Post on 23-Feb-2016
description
transcript
Determinants of the GM cotton adoption: evidences for Brazil
Paper presented in the XVI ICABR Conference- Ravello, Italy 25-27 June -2012.
Alexande Gori Maia- IE/UnicampBruno C. B. Miyamoto – IE/UnicampJosé Maria F.J. da Silveira – IE/Unicamp
2
cotton in Brazil• High growth of quantity due to increasing
productivity;• High price variability;
Graphic 1 - Index for acreage, quantity produced and production value (R$/t) of cotton in
Brazil (1994 = 100)
Source: IBGE - Municipal Agricultural Production
3
Spatial Distribution of Cotton• Mato Grosso and Bahia accumulate the
highest shares of total production (49% and 34%, respectively);
Figure 1 - Spatial distribution of the quantity produced (circles, in tons) and productivity
(colors, in tons/ hectare) of cotton - Brazil, 2010
Source: IBGE - Municipal Agricultural Production Elaborated with Phicarto. Available http://philcarto.free.fr/.
First GM cotton approval: 2005, Mon 810 (Cry1AC) and three years after,
HT Liberty link varietiesMain reason for a slow diffusion:
• Duopal and Nuopal were not considered good varieties in quality for growers and industry;
Alabama argilaceaDelay in the approvals of other Bt Genes, like Cr2Ab +
Cry1Ac), Vip3A e (Cry1Ac +Cry1F).Cry1F)- control of Spodoptera frugiperda and plusídios
I. Brazil: GM cotton in main production areas
Some features
Panels in main cotton regions
• Evaluate cost differentials of GM cotton in Brazil, 2010/2011 harvest.
• Comparison between conventional x GM production systems.
• Additionally, evaluate the potential general equilibrium impacts.(we skip this in the paper)
Methodology• First step: field survey by CEPEA(ESALQ-USP) on the
main cotton production regions. Cost differentials. – Sorriso (MT)– Campo Novo Parecis (MT)– Campo Verde (MT)– Mineiros (GO)– Luiz Eduardo Magalhães (BA).
• Surveys: “Panel” method. Costs and area with GM and conventional varieties estimates.
Survey results: adoptionRegion State Season Row spacing GM use
Luiz Eduardo Magalhães BA First harvest conventional No
Mineiros GO First harvest conventional Yes (LibertyLink) Second harvest conventional Yes (LibertyLink)
Primavera do Leste MT First harvest conventional No
Campo Verde MT First harvest conventional No Second harvest conventional Yes (LibertyLink)
Campo Novo do Parecis MT First harvest conventional No Second harvest Second harvest
conventional narrow
Yes (LibertyLink) Yes (LibertyLink)
Lucas do Rio Verde / Sorriso MT First harvest conventional No Second harvest Second harvest
conventional narrow
No Yes (LibertyLink)
Region 1st. Harvest 2nd. Harvest Conventional GM Conventional GM
Sorriso/MT 100% 0% 50% 50% Campo Novo do Parecis/MT 100% 0% 8% 92% Campo Verde/MT 100% 0% 0% 100% Mineiros/GO 60% 40% 60% 40% Luiz Eduardo Magalhães/BA 100% 0% 0% 0%
GM adoption in regions
• Only herbicide tolerant (HT) GM cotton in use.– Mineiros: low potential of GM varieties.– Luiz Eduardo: low potential, BT cotton do not control
other common pests in the region.• Only Mineiros used GM cotton in the first harvest.• Use of HT in the second harvest to facilitate weed
control.• Low availability of seeds reported for this year.• No price differentiation for the fiber.
Cost comparison, R$/ha. No GM cotton in Bahia.
Cost item Cotton 1st Harvest
Cotton 2nd. Harvest
MNR
MNR CNP OGM NOGM
OGM NOGM OGM NOGM
Fertilizers 875.05 875.05 551.55 551.55 759.60 759.60 Chemical inputs 739.38 847.06 541.53 693.21 690.83 669.26 Herbicides 270.24 377.92 186.80 338.48 218.89 197.33 Insecticides 369.04 369.04 304.68 304.68 400.64 400.64 Fungicides 100.10 100.10 50.05 50.05 71.29 71.29 Seed treatment 0.00 0.00 0.00 0.00 0.00 0.00 Seeds 117.94 79.04 117.94 79.04 81.00 42.00 Emulsionable oil 95.66 95.66 95.66 95.66 39.65 39.65 Mechanical operations 406.54 417.11 397.88 408.45 335.33 367.80 Transportation 0.00 0.00 0.00 0.00 0.00 0.00 Labor 111.97 188.40 106.41 182.84 295.63 358.10 Trade/Storage 795.27 795.27 745.81 745.81 648.23 648.23 Taxes 148.66 148.66 141.23 141.23 202.80 202.80 Insurance 18.21 18.83 19.70 20.32 19.86 22.05 Technical assistance 46.93 50.05 36.22 40.21 44.04 44.73 Interest over capital 233.85 250.63 312.50 336.92 228.86 232.44 CO 3589.44 3765.76 3066.41 3295.24 3345.83 3386.65 COT 3825.54 4011.77 3327.86 3566.61 3600.74 3669.99 CT 4195.73 4389.96 3722.82 3969.57 4053.51 4145.69
Operational costs differentials GM – conventional
(Important reduction in labor)
-8.00%
-7.00%
-6.00%
-5.00%
-4.00%
-3.00%
-2.00%
-1.00%
0.00%
MNR-GOFirst harvest
MNR Second harvest, 0.76 m.
Campo Novo dos Parecis
Profitability comparison
MNR MNR CNPAlgodão Safra Algodão 2ᵃ safra 0,76 m
0.0%
10.0%
20.0%
30.0%
40.0%
50.0%
60.0%
70.0%
80.0%
54.0
% 64.9
%
68.1
%
47.2
%
54.7
% 64.4
%
OGM Convencional
Final remarks: Brazil main production areas
• GM cotton is a labor saving technology.• Labor will be released from cotton to other
economic activities, with beneficial effects under full employment.
• Regions that do not adopt the GM technology will tend to reduce production.
14
Spatial Distribution of Cotton• Where are the small holders?
Figure 1 - Spatial distribution of the quantity produced (circles, in tons) and productivity
(colors, in tons/ hectare) of cotton - Brazil, 2010
Source: IBGE - Municipal Agricultural Production Elaborated with Phicarto. Available http://philcarto.free.fr/.
Cotton in Brazil: small holders share in 2006
Less than 100 ha 11 997 41 971 40 915 36 696 36 821
% of the total 90 2 2 1 5
Categories Number of properties
Quantity produced (ton)
Quantity Sold (ton)
Value ( 1000 R$ 2006)
Harvested area (ha)
Review of literature
Papers on seed adoption
• HUBBELL et al 2001; Bt cotton in US: revealed and declared preferences;
• FERNANDEZ-CORNEJO et al, 2001): Bt and Precison Farming;
• QAIM e de JANVRY, 2003 Bt cotton in Argentina (Willingness to pay; even commercial growers would like to pay half a royalty value. They conclude that profits would be higher if the firm reduce seed prices);
• KOLADY e LESSER, 2006;- Bt eggplants (Hybrid and open pollinated varieties, imperfect substitutes): production systems conditioning adoption;
• BREUSTEDT et al, 2008; Colza GM (two types) and a conventional variety,( probit multinomial, ex-ante);
17
Survey details• Sample: 175 small producers (aprox. 1,5% of the population)
• Coverage: States of Bahia, Paraíba, Rio Grande do Norte, Minas Gerais and Goiás;
This is a “first year” analysis
Typology• Descriptive analysis:
Multiple Correspondence Analysis (ACM) in order to identify patterns of association among types of cotton and farmers’ characteristics.
What is the preference for cotton varieties?
• Ex-post analysis: Conjoint Analysis in order to estimate utilities for each characteristics of cotton production.
19
Multiple Correspondence Analysis
Dimension 1(16% of the total variability)
Negative Values
Positive Values
Area up to 1 ha Absence of boll weevil
Organic Cotton
Dimension 2(9% of the total variability)
Dimension 3(8% of the total variability)
Bt cottonArea between 2 – 5 ha
Pink BollwormLow use of pesticide
HT CottonSilverleaf
No mechanization
Area over 5 ha Mechanization
Area up to 1 ha Permanent Employee
White Cotton
Access to credit
ACM: methodology• From a contingency list with multiple combinations of qualitative categories, MCA
determines the number of relevant dimensions to understand the structure of associations among the categories of analysis (GREENACRE, 1984).
• The MCA is based on the technique of principal components to simplify the data structure, (CUADRA, 1981).
• The technique decomposes the structure of the distances between the categories of interest (distances c2) in (i) eigenvalues representing the partial contributions of each dimension to the total variability, and (ii) eigenvectors representing geometric projection planes of the sub-populations characteristics (GREENACRE & HASTIE, 1987).
• The total inertia represents the average degree of separation of the multiple combinations of frequencies in relation to the average behavior of the population. The K eigenvalues l1, ...,lK resulted from the decomposition of the total inertia are called main inertia and correspond to the partial contributions of the respective dimensions.
• The geometric dispersion of the categories in the space defined by the dimensions of correspondence analysis shows the nature of associations between qualitative variables of the problem.
• Groups of categories close together reveal similarities in the patterns of associations, while groups far apart mean repulsion between the categories (HOFFMANN & FRANKE, 1986).
21
Multiple Correspondance Analysis
Figure 2 - Distribution of categories in the three main dimensions of the MCA
Source: Research data
• Pattern 1: Organic + small area + low education + no credit;• Pattern 2: Colored + no mechanization + no employee;• Pattern 3: White + higher education + mechanization + employee;• Pattern 4: Bt + area 2-5 ha + credit + lower use of pesticide;• Pattern 5: HT + Budworm + Silverleaf whitefly;
Table 2 – Utilities based on the non-metric conjoint analysis of the ordered-rank evaluations
Variable ̂ ̂S ˆ
ˆ S
Intercept 2,336 0,051 45,681
Bt Pirate -0,179 0,115 -1,555
Bt R$20/ha 0,328 0,138 2,370
Bt R$55/ha -0,298 0,313 -0,950
Bt R$90/ha -0,533 0,121 -4,418
RR Pirate 0,154 0,145 1,064
RR R$20/ha -0,102 0,174 -0,587
RR R$55/ha 0,044 0,169 0,261
RR R$90/ha -0,473 0,129 -3,661
Organic 0,987 0,102 9,634
White conventional 0,028 0,127 0,221
Colored 0,043 0,140 0,311 Source: Research data
C.A. methodology• Data were collected by asking farmers about their preferences
for different characteristics of cotton systems. • The CA decomposed rank-ordered evaluation judgments of
cotton systems into components based on qualitative cotton characteristics.
• For each characteristic of interest, a numerical "part-worth utility" value was computed. The sum of the part-worth utilities for each product is an estimate of its utility.
• The aim is to compute part-worth utilities in such a way that the product utilities are as similar as possible to the original rank ordering.
The methodology applied• Two attributes of interest were considered: type of cotton and price (Table 1). • The type of cotton represents the five most usual productions in Brazil: GM Bt, BM
RR, white conventional, organic and colored. • Since there were no expressive differences between the sales prices of these seeds
in Brazil, the attribute price expressed the payment of royalties for GM seeds. • Thus, no prices were presented for white conventional, organic and colored
cotton. Based on a range of values practiced in Brazil, three values of royalties were considered: R$ 20 / ha, R$ 55 / ha and R$ 90 / ha. We also considered the option for a pirate GM seed (with no royalty payments).
• Such design would imply a total of 11 possible alternatives (4 prices of GM Bt + 4 prices of GM RR + 1 choice of white conventional + 1 choice of organic + 1 choice of colored cotton), which were randomly distributed in sets of four alternatives for each interviewed.
Since all cotton characteristics are nominal, the CA can be similarly represented by a main-effect ANOVA, where the attributes are the independent variables and a function of the rank order comprise the dependent variable (KUHFELD, 2010):
ijjijy )(
0 j
where
The yij represent the stated preference of farmer i for a cotton production with characteristics j and designates its monotonic transformation. Analyses were done using TRANSREG procedure of SAS System (SAS, 2012) and non-metric conjoint analysis models were fit using an alternating least squares algorithm (Young, 1981; Gifi, 1990).
Cotton characteristics of valuation in the contingent ranking
Type Price
Bt Pirate (zero)R$ 20 / haR$ 55 / haR$ 90 / ha
RR Pirate (zero)R$ 20 / haR$ 55 / haR$ 90 / ha
Organic -White Conventional -Naturally Colored -
27
Conjoint Analysis
• Organic cotton has the higher utility;
• Bt cotton in second place, with low royalties (R$ 20/ha);
• High rates of royalties imply negative utility in comparison with other choices;
Agroecological networkThe existence of a complex network established in order to support market access for cotton growers in the semi-arid areas, where most actors do not obtain direct financial returns and many do not aim to.;
This goal has been achieved thanks to collective work, which entails organizational, technical and relational issues. It can therefore be concluded that the collective actions performed by the Network are driven by various motivations in addition to economic goals.;
These motivations are not easy to identify or measure. However, where actions geared to equality, justice and solidarity are concerned, subjective values are necessarily present.
29
Next Steps• Multinomial logit model (MLM) in order to estimate the
main determinantes of the producer’s revealed choices;• Based on the MLM results, to apply propensity score (PS)
approach in order to indentify groups of producers with similar socioeonomic characteristics and different revealed choices for cotton seeds (Bt, RR, white conventional, colored and organic);
• Based on the PS results, to apply a new conjoint anlysis for selected farmers (relatively similars) in order to estimate stated preferences for cotton seeds with no selection bias.
Final Remarks• Diffusion process of GM cotton crops is still under way: is
depends crucially of the quality of the varieties;• Impacts in Brazil of Bt cotton and even stacked varieties is
not as high as in other countries, like China and India;• There is limit to accept the payment of royalties in cotton
by small grower. However, the case of Catuti shows that Bt cotton could be important in certain situations;
• Agro-ecological networks are complex, demanding a huge effort by different types of stakeholders, part of them not directly involved in profit seeking activities.