jeffrey vitale gaspard vognan. source: isaaa 2011
TRANSCRIPT
The socio-economic impacts of GM cotton in Burkina Faso:
Does farm structure affect how benefits are distributed?
Jeffrey Vitale
Gaspard Vognan
Bollgard II in Burkina Faso:
< 2003Success -> Stagnation
2003-05 Confined Field Trials Ref: Huma et al. 2007;Vitale et al. 2008
2006 Demonstration Plots
2007 On-farm trials
2008 Limited Commercial release
2009-14 Large-Scale Commercial release
Testing
Legal Framework
Biosafety Protocols
Monitor & Evaluate
Ref: Sustainability paper, Sanders et al., Tom Bassett
Summary of Bollgard II® in Burkina Faso: Documented Findings from Surveys
• Six years of commercial use (2009-2014)– Approaching “full” adoption threshold
• Higher BGII yields in all years (20.5%)• Lower pesticide use (2/3 reduction)• Higher economic returns in all years– Consistent with yield increase since no significant
change in costs (higher seed cost offset by insecticide cost savings)
• Health benefits (self-reported) ($1 million annual)
BGII Adoption Profile
2009 2010 2011 2012 2013 20140
100,000
200,000
300,000
400,000
500,000
600,000
700,000
800,000
129,000
256,000 251,000
312,000
450,000 454,000420,000
386,000429,000
600,000
750,000
650,000
BGII Total
Year
Cott
on P
lant
ed A
rea
(ha)
Conventional Cotton = “Gray” – “Red”
BGII Adoption Profile
2009 2010 2011 2012 2013 2014 Ave0
10
20
30
40
50
60
70
80
30.7
66.3
58.5
52.0
60.0
69.8
57.2
% Adoption
Year
Cott
on P
lant
ed A
rea
(ha)
Roger’s 80% upper limit
Seed Supply Issues
Refugia
BGII vs. Conventional Cotton Yields: 2009-2014
2009 2010 2011 2012 2013 2014 Ave0
200
400
600
800
1,000
1,200
1,400
CV BGII
Year
Yie
ld (
kg/h
a)
Ave Yield Increase = 20.5%
INERA Producer Surveys
b a b a b a b a b a b a b a
Summary of Bollgard II® in Burkina Faso: Documented Findings from Surveys
• Six years of commercial use (2009-2014)• Approaching “full” adoption threshold
• Higher BGII yields in all years (20.5%)• Lower pesticide use (2/3 reduction)• Higher economic returns in all years• Proportional with yield increase since no significant
change in costs: higher seed cost offset by insecticide cost savings → Economic Impact ≈ Pcott*∆Yield
• Significant yield impact → Significant economic impact
• Health benefits (self-reported) ($1 million annual)
Are all farmers benefitting?
• Does location matter?– Modest regional difference but producers in all
zones obtained significantly higher yields growing BGII compared to conventional cotton
• Does “farm size” matter?– “Larger” farms were found to have higher yields but
farms of all size, including smallholder farms, obtained significantly higher yields growing BGII compared to conventional cotton
INERA Farm Type Classification Cotton Production Zone
Yield Item SOFITEX (NBT=109) SOCOMA (NBT=73) Faso Coton (NBT=75) All Zone
(kg ha-1) Large nBT=66 nCV=77
Med nBT=31nCV=47
Manual nBT=12nCV=30
Ave Large nBT=23 nCV=21
Med nBT=42nCV=32
Manual nBT=8nCV=9
Ave Large nBT=14 nCV=15
Med nBT=61nCV=56
Manual nBT=0nCV=0
Ave nBT=257 nCV=287
BG II 1,293A 1,169AB 1,297AB 1,258A 1,192AB 1,286A 1,088ABC 1,235A 1,173AB 954C - 995B 1,175aConventional 1,105B 1,084BC 870C 1,053B 948BC 964BC 1,060ABC 972B 866C 825C - 834C 981bAverage Yield 1,199a 1,127a 1,083b 1,155a 1,070a 1,125a 1,074a 1,103a 1,019ab 890b - 914b 1,078Yield inc (kg ha-1) 188 85 427 206 244 322 27 262 307 129 - 161 193Advantage (%) 17.0 7.9 49.1 19.5 25.7 33.4 2.6 27.0 35.4 15.6 - 19.3 19.7
Inc Rev: $ ha-1102.46 46.44 232.38 111.88 132.58 175.49 14.84 142.69 167.03 70.15 - 87.69 105.30
Farm Size (ha) BGII 5.83 3.29 2.42 4.73 4.60 2.46 1.53 3.03 2.25 1.39 - 1.55 3.32 Conventional 4.26 2.77 2.00 3.37 3.05 2.07 1.67 2.34 1.57 1.08 - 1.18 2.60
Source: Vitale et al. (2010) AgBioforum
Yields by Production Zone
SOFITEX SOCOMA FASO COTON0
200
400
600
800
1,000
1,200
1,400
988900
785
1,1581,090
941
CV BT
Production Zone
Cott
on Y
ield
(kg
per
ha)
b a b a b a
Why the Concern over Farm Size/Farm Structure?
• Welfare of the “smallholder farmer” is explicitly mentioned in Burkina Faso’s biosecurity legal framework
• Welfare of the smallholder farmer overarching theme of CGIAR and many NARS
How can Farms be Classified by Size?
• Biosecurity framework provides no specific definition of “smallholder farm”
• Our analysis has followed the classification used by INERA based on the # of draft animals owned by the household
• Given the importance of addressing the welfare of smallholders, we have been investigating whether another classification could provide a more accurate depiction of farm size
Farm Size: INERA Classification
Manual (hand-hoe) Medium (1 Pair of Bullocks) Large (>1 pair of Bullocks)0
5
10
15
20
25
30
5.3
23.1
19.9
4.6
24.222.8
CV BT
Farm Type (INERA CLassification))
% o
f Far
ms
in E
ach
Clas
s
Farm Size: Yields by INERA Class.
Manual (h
and-hoe)
Medium (1
Pair of B
ullock
s)
Large (>
1 pair of B
ullock
s)0
200
400
600
800
1,000
1,200
1,400
807884
9759541,022
1,189
CV BT
Farm Type (INERA CLassification))
Cott
on Y
ield
(kg
per h
a) A
DC
BC
E
R2 = 0.198
Farm Size: Planted Area
0-0.5 0.5-1 1-3 3-5 5-8 8-10 >100
10
20
30
40
50
60
3.7
17.3
51.7
14.3
8.1
2.4 2.5
%
Cotton Planted Area (ha)
Perc
ent o
f far
ms
(%)
Farm Size: Yields by Planted Area
0-0.5 0.5-1 1-3 3-5 5-8 8-10 >100
200
400
600
800
1,000
1,200
1,400
1,600
912852
901971 976 983
1313
1031 1007 10471141 1173 1180
1368
CV BT
Cotton Planted Area (ha)
Cott
on Y
ield
(kg
per h
a)
Small letters: means testing within each land class
a a
a ab a b ab ab ab a
R2 = 0.187
Farm Size: Household Labor
1-3 4-5 6-10 11-15 >150
5
10
15
20
25
30
35
40
45 42.5
16.7
20.1
4.52.9
Labor_Supp
Household Labor Supply (# persons)
Perc
ent o
f far
ms
(%)
1-3 4-5 6-10 11-15 >150
200
400
600
800
1,000
1,200
1,400
909 917 956 948870
1,050 1,074 1,1111,211
1,159
CV BT
Household Labor Supply (#persons)
Cott
on Y
ield
(kg
per h
a)
Farm Size: Yields by Household Labor
R2 = 0.167
b ab a b a b a b a
Summary of Initial Findings• BGII provided significantly higher yields and economic
returns for all types of farmers, in particular smallholder farmers, using three alternative classifications (#animals, land size, HH labor)
• All three classifications provided about the same level of explanatory power, 17-20%
• On-going research: Would using all three farm structure variables provide a better farm classification?
Empirical Evidence
• Investigate the effect of farm size and farm structure on BG II yield performance using six years of cotton producer survey data
• Test whether farm size/farm structure related variables have a significant effect on cotton yield:– Farm size (area)– Household labor – Number of bullocks
Empirical Model Structure
• Y = α0 + α1Year + α2Type + α3Zone + α4Sprays + β1Animals + β2Area + β3Labor + interactions
• Variables:– Type BGII or conventional– Year 2009-2014– Zone SOFITEX,SOCOMA, or Faso Coton– Sprays Late season sprays (secondary pests)– Animals # of working animals (bullocks)– Area # hectares of cotton planted by household– Labor # of household members working on-farm
Model ResultsSource Estimate Pr>F Estimate Pr>F Estimate Pr>F
Intercept 853.7 <.0001 854.9
Year … <.0001 <.0001 <.0001
Type_Coton (BT=1) 139.62 <.0001 146.41 <.0001 <.0001
Year*Type_Coton … 0.0018 0.002 0.0048
Zone … <.0001 <.0001 <.0001
Type_Coton*Zone … 0.0102 0.0176 0.0358
Sprays … 0.118 0.1637 0.1121
Type_Coto*Sprays … 0.0014 0.0015 0.0465
Labor 0.081643 0.9423 -1.53464 0.4694 -0.33358 0.8609
Land 18.65116 <.0001 16.78885 0.008 15.93261 <.0001
Animals 26.68308 <.0001 29.64985 <.0001 32.24178 <.0001
Labor*Land 0.674873 0.0141
Land*Animals -0.59433 0.6084
Labor*Animals -0.32312 0.5855
Labor/Land 0.310934 0.9114
Animals/Land -11.5453 0.0777
R2
Model 1 Model 2 Model 3
0.212 0.215 0.229
Model Results: Estimated BGII Yield Function: Land, Labor, and Animals
0 2 4 6 8 10 12 140
200
400
600
800
1,000
1,200
1,400
0
2
4
6
8
10
12
BGII Yield HH Labor #Bullocks
Area of Cotton Planted (ha)
Cott
on Y
ield
(kg
per h
a)
HH
Lab
or (#
per
sons
) & #
Bul
lock
s
3 animals
4 laborers
1,100 kg yield
4 animals
8 laborers
1,250 kg yield
Practical Implication
• Research findings suggests that the prior definition used to define farm types, based only on # bullocks, is as good as alternative classifications using land and labor.
• Provide policy makers, and the on-going legal framework, with an alternative approach to define what a “smallholder” producer is , i.e. include a “3-D” definition.
The End
Does Farm Size Affect BGII Impact?
• BG II is a scale neutral technology:– Control effectiveness of BGII independent of field size
(~95%) – No new equipment needed (or that could be sold) – Seed cost on a per ha basis (no scale effect)– Insecticide costs on a per ha basis (no scale effect)
• Farm size and farm structure can affect yield performance …
How Farm Size Can Matter: Stylized Facts
• Economy of scale: larger farms can more easily cover fixed costs compared to smaller ones– Higher profits earned by larger farms
• Wealth Effect: Higher profits enable larger (& better managed) farms to make more investments in equipment (e.g. animal traction) and resources– Bigger farms are wealthier & better equipped than
smallholders who remain resource constrained– Greater efficiency, risk mgmt easier, access to capital
How Farm Size Can Matter: Stylized Facts
• “Rich get richer” while smallholders remain trapped in subsistence farming – Increase land holdings, access to quality lands, and
political power while smaller producers are pushed to the margins
Empirical Evidence
• Investigate the effect of farm size and farm structure on BG II yield performance using six years of cotton producer survey data
• Test whether farm size/farm structure related variables have a significant effect on cotton yield:– Farm size (area)– Household labor – Number of bullocks
Scale Effects in Cotton Production• Farm equipment (+)
– Manual farms vs. animal powered vs. mechanized• Deeper plowing, increased speed of operation for bigger farms
• Household farm labor (+/-)– Big farms likely to have larger workforce but also larger field size – Small farms could have more labor per ha but likely have greater labor demand since
they are less well equipped– So this variable is difficult to predict a priori and likely to depend on other variables
(interaction terms)
• Farm Size (+/-) – Larger farms are more difficult to manage since they are more complex and have
larger area, e.g. pest scouting and nutrient management – Larger farms likely better equipped and more efficient– So this variable is difficult to predict a priori and likely to depend on other variables
(interaction terms)
Model Structure• Test alternative regression models and identify which
variables are significant and which model best fits the data
• Include interaction terms and create new variables to place farm structure variables on a unit basis, e.g. labor per ha
• Include other variables to explain cotton yield:– Year– Zone– Insecticide sprays
Empirical Model Structure
• Y = α0 + α0Zone + α0Zone + α0Sprays + β1Area + β1 + α0Year
• Variables:– Type (BGII or conventional)– Year– Zone– Insecticide sprays– Village
Model ResultsSource Estimate Pr>F Estimate Pr>F Estimate Pr>F
Intercept 853.7 <.0001 854.9
Year … <.0001 <.0001 <.0001
Type_Coton (BT=1) 139.62 <.0001 146.41 <.0001 <.0001
Year*Type_Coton … 0.0018 0.002 0.0048
Zone … <.0001 <.0001 <.0001
Type_Coton*Zone … 0.0102 0.0176 0.0358
Late_spray_Cat … 0.118 0.1637 0.1121
Type_Coto*Late_spray … 0.0014 0.0015 0.0465
Actifs_agricole 0.081643 0.9423 -1.53464 0.4694 -0.33358 0.8609
Surface_parcelle 18.65116 <.0001 16.78885 0.008 15.93261 <.0001
Nombre_animaux_trait 26.68308 <.0001 29.64985 <.0001 32.24178 <.0001
Actifs_ag*Surface_pa 0.674873 0.0141
Surface_p*Nombre_ani -0.59433 0.6084
Actifs_ag*Nombre_ani -0.32312 0.5855
Actifs_ha 0.310934 0.9114
Animals_ha -11.5453 0.0777
R2
Results
0 2 4 6 8 10 12 140
200
400
600
800
1,000
1,200
1,400
0
2
4
6
8
10
12
BGII Yield HH Labor #Bullocks
Cott
on Y
ield
(kg
per
ha)
HH
Lab
or (#
per
sons
) & #
Bul
lock
s
Two-Dimensional View of Smallholder Farms
Lighter colors represent larger residuals
Practical Implication
• Research findings suggests that the prior definition used to define farm types, based only on # bullocks, is not the best one, but is still consistent with our more general findings.
• Provide policy makers, and the on-going legal framework, with an alternative approach to define what a “smallholder” producer is , i.e. include a “3-D” defintion.
Conclusions/Policy Implications• Smallholder farmers benefit the same as larger, better equipped
farms on a proportional basis no matter how “smallholder” is defined
• Larger, better equipped farms have higher yields and do achieve higher overall yield and economic benefits
• ALL farm types and size perform significantly better with BGII than conventional cotton
• Policy makers need to focus assisting farmers to become better equipped and to utilize increased profitability of BGII cotton to invest in farm equipment
• More efficient farms are expected to improve yields as suggested by the survey results. – Larger farm sizes will be an outcome of the increased farm capital but
increasing farm size just for the sake of larger farms will not increase yields.
Farm TypeYear Manual Small Large Motorise Ave2009-2011 BGII 1012aC 1064aC 1207aB (1555A) 1094a
CONV 915aA 909bA 973bA - 933bDiff 97 155 234 - 128% Diff 10.6 17.1 24.0 17.2
2012-2013 BGII 978aB 1028aB 1162aA - 1056aCONV 882aB 863bB 947bA - 897bDiff 96 165 215 158% Diff 10.9 19.1 22.7 17.7
2014 BGII 782aD 962aC 1144aB (1310aA) 962aCONV 760aC 872bB 979bA (1000bA) 870bDiff 22 90 165 (310) 92% Diff 2.9 10.3 16.9 (13.1) 10.5
Ave2009-2014
BGII 953aD 1021aC 1174aB (1352aA) 1049a
CONV 860bB 887bB 979bA (1000bA) 908bDiff 92 135 196 (352) 141% Diff 10.8 15.1 19.9 (13.5) 15.5
2009/2010 - 2014/2015 – Yields*old farm type (animals)*type of cottonSofitex + Socoma + Faso Coton
Notes statistical analysesMean separation indicated by letters a,b,c is comparing cotton types within same farm typeMean separation indicated by letters A,B,C is comparing farm types within same cotton type
ALL FARM TYPES BENEFIT FROM GROWING BOLLGARD II®
2009/2010 – 2013/2014BGII yield benefit for all field sizes
Higher yieldsMore consistent yield
2009/2010 – 2013/2014BGII yield benefit for all HH Labor
Higher yieldsMore consistent yield
Better Way to Classify Farms
Monsanto Company Confidential
Conclusions
This presentation focused on cotton yield impact of BGIIThe close proximity of the average production costs of BGII cotton and conventional cotton indicates that the primary source of the increase in cotton profit from growing Bollgard II across years (2009/2010 – 2011/2012) was from a combination of the yield increase and the higher cotton price that placed a greater value on output compared to the previous two years.
Other socio-economic benefits directly linked to yield are available but not discussed in this presentation - Economic return ($/ha)
Cotton income – Cotton Production cost- Household income (related to hectares of BGII)- Return to labor ($/day)- more consistent production => this will become visible in presentation
(less variable, target pest control)
Consistent benefits not directly linked to yield are available but not discussed in this presentation- improved human health (and related reduced health care cost)
reduced insecticide exposure (2 treatments vs 6 treatments)chemical storage, preparations spray solutions,exposure during applicationswaste handling
- Reduced environmental impact from reduced insecticide usage- labor saving and the related time spending
Further Research
• Is there more land available in all villages?• Is additional training needed for increasing
animal traction