Upload
international-water-management-institute-iwmi-cgiar-water-land-and-ecosystems-program
View
2.281
Download
2
Tags:
Embed Size (px)
DESCRIPTION
by A. Zewdie, J. Cadilhon and C. Werthmann Presented at the Final Volta Basin Development Challenge Science Workshop, September 2013
Citation preview
A Partner of
Impact of Innovation Platforms on Marketing Relationships: The Case of Volta Basin Integrated
Crop-Livestock Value Chains in Ghana
Volta Basin Development ChallengeFinal Scientific Workshop
Ouagadougou, Burkina Faso17-19 September, 2013
Zewdie A., J. Cadilhon and C. Werthmann
Andes • Ganges • Limpopo • Mekong • Nile • Volta
Content1. Main message
2. Background and Purpose
3. Methodology
4. Results
5. Lessons
6. Recommendations
2
Andes • Ganges • Limpopo • Mekong • Nile • Volta
Main messages
The IPs have created an additional option for value chain actors
The interaction through the platforms contributes to reduction in transaction costs and improvement in access to markets
No sizable market niche has been created although there is a potential for the future
3
Andes • Ganges • Limpopo • Mekong • Nile • Volta
1. Background and purpose
The changing nature of Agricultural ResearchAgricultural innovations as multi-dimensional and
co-evolutionary processes technological, organizational and institutional
innovations
creating synergies “convergence of agricultural sciences” (Huis et al.
2007)
Growing use of Innovation Platform (IP) approaches
4
Andes • Ganges • Limpopo • Mekong • Nile • Volta
What is an Innovation Platform?
Source: PAEPARD, October 2012
IP meeting in Lawra, June 27, 2013
5
Andes • Ganges • Limpopo • Mekong • Nile • Volta
The Volta2 Innovation Platforms
Ghana Volta2 IPs were established in July 2011
Main goals of the Volta2 IP project:
to facilitate communication and collaboration among
actors
to support value chains development
to serve as spaces for Participatory Action Research
(PAR)
6
Andes • Ganges • Limpopo • Mekong • Nile • Volta
Purpose of the study
to assess the interrelationships between various
actors
to investigate the impact of communication and
information sharing on market access
test a new conceptual framework for evaluation of
IPs
7
Andes • Ganges • Limpopo • Mekong • Nile • Volta
Why to use a new framework?
Limitations with conventional methods
difficulty to check cause-effect relationships
only few econometric methods
E.g. DID and IV methods, but not applied to IPs (and use control
groups)
Statistics and record on agricultural data in LDCs is poor
Perception of respondents based on Likert-scale as an
alternative?
8
Andes • Ganges • Limpopo • Mekong • Nile • Volta
2. Methodology
Data Four communities (two IPs)
Lawra: Upper West region Orbilli and Naburinye
Tolon-Kumbungu: Northern region Digu and Golinga
43 IP members 34 farmers, 6 traders, 3 processors
9 key respondentsSource: Diamenu and Nyaku 1998
9
Andes • Ganges • Limpopo • Mekong • Nile • Volta
Data cont…
Socio-economic information and 5 point Likert-scale data from IP members (1 = strongly disagree, 2 = disagree, 3 = undecided, 4 = agree, 5 = strongly agree)
Assumption: 5-4 = 4-3 = 3-2 = 2-1
10
Simplifying the Likert-scale, to ensure better response quality
Andes • Ganges • Limpopo • Mekong • Nile • Volta
Method of analysis
Mixed methods using a new framework
Qualitative
Quantitative response summaries and averages on various
statements factor analysis on the Likert-scale variables
Principal components factor
regression Ordinary Least Squares (OLS)
11
Andes • Ganges • Limpopo • Mekong • Nile • Volta
The conceptual framework
Based on a mix of concepts from various disciplines
Industrial Economics
Structure-Conduct-Performance (SCP) hypothesis
New Institutional Economics of Markets
transaction costs
governance structures Concepts from the marketing literature
value chain relationship12
Andes • Ganges • Limpopo • Mekong • Nile • Volta
“A priori linear relationship between Structure, Conduct and Performance”
The SCP hypothesis
13
Andes • Ganges • Limpopo • Mekong • Nile • Volta
The regression model
Semi-logarithmic multiple regression
14
Andes • Ganges • Limpopo • Mekong • Nile • Volta
Explanation of terms marketaccessj = the jth dependent variable of market access IP = a dummy variable that assumes 1 for Tolon-Kumbungu and 0
for Lawra to account for possible differences between the two IPs gender = 1 for male and 0 for female age = age of the IP member lnnbhous = natural logarithm of household size Incestm = annual income of the participant communicationi = the ith variable that represents the level of
communication and information sharing of by an IP member , , ….. are partial effects of each respective explanatory variables
on market access. is an intercept term is the error term or model residual i and j depend on the outcome of the factor analysis
15
Andes • Ganges • Limpopo • Mekong • Nile • Volta
4. Results
16
Andes • Ganges • Limpopo • Mekong • Nile • Volta
17
More descriptive
Andes • Ganges • Limpopo • Mekong • Nile • Volta
Three factors* on communication and information sharing
Four factors* on market access
Results of the factor analysis
*Based on the Kaiser criteria is used
18
Andes • Ganges • Limpopo • Mekong • Nile • Volta
Results for regression equation 1Regression Equation
Dependent Variable
Explanatory variables
Coefficient Beta t P>|t|
1
factor11 IP 0.0638 (0.617) 0.032 0.10 0.918 gender 0.2767 (0.349) 0.139 0.79
0.436
lnnbhous -0.0182 (0.542) -0.007 -0.03
0.973
age -0.0014 (0.011) -0.020 -0.13
0.896
incestm -0.0003 (0.000) -0.240 -1.19
0.247
focq50i 0.5782 (0.255)
0.365** 2.26
0.032
factor1 0.1543 (0.256) 0.156 0.60
0.553
factor2 -0.0642 (0.231) -0.069 -0.28
0.783
factor3 -0.1543 (0.619) -0.157 -0.95
0.349
constant -2.1627 (1.472)
. -1.47
0.154
- Robust) standard errors are shown in brackets and betas are standardised coefficients. - * and ** represent statistical significance of betas at 1% and 5% levels of significance - focq50i = individual statement about improvement in communication in the past 2 years
19
Andes • Ganges • Limpopo • Mekong • Nile • Volta
Results for regression equation 2
- Robust) standard errors are shown in brackets and betas are standardised coefficients. - * and ** represent statistical significance of betas at 1% and 5% levels of significance - focq50i = individual statement about improvement in communication in the past 2 years
Regression Equation
Dependent Variable
Explanatory variables
Coefficient Beta t P>|t|
2
factor12
IP 0.6432 (0.443) 0.324 1.45
0.159
gender -0.2612 (0.296) -0.131 -0.88
0.387
lnnbhous 0.1248 (0.242) 0.054 0.51
0.611
age -0.0122 (0.012) -0.169 -0.96
0.344
incestm -0.0001 (0.000) -0.026 -0.17
0.867
focq50i -0.1968 (0.258) -0.124 -0.76
0.453
factor1 0.2535 (0.2252) 0.257 1.13
0.271
factor2 0.3339 (0 .117)
0.359* 2.84 0.009
factor3 -0.0460 ( 0.142) -0.047 -0.32
0.749
constant 1.068 (1.3416) . 0.80
0.433
20
Andes • Ganges • Limpopo • Mekong • Nile • Volta
Results for regression equation 3
- Robust) standard errors are shown in brackets and betas are standardised coefficients. - * and ** represent statistical significance of betas at 1% and 5% levels of significance - focq50i = individual statement about improvement in communication in the past 2 years
Regression Equation
Dependent Variable
Explanatory variables
Coefficient Beta t P>|t|
3
factor13 IP -0.3810 (0.553) -0.192 -0.69
0.497
gender -0.8305 (0 .381)
-0.418**
-2.18 0.039
lnnbhous 0.0440 (0.422) -0.418 0.10
0.918
age -0.0228 (0.014) 0.019 -1.55
0.132
incestm 0.0002 (0.000) -0.314 0.71
0.486
focq50i 0.3342 (0.322) 0.122 1.04
0.308
factor1 0.3007 (0.272) 0.305 1.10
0.279
factor2 0.0222 (0.162) 0.023 0.14
0.892
factor3 -0.1405 (0.223) -0.143 -0.63
0.534
constant 0.01651 (1.864) . 0.01
0.993
21
Andes • Ganges • Limpopo • Mekong • Nile • Volta
Results for regression equation 4
- Robust) standard errors are shown in brackets and betas are standardised coefficients. - * and ** represent statistical significance of betas at 1% and 5% levels of significance - focq50i = individual statement about improvement in communication in the past 2 years
Regression Equation
Dependent Variable
Explanatory variables
Coefficient Beta t P>|t|
4 factor14 IP 1.8330 (0.402)
0.923* 4.55
0.000
gender -0.2039 (0.297) -0.102 -0.69
0.499
lnnbhous -1.0078 (0.429)
-0.438**
-2.35
0.027
age 0.0123 (0.011) 0.170 1.14
0.265
incestm 0.0006 (0.000)
0.449* 3.02
0.006
focq50i -0.0157 (0.228) -0.009 -0.07
0.946
factor1 -0.1235 (0.165) -0.125 -0.75
0.463
factor2 -0.3224 (0.137)
-0.347**
-2.35
0.027
factor3 -0.0318 (0.118) -0.032 -0.27
0.790
constant 0.4784 (1.324) . 0.36
0.721
22
Andes • Ganges • Limpopo • Mekong • Nile • Volta
Regression results
Improvement in access to input and output markets is related to improvements in overall communication or interaction
Women have better access to market than men
Participants in Tolon-Kumbungu IP have better market access than those in Lawra: same information from baseline survey?
The higher the annual income the higher the level of market access
(log) household size (local indicator of wealth) is negatively related to market access
Use of the media reduces the likelihood of bypassing intermediaries although it improves access to market information and hence to markets
23
Andes • Ganges • Limpopo • Mekong • Nile • Volta
Results from qualitative information
Improved knowledge because of trainings and meetings
trainings by urban traders and processors to their rural counterparts and to farmers
value chain training by a marketing expert from ILRI
Knowledge on price standardization, commercialization and use of weighing scales
New, but limited additional market options because of the IP
members met potential trade partners due to the IP meetings
farmers and traders check prices before engaging in transactions
24
Andes • Ganges • Limpopo • Mekong • Nile • Volta
Results from qualitative information cont…
Improved knowledge on crop and livestock production
Improved post-harvest management and timing of sale
Knowledge on cooperatives
Shelter for small ruminants, constructed after a training under the IP project : at Digu village
25
Andes • Ganges • Limpopo • Mekong • Nile • Volta
5. Lessons
The IPs played a ‘role’ in improving communication and
information sharing and opened new options
Proximity to market centers and level of income of the
members seems to be key determinants of access to
market
So, were the IPs helpful to the communities?
It is not time to judge whether the IPs had significant
impact
26
Andes • Ganges • Limpopo • Mekong • Nile • Volta
6. Recommendations
IPs should not be left alone at this stage
further education on commercial production, marketing,
record keeping, cost benefit analysis and business plan
development
more effort to link the value chain actors to engage in
commercial transaction and hence creating new markets
further efforts on cooperative formation
linking them to other development partners
Evaluating the overall impact on livelihoods is also
required
27
Andes • Ganges • Limpopo • Mekong • Nile • Volta
Annex
28
Andes • Ganges • Limpopo • Mekong • Nile • Volta
29
Name of factor
Statements contributing to the variances in the respective factors representing communication and information sharing
Remark (assigning name to the factors)
Factor1 I exchange information with my value chain partners about my on-going activities
Information sharing?
My value chain partners exchange information about their on-going activities with me
Factor2 I listen to weekly radio announcements to get market information
Using media to acquire information/better communication?I am satisfied with the quality of communication I
was having with my business partners in the last two years
Factor3 I am satisfied with the communication frequency I had with value chain actors in recent business relationships
Frequent communication to obtain market information?I ask relatives and friends in the village for market
information
Underlying factors for communication and information sharing
Andes • Ganges • Limpopo • Mekong • Nile • Volta
30
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43
5 5 5 5 5
4
5 5
4 4 4
5
4 4
5 5 5 5 5
4
5 5
4
5
3
4 4
1
2
1
5
3
5
3
4 4
3
4
3
4 4
3
4
5
4
5 5 5
4
4
5
4
5
4
5
5
4
5 5 5 5 5
4
5
4
4
4
4
1
4
3
1 2
4
3
5
3
4 4
3
3
3
4 4
3
4
Relationship between responses on statements 28_a (I exchange in-formation with my VC partners about my ongoing activities) and 28_b (my VC partners exchange information about their ongoing activities
with me) which together constitute Factor1 (information
28_a 28_b
Andes • Ganges • Limpopo • Mekong • Nile • Volta
31
Name of factor
Statements contributing to the variances in the respective factors representing market access
Remark (assigning name to the factors)
Factor11 The number of marketing companies buying products from the villagers has increased in the past two years
Improved access to input and output markets?My access to input markets has improved in the past two
years My access to output market has improved in the past two years
Factor12 Information on the market is easily accessible to value chain actors
Better access to market information?Farmers in the IP negotiate with buyers as a group
Factor13 I can now better negotiate market prices than two years ago Improved negotiation for better price?I am satisfied by the prices I get from my customers for my
products
Factor14 I sell my output directly to processers or consumers Bypassing market intermediaries?There is a ready market for farm produce during harvesting
seasons in my area
Underlying factors for market access
Andes • Ganges • Limpopo • Mekong • Nile • Volta
32
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43
5
4
5 5
4 4
5
4
5
4 4
5 5
4
3
4 4 4 4 4 4
3
5 5 5 5
4
3
5
4
5
3
4
2
4 4 4 4 4
5
4 4 4
5
3
5 5
5
4
5
4
5
5 5
5
4
4
3
4
2
4
1
4 4 5
4
5 5 5
5
4
4
4
4
4
4
4
4
2
4 4 4
5
3
5
4
Relationship between responses FocQG_55m (my access to input market has improved in the past two years) and FocQG_55n (my
access to output market has improved in the past two years)
FocQG_55m FocQG_55n
Andes • Ganges • Limpopo • Mekong • Nile • Volta
33
Factor analysis
Kaiser-Meyer-Olkin (KMO) Measure of Sampling Adequacy
Bartlett’s test of sphericity
Cronbach’s Alpha
Chi-square p-value
Conduct 0.748 142.887* 0.000 0.81Performance 0.641 93.161* 0.000 0.72 H0: variables are
not intercorrelated
• * implies that the test rejects the null hypothesis at the 1% level of significance
• KMO > 0.6; Cronbach’s alpha > 0.7 and P-value < 0.05 for Bartlett’s tests all suggest that conducting factor analysis is appropriate both for Conduct and Performance indicators
Tests of factorability and reliability
Andes • Ganges • Limpopo • Mekong • Nile • Volta
34
Variable VIF Tolerance = 1/VIF Variable VIF Tolerance = 1/VIF
IP 2.84 0.352489 lnnbhous 1.24 0.808407
factor1 2.60 0.385266 age 1.22 0.822854
factor2 1.40 0.712389 incestm2 1.21 0.823299
gender 1.32 0.757049 factor3 1.13 0.888709
focq50i 1.29 0.776226 Mean VIF = 1.58
VIF < 5 implies absence of serious multicollinearity problem
Multicollinearity test for explanatory variables using VIF
Andes • Ganges • Limpopo • Mekong • Nile • Volta
35
Shapiro-Wilk W test for normality Breusch-Pagan / Cook-Weisberg test Variable W V Z P>Z Variable chi2(1) P>chi2Resid1 0.957
1.794
1.235 0.108 fitted values of factor11
0.38 0.539
Resid2 0.946
2.256
1.719**
0.042 fitted values of factor12
3.99** 0.045
Resid3 0.941 2.434 1.880** 0.030 fitted values of factor13
1.32 0.249
Resid4 0.964
1.464
0.805 0.210 fitted values of factor14
0.16 0.687
Ho: error term is normally distributed Ho: dependent variable has constant variance
- ** implies that the test rejects the null hypothesis at the 5% level of significance.- Resid refers to the residuals of the corresponding regression equations- No serious deviation from normality when the 1% level of significance is
considered
Tests of variance equality and residual normality in each of the four equations
Andes • Ganges • Limpopo • Mekong • Nile • Volta
36
Regression Equation no.
Dependent Variable
F-value
Prob > F R-squared
1 factor11 0.35 0.7920 0.33242 factor12 0.54 0.6584 0.50783 factor13 0.43 0.7315 0.27924 factor14 0.42 0.7396 0.5264 Ho: model has no omitted variables
All the four models are well specified, no serious problem of omitted variables
R-square is quite low particularly in equation 3
Ramsey regression equation error specification test (RESET) and overall fit of the models
Andes • Ganges • Limpopo • Mekong • Nile • Volta
Pictures from focus group discussions
37
Andes • Ganges • Limpopo • Mekong • Nile • Volta
Thank you!
Also credit to the contributors, CPWF, CSIR-ARI, SNV, ILRI and others
38