Ravello, June 17 th 2015 Jaim J. da Silva Jr. GM COTTON SEEDS DIFFUSION: IMPACT ASSESSMENT ATBRAZILLIAN AGRIBUSINESS 1
Dec 23, 2015
1
Ravello, June 17th 2015Jaim J. da Silva Jr.
GM COTTON SEEDS DIFFUSION: IMPACT
ASSESSMENT ATBRAZILLIAN AGRIBUSINESS
Cotton Production in Brazil
• Total Cotton area 2013/2014- 1.1 million hectares
• Concetrated in 2 regions: Mato Grosso (57%) and Matopiba (32%)
• Second season for soybean
Yields (ton/ha)
Quantity (ton)
lack of information
1976/77
1978/79
1980/81
1982/83
1984/85
1986/87
1988/89
1990/91
1992/93
1994/95
1996/97
1998/99
2000/01
2002/03
2004/05
2006/07
2008/09
2010/11
2012/13
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
0
500
1,000
1,500
2,000
2,500
3,000
3,500
4,000
4,500
MATOPIBA NORTHEAST MATO GROSSO REST OF THE COUNTRY SOUTH BRASIL (Right Axis)
1,000 Hectares
Northeast and South decline
Mato Grosso and Matopiba growth
• Average area of the plots in the sample – 1,296 Hectares
• Data on cotton production among 2009-2013 seasons at the Mato Grosso State.
• The data comprises values of 157 farms, separated by plots at farm level.
5
Seed Market Structure and Sample data
2009 2010 2011 2012 20130
1
2
3
4
5
6
Average age of Cotton cul-tivars
Where: -WA is the weighted average age of varieties of the sample,- p is the proportion of the area sown to variety i in year t and,- R is the number of years (time t) since the registration (RNC) of variety i.
6
Seed and Trait Market structure
National Market Registered Firms Sample Registered FirmsTotal Cultivars 181 14 Total Cultivars 60 10GM Cultivars 47 6 GM Cultivars 25 4Traits 6 3 Traits 4 3
Firm Market Share (%) 2010/2011Bayer 50%FMT 23%MDM 14%IMAMT 8%EMBRAPA 1%Others 4%
Source: Céleres Consulting
7
Introduction and Outline
• Demand for agricultural biotechnologies in developing countries are growing fast.
• Commercial approval of GM cotton seeds in 2003. Nowadays it is widespread in about 50% of the cotton area.
• Although Brazil holds the 2nd position in the global ranking of GM seeds, research on socioeconomic and environmental impacts assessment at farm level still incipient; Céleres (2014), Alves et al. (2012), Seixas e Silveira (2014).
•General objective - evaluate profit and yield impacts from using GM cotton seeds in Brazillian agribusiness.
8
Introduction and Outline
• Research looks the diffusion of GM cotton seeds at the Mato Grosso State.
• Econometric models based on farm level dataset to assess net profit and yield changes per hectare.
• Identification strategy: interfarm variation in gross operational margin and productivity per hectare for different seed traits (Conventional, Bt, HT) along of the 2009 - 2013 seasons.
• Main results:• Possible to identify a reduction in the number of insecticides application and insecticide spending per hectare when comparing Bt and conventional seeds
• Models does not confirm that there are differences in profit and yield between conventional and transgenic cotton seeds (Bt and HT)
9
Dataset and Descriptive Statistics
Plot Level Information 2009-2013 Conventional Ht Bt
Seed Cost (R$/Hectare) 149.26 260.55 368.83 (71.91) (106.11) (79.84)
Insectiticide applications (number/hectare) 13.10 12.90 10.07
(3.63) (3.50) (2.10)
Herbicide applications (number/hectare) 4.13 4.64 3.82
(1.53) (1.60) (1.51) Insecticide Spending(R$/Hectare) 641.51 606.29 458.57
(223.04) (260.57) (203.76)
Herbicide Spending (R$/Hectare) 334.62 286.06 341.08
(149.47) (121.17) (120.80)
Yield (Arroba/Hectare) 272.45 263.09 267.05 (31.33) (28.00) (30.25) Direct Costs (R$/Hectare) 4,081 3,721 3,623
(588.74) (498.43) (524.10)
Gross Margin (R$/Hectare) 3,047 3,339 3,387
(1,284) (938) (1,261)
No.of plots 164 83 56 Source: Research data
10
Dataset and Empirical Strategy
• Survey on costs, revenues, production and biotech adoption that has been conducted by a private consulting firm during 2009-2013 at the State of Mato Grosso.
• The sample used in the estimations comprises 303 observations, separated according to the plots on the farm. Thus, each observation corresponds to plot i, in the year t, producing cotton using a determined type of seed (Conventional, Bt, HT).
11
The Model
• Net impact of Seed type on cotton yield and profit per hectare (Kathage and Qaim, 2012).
• Pooled data and panel random effects specifications of a cotton yield function and a cotton profit function
• Log log models with dummies for controling time, region and seed type.
•Model 1 - Pooled data - Estimated equation:
Y = α + βX + δt + е
•Model 2- Panel data (random effects) – Estimated equation:
Yit = β0 + β1Xit + wit where wit = ci + eit
12
The Model
• HT seed: reduced loss as side effect of herbicide increases the value of marginal product of herbicide
VMP1
VMP0
𝑥1∗𝑥0
∗ x
$
w
• Prediction: HT seed increases the amount of pr0duction do to better control of weed.
• Net environmental effect is ambiguous• Scale?• Substitution?
13
The Model
VMP0
VMP1
𝑥0∗𝑥1
∗ x
$
w
• Prediction: Bt seed decreases the
amount of insecticides and
increases the profitability
14
Cotton Yield Model – Dependent Variable: Productivity (Kg of fiber and feather per hectare)
*,**,*** indicates that the estimated coefficient is statistically significant to 10%, 5% and 1%, respectively. In both models, the dependant variable is the gross operating margin, measured in R$ per hectare. The values in parenthesis correspond to the standard deviation.
Independent Variable Model 1 Model 2 Constant 4.750*** 4.755*** (0.19) (0.19) Nitrogen 0.083*** 0.082*** (0.020) (0.021) Potassium 0.0712*** 0.0717*** (0.020) (0.021) Year 2011_Dummy -0.077*** -0.075*** (0.019) (0.019) Year 2010_Dummy 0.0508*** 0.0505*** (0.017) (0.0175) Area 0.0146*** 0.0145*** (0.005) (0.005) Region 2_Dummy -0.0827*** -0.0825*** (0.014) (0.014) Ht_Dummy -0.027 -0.022 (0.018) (0.018) Bt_Dummy 0.002 0.008 (0.021) (0.0212) VIF 1.72 - R2 0.3947 - Hausmann - 0.2948
15
Cotton Profit Model – Dependent Variable: Gross margin (R$/Ha)
*,**,*** indicates that the estimated coefficient is statistically significant to 10%, 5% and 1%, respectively. In both models, the dependant variable is the gross operating margin, measured in R$ per hectare. The values in parenthesis correspond to the standard deviation.
.
Independent Variable Model 1 Model 2 Productivity 2.865*** 2.736*** (0.309) (0.225) Cotton Fiber Price 2.523*** 2.21*** (0.388) (0.228) Inseticide Cost -0.307*** -0.253*** (0.061) (0.058) Year 2012_Dummy 0.276*** 0.240*** (0.051) (0.060) Year 2009_Dummy -0.308* -0.230** (0.173) (0.116) Ht_Dummy 0.066 0.060 (0.059) (0.070) Bt_Dummy -0.009 -0.064 (0.079) (0.081) VIF 2.38 - R2 0.5665 - Hausmann - 0.4710
16
Wrapping Up
• Econometric models showed non evidence that GM cotton seeds (IR or Bt trait) increases profit or yield.
• It is possible for cotton farms in Brazil to attain the same level of productivity using either the insect-resistant seeds or using insecticides to manage insects;
• New and old pleagues not controlled by the insect resistance trait. Grass resistance on herbicide tolerant.
• Seed price increases and weak substition of pesticides captures most of the positive impacts of GM seed adoption.
17
Wrapping Up
•In the particular case of Brazil, where farmers have a well-defined cotton production function, the results are in keeping with the reality observed. This is due to the fact that in the growing environment there are different pests against which the insect-resistant trait is not effective in its control;
•It can therefore be deduced that the delay in diffusion process of transgenic cotton seeds is associated not to a lack of information on the part of the farmers, but to their deep knowledge on pest and weed control methods, therefore the focus on productivity and profit gains changes when they decide to adopt new technologies.
18
Control Moderate control No control
Alabama argillacea Helicoverpa zea Anthonomus grandis SP
Chrysodeixis includens Helicoverpa armigera Aphis gossypii
Heliothis virescens Spodoptera frugiperda Nezara viridula
Pectinophora gossypiella Euschistos heros
Spodoptera eridania
Spodoptera cosmioides
New Plagues
19
THANK YOU! [email protected]