28,51 28,5100 ,3 52,56 77,48 66,5067 12,7233

3 17,24 34,88 25,1200 8,96903 5,28 12,56 8,3733 3,76111





Soil Class 6

5 4,00 4,50 4,2000 ,21215 1,30 1,50 1,3600 8,944E-025 ,45 2,68 1,2240 ,88345 ,47 3,30 1,6040 1,0862

5 ,04 ,06 5,200E-02 8,367E-035 ,24 1,07 ,5700 ,32895 6,76 10,14 8,5600 1,28495 6,07 9,23 7,6320 1,44495 76,92 84,92 81,3200 2,9665

5 8,52 12,52 10,5200 2,00005 6,56 10,56 8,1600 1,67335





77

A 10: Reduced costs agricultural product price sensitivity analysis Subsistent Farm Household: Sensitivity Analysis Agricultural Product Price Increase % increase of current price

0 10 20 30 40 50 60 70 80 90 100


182 211 313 188 252 364

181 203 304 186 239 348

180 195 294 185 226 332

180 188 284 184 214 316

179 180 275 182 201 300

178 172 266 182 188 284

178 164 256 180 175 268

177 157 247 179 162 253

201 196 303 203 198 304

265 315 470 267 317 471

330 434 637 332 437 639

Area adopted (ha)

0 0 0 0 0 0 0 0 0 0 0

Semi-subsistent Farm Household: Sensitivity Analysis Agricultural Product Price Increase % increase of current price

0 10 20 30 40 50 60 70 80 90 100


505 832

1154 464 792

1114

534 886

1233 500 852

1199

563 939

1311 536 911

1284

593 992

1389 572 971

1369

622 1045 1468 608

1031 1454

651 1098 1546 644

1091 1539

680 1151 1624 679

1150 1624

702 1185 1674 704

1187 1676

718 1206 1706 720

1208 1708

735 1227 1737 737

1229 1739

751 1248 1769 753

1250 1771

Area adopted (ha)

0 0 0 0 0 0 0 0 0 0 0

Commercial Farm Household: Sensitivity Analysis Agricultural Product Price Increase % increase of current price

0 10 20 30 40 50 60 70 80 90 100


213 255 368 189 238 346

213 239 349 203 244 353

215 228 335 218 251 362

250 284 415 260 317 455

298 366 531 309 403 574

326 409 593 338 448 639

354 452 655 367 493 704

382 495 717 395 538 767

410 537 779 424 584 833

417 538 782 431 586 839

417 523 764 431 573 823

Area adopted (ha)

0 0 0 0 0 0 0 0 0 0 0

78

A 11 a: Production Structures Subsistence Farm Household Type

Production Structure (Baseline Scenario)

Maize l/ Groundnuts

1%

Cassava5%

Coffee/ Banana41%

Maize l/ Beans31%

Sweet Pot Bu22%

Production Structure (Scenario 1)

Coffee/ Banana40%

Maize l/ Groundnuts

1%Cassava5%

Maize l/ Beans 31%

Sweet Pot Bu22%

Onions1%


Maize l/ Beans31%

Coffee/ Banana

22%Sweet Pot Bu

22%

Onions6%

Tomato12%

Cassava5%

Maize BRP1%

Maize l/ Groundnuts

1%

79

A 11 b: Production Structures Trial Farm Household Type


Coffee l30%

Sweet Pot Silk5%

Groundnuts i6%

Sweet Pot Bu7%

Maize / Beans15%

Cassava18%

Maize i/ Cassava

19%

Production Structure of (Scenario 3)

Tomato irr30%

Maize i/ Beans18%Onions

17%

Cassava9%

Ground Nuts Improved

8%

Sweet Pot Bu9%

Coffee cl irr5%

Maize l/ Beans1%

Banana3%


Onions49%

Maize l/ Beans1%

Passion1%

Banana5%Ground Nuts i

8%Sweet Pot Bu

9%

Cassava9%

Maize i/ Beans18%

80

A 11 c: Production Structures Commercial Farm Household Type


Maize i/ Cassava

17%

Cassava32%

Maize i22%

Ground Nuts i3%

Sweet Pot Silk5%

Sweet Pot Bu11%

Coffee cl5%

Maize i/Beans5%


Onions42%

Passion18%

Ground Nuts i4%

Maize i/ Cassava

18%

Sweet Pot Bu8%

Maize l/ Beans5%

Maize BRP5%


Sweet Pot Bu7%

Cassava9%

Maize BRP6%

Groundnuts i3%Maize l/ Beans

4%Passion

36%

Tomato irr18%

Onions17%

81

References Agricultural Policy Secretariat Uganda (2000): Report on Characterization of Policies and

Institutions Affecting Land Management.

Backhaus, K., B.Erichson, W.Plinke and R.Weiber (1994): Multivariate Analysemethoden,

7.Auflage, Springer-Verlag, Berlin.

Baffoe, John K. (2000): Structural Adjustment and Agriculture in Uganda. International

Labour Organization, Working Paper No.149. Geneva.

Barber, R. G., T. R. Moore, D. B. Thomas (1979): The Erodibility of Two Soils from Kenya.

In: J. Soil Sci. 30, 579-591.

Barbier, B. (1996): Impact of Market and Population Pressure on Production, Incomes and

Natural Resources in the Dryland Savannas of West Africa: Bioeconomic Modeling at

the Village Level. EPTD Discussion Paper No. 21. IFPRI, Washington.

Barbier, B. (1997): The Economic Determinants of Land Degradation in Developing

Countries. In: Philosophical Transactions: Biological Sciences. Land Resources: On

the Edge of the Multhusian Precipice? Vol. 352.

Barbier, B. and G. Bergeron (1998): Natural Resource Management in the Hillsides Of

Honduras: Bioeconomic Modeling at the Micro-Watershed Level. EPTD Discussion

Paper No. 32. IFPRI, Washington.

Beven, K. (1989): Changing Ideas in Hydrology- the Case of Physically Based

Models. Journal of Hydrology 105, 157-172.

Berger, T. (2001): Agent-based Spatial Models Applied to Agriculture: A Simulation Tool for

Technology Diffusion, Resource Use Changes and Policy Analysis. Agricultural

Economics Volume 25, Issue 2, 01-September-2001.

Bishop, Christopher M. (1995): Neural Networks for Pattern Recognition, New York.

Brandes, W. (1985): Über die Grenzen der Schreibtischökonomie. Tübingen.

Esilaba, A. O., J.B. Byalebeka, A. Nakiganda, S. Mubiru, R. J. Delve, H. Ssali and M.

Mbalule (2001): Integrated Nutrient Management Strategies in Eastern Uganda. (in

review)

Defoer, T., A. Buldeman, A. Toulman and S. Carter (2000): Building Common Knowledge.

Participatory Learning and Action Research (Part 4). In: Defoer, T. and A. Buldeman

(eds.): Managing Soil Fertility in the Tropics. A Resource Guide for Participatory

Learning and Action Research. Royal Tropical Institute, Amsterdam.

82

de Janvry, A., M. Fafchamps and E. Sadoulet (1991): Peasant Household Behaviour with

Missing Markets: Some Paradoxes Explained. In: The Economic Journal 101

Fafchamps, M. (1993): Sequential Labor Decisions under Uncertainty: An Estimable

Household Model of West African Farmers. Econometrica 61.

FAO (1999): Uganda Soil Fertitily Initiative, Draft Concept Paper. Rome.

International Fertilizer Development Center (1999): The Fertilizer Market in Uganda: An

Assessment and a Strategy for Development (Draft Paper).

Feder, G., R.E. Just and D. Zilberman (1985): Adoption of Agricultural Innovation in

Developing Countries: A Survey. In: Journal of Economic Development and Cultural

Change 33 (2).

Government of the Republic of Uganda (2000): Plan for Modernization pf Agriculture:

Eradicating Poverty in Uganda. Government Strategy and Operational Framework.

Hair, J.S., R.E. Anderson, R.L. Tatham, W.C. Black (1998): Multivariate Data Analysis, 5th

Edition, London.

Hanf, C.H. (1989): Agricultural Sector Analysis by Linear Prgramming Models. Kiel.

Hazell, P.B.R. and R.D. Norton (1985): Mathematical Prgramming for Economic Analysis in

Agriculture. New York.

Hoosbeek, M. R., R. B. Bryant, (1992): Towards the Quantitative Modeling of

Pedogenesis. A Review. In: Geoderma 55, 183-210.

Kruseman, G., H. Hengsdijk, R. Ruben, P. Roebeling, J. Bade (1997): Farm Household

Modelling System for the Analysis of Sustainable Land Use and Food Security:

Theoretical and Mathematical Description. DLV-Report No.7.

Latham, M. C. (1997): Human Nutrition in the Developing World. FAO Food and Nutrition

Series 29.

Lynam, J. K., S. M. Nandwa, E. M. A. Smaling (1998): Nutrient Balances as Indicators of

Productivityn and Sustainability in Sub-Saharan African Agriculture. In: Agiculture,

Ecosystems and Environment Vol.71.

McCarl, B. A. and J. Apland (1986): Validation of Linear Programming Models. In: Southern

Journal of Agricultural Economics.

Metcalfe, J. S. (1988): The Diffusion of Innovations: An Interpretative Survey. In: Dosi, G.,

et al. (Ed.): Technical Change and Economic Theory.

Park, S.J. and P.L.G. Vlek (2002): Environmental Correlation of Three-Dimensional Soil

Spatial Variability: A Comparison of Three Adaptive Techniques. (in press)

83

Parrot, L. and R. Kok (2000): Incorporating Complexity in Ecosystem Modelling.

http://www.csu.edu.au/ci/vol07/lparro01

Pender, J., P. Jagger, E. Nkonya and D. Sserunkuuma (2001): Development Pathways and

Land Management in Uganda: Causes and Implications. Paper presented at a

Workshop on Policies on Improved Land Management in Uganda, Kampala, Uganda.

Pomarici, E. and C.H. Hanf (1996): Outline of a Dynamic Mathematical Programming Model

for Regional Analysis of Agricultural Development. Materialsammlung zur

Agrarunternehmenslehre. Kiel.

Principe, J.C., N.R. Euliano, W. C. Lefebvre, (2000). Neural and adaptive systems:

Fundamentals through simulations. John Wiley & Sons, New York.

Renard, K. G., J. M. Laflen, G. R. Foster, D. K. McCool (1994). The Revised Universal Soil

Loss Equation. In: Lal, R. (editor): Soil Erosion: Research Methods, 2nd Edition.

Roe, T. and T. Graham-Tomasi (1986): Yield Risk in a Dynamic Model. In: Agricultural

Household Models: Extensions, Applications, and Policy, ed.: I. Singh, L. Squire and

J. Strauss. Baltimore and London.

Rogers, E. (1995): Diffusion of Innovations. 4.edition. New York.

Ruben, R., H. Menk and A. Kuyvenhoven (1998): Integrating Agricultural Research and

Policy Analysis: Analytical Framework and Policy Applications for Bio-economic

Modeling. In: Agricultural Systems, Vol. 58, No.3.

Sharma, S. (1996): Applied Multivariate Techniques, John Wiley & Sons, New York.

Singh, I., L. Squire and J. Strauss (1986): Agricultural Household Models: Extensions,

Applications, and Policy. Baltimore and London.

Stoorvogel, J.J. and E.M.A. Smaling (1990): Assessment of Soil nutrient Depletion in Sub-

Saharan Africa: 1983-2000. Wageningen.

Upton, M. (1996): The Economics of Tropical Farming Systems. Cambridge.

Valente, Thomas W. (1995): Network Models of the Diffusion of Innovations. Cresskill, New

Jersey.

Valente, T.W. (1996): Social Network Thresholds in the Diffusion of Innovations. In: Social

Networks 18 (1996) 69-89.

Van den Bosch, H., J. N. Gitari, V. N. Ogaro, S. Maobe and J. Vlaming (1998): Monitoring

Nutrient Flows and Economic Performance in African Farming Systems (NUTMON),

In: Agriculture, Ecosystems and Environment Vol. 71.

84

Vosti, S.A. and T. Reardon (1997): The Critical Triangle of Links among Sustainability,

Growth, and Poverty Alleviation. In: Vosti, S.A. and T. Reardon (ed.): Sustainability,

Growth and Poverty Alleviation. Baltimore and London.

Vredeseilanden-Coopibo-Uganda, SNV-Uganda and Oxfam-Uganda (1998): Participatory

Research Carried Out in Arua, Iganga, Kitgum and Pallisa Districts.

Williams, J.R., C.A. Jones and P.T. Dyke (1984): A Modeling Approach to Determining the

Relationship between Erosion and Soil Productivity. Trans.ASAE 27.

World Bank (1993): Uganda Agriculture. Washington.

Wortmann, C.S. and C.K. Kaizzi (1998): Nutrient Balances and Expected Effects of

alternative Practices in Farming Systems of Uganda. In: Agric. Ecosystem.

Environ.115.129.

Documents

Land Management and Technology Adoption in Eastern Uganda