International Food Policy Research Institute (IFPRI) and Ethiopian Development Research Institute (EDRI) in collaboration with Ethiopian Economics Association (EEA). Eleventh International Conference on Ethiopian Economy. July 18-20, 2013
"Subclassing and Composition – A Pythonic Tour of Trade-Offs", Hynek Schlawack
Dynamic Acreage Demand and Supply Responses in Ethiopia
1. Fantu Nisrane Bachewe and Alemayehu Seyoum Taffesse
International Food Policy Research Institute (IFPRI)
(Ethiopia Strategy Support Program, ESSP-II)
The 11th International Conference on Ethiopian Economy
20 July 2013
Addis Ababa, Ethiopia
Dynamic Acreage Demand and Supply Response
of Farm Households in Ethiopia
2. Presentation Outline
1. Background of study
2. Linear rational expectations model
3. Elasticities
4. Data description
5. Results
6. Summary of findings.
3. Background
• Broadly agreed that farmers in developing countries respond
little for market incentives, such as prices
• Most argue production is for subsistence,
• In Ethiopia, crop production is dominated by subsistence farm
households with little marketed surplus.
• Small holders accounted for about 97.5% of both area and output of grains
and all temporary crops in 2008/9 (2001 EC)
• Shares declined in grains to 96.6% of area and 96% of output in 2010/11
• During 2008/9-10/11 grains consumption and sales were 62 & 21% output
• Previous works on farmers’ responses to prices covering the pre-
1990 Ethiopia include Seyoum Taffesse (2003)
4. Background…contd.
• Major changes that affected crop production since mid-1990s
• Liberalization of inputs and outputs market
• Emphasis on agriculture and agriculture led development
• Recent fast agricultural and economic growth and high food prices
• Farmers’ responses to recent changes studied little
• Contributing to knowledge in area among objectives of this work
• Objectives: study acreage demand responses for changes in crop
prices and costs of production
• Dynamic rational expectations (RE) farm household model used
• Sargent (1987) and Hansen and Sargent (1980) – generic firm
• Applications in agriculture: Eckstein (1984, 1985), Tegene, Huffman,
Miranowski (1988), and Seyoum Taffesse (2003)
5. Model
• Crop production is linear in acreage allocated for 2 types of crops
• Non-land costs linear in acreage and incurred in periods t-1 and t
• With rational expectation assumption the Euler’s equation of the
optimization problem is:
2
, 1 , , 1 , 1 , 1 , 2, ( )
2
i
i t i t i t i t i t i i t i t
b
C v f A A d A A
, 1 ,, ; 0, 0,1,..., and =1,2i t i i t i t iQ y A y t i
1
1, 1 1, 1 1 1, 1 1 1, 1 1, 1
0
( ) [ ]i
t t t t i t i t i
i
A A E P y R V k
d
6. Model.
• In the empirical analyses teff takes the place of crop 1
• Selected due to high acreage share and proportion marketed
• Empirical teff acreage demand and price equations are:
1, 0 1 1, 1 2 1, 2 3 1, 4 1, 1 5 1, 6 1, 1
1, 1, 1 ,
3
| | 1, 2
t t t t t t t t
P P
t t i i t
A A A P P R R t
P P t
7. RE: Elasticities
• The long-run acreage elasticity is positive with respect to (WRT)
price and negative WRT opportunity costs of teff acreage
• Sign of SR elasticities depend on the sign of d and the period i in
which prices and costs are expected to change.
• Particularly
and if d<0 or d>0 and i is odd.
1 1
1 1 1
,
1 1 1
0
(1 )(1 )
L
A P
y EP
d EA
1 1
1 1
,
1 1 1
0
(1 )(1 )
L
A R
ER
d EA
1 1
1
, 0i
A P 1 1
1
, 0i
A R
8. Data
• Zonally aggregated data from FDRE Central Statistical Agency (CSA)
annual AgSS household used. The data
• Cover the 2003/4-2010/11 period
• Include zones in Tigray (5), Amhara (10), Oromiya (17), and SNNP (21)
• Grains (cereals, pulses, and oilseeds)
• Zonal aggregation needed because:
• HHS resampled each year and zones are representative of regions
• Price taking assumption made for HHs may not apply at zone level
• Tests indicate no causality in both directions (price and acreage)
• Average monthly producers’ prices from CSA are also used.
9. Data…contd.
Mean regional acreage & output shares and prices 2003/4-2010/11.
Variable National Tigray Amhara Oromiya SNNN
Grains out of total agricultural output (%) 86 91 92 81 45
Growth rate in grains output (%) 12 14 11 13 12
Teff output out of grains (%) 18 17 22 16 16
Growth rate in teff output (%) 12 12 6 9 16
Grains out of total agricultural area (%) 92 91 92 81 45
Growth rate in grains area (%) 4.6 5.9 4.5 6.2 4.7
Teff area out of grains (%) 23.2 18.0 25.2 19.4 22.7
Growth rate in teff area (%) 4.7 3.1 3.1 6.8 6.7
Nominal teff price (birr/KG) 4.4 4.8 4.6 4.2 4.2
Growth rate in nominal teff price (%) 20 20.7 19.4 20.5 20.1
Real teff price (December 2006 birr/KG) 3.6 3.8 3.8 3.5 3.5
Growth rate in real teff price (%) 3.5 5.3 3.0 4.2 3.0
10. Results: Acreage equation
• Teff acreage and price equations estimated using Arellano-Bond and
AB-BB systems GMM
• Sign of coefficients as expected and significant except teff price in AB
• Acreage a positive function of both of its lags with declining effect
• Lagged price has slightly stronger effect
• Effect of current opportunity cost negative on teff acreage
11. Variable Systems (AB-BB) Arellano-Bond
Constant 0.022 -0.039***
Acreage (lagged) 0.419*** 0.573***
Acreage (twice lagged) 0.074** 0.283***
Teff price 0.019*** 0.001
Teff price (lagged) 0.021*** 0.022***
Rat -0.00016*** -0.00005***
Rat (lagged) 0.00005** 0.00003***
χ2/F statistics 3943 *** 1489 ***
N 318 318
Results…contd.
Dynamic panel data estimates of acreage demand function.
12. Results: Elasticities
• Elasticities computed from coefficient estimates have expected signs;
• Acreage demand elasticity WRT own price
• Long-run: 1.4 in systems and 2.7 in AB, and
• Short-run: 0.34 in systems and 0.01 in AB.
• Acreage demand elasticity WRT opportunity cost of teff acreage
• Long run: -0.36 in AB and -0.54 in systems, and
• Short-run: -0.13 in AB and -0.38 in systems.
13. Results: Elasticities.
Dynamic panel
data model
Long-run acreage
elasticity WRT
Short-run acreage
elasticity WRT
Teff
price
Opp. Cost
of teff
acreage
Teff
price
Opp. Cost
of teff
acreage
System (AB-BB) 1.38 -0.54 0.34 -0.38
Arellano-Bond 2.71 -0.36 0.01 -0.13
Calculated elasticities of acreage demand .
14. Further Results
• We estimated unrestricted reduced form equations
• Not possible to recover all structural parameters and test their
implications,
• However, was possible to uniquely identify some structural
parameters
• The coefficients of lags of acreage ensure the stationarity conditions
• They also imply , which with calculated elasticities imply d<0
• The latter satisfies the requirement that d and have opposite signs,
• That d is negative, in turn, implies decline in land preparation costs
outweigh the decline in marginal productivity
1 0.552
1
15. Results of previous analyses using rational expectations.
Summary and remarks.
• LR price elasticity higher and SR comparable with those in relevant work
• LR elasticity of opportunity cost of acreage lower and SR higher
• Own price elasticities range from 0.2-2 (LR) & 0.01-1.2 (SR) in other works
Study Crop/ acreage
Price (Opp.
Cost)
Elasticity
Long run Short run
Seyoum Taffesse (2003) Teff
Own price 0.48 0.31
Rat -0.93 -0.13
Mythili (2008)
Wheat own
price/AVC
0.24 0.07
Sugarcane 1.51 0.27
Ahouissoussi et al 1995
Soybean Own price 1.98 1.21
Soybean Wheat price -1.57 -2.11
Eckstein (1994) Cotton Rt -0.13 -0.11
Tegene et al. (1988)
Corn Own price 0.2 0.011
Development economists broadly agree that subsistence producers respond little for market incentives, such as pricesAmong factors, production is mostly subsistence with little marketed surplus,In Ethiopia, crop production is dominated by subsistence households/farmers with little marketed surplus,In 2008/9 (2001 EC) out of the total area under temporary (perennial) crops the share cultivated by small private holders was 97.5% while out of the total area under grains they accounted for 97.3%. Similarly, small holders accounted for 97.4 and 97.2% of all temporary and grains output. In 2010/11 small holders’ share of the area under grains, which grew annually at 2.3%, declined to 96.6 percent while commercial farms’ share grew annually at 16%. Similarly, small holders’ share of total grains output, which grew annually at 9.4%, declined to about 96 percent while commercial farms’ share grew annually at 39%. Although growth in area and output among commercial farms was 4 and 7 times faster than growth in small holders its contribution to totalgrowth was small accounting for only 11 and 16 percent of the growth in grains output and area in the country,During the 3 years of 2008/9-2010/11 household consumption accounted for 62 percent of grains production, which increased marginally during the period, while sales accounted for 21 percent, which declined at 1 %.Farmers’ responses to recent changes in market incentives studied little
Objectives: study acreage demand responses for changes in crop prices and costs of production Period: 2003/4-2010/11Geographic coverage: zones in Tigray, Amhara, Oromiya, and SNNPCrops included: grains (21 types of crops)Model: dynamic rational expectations farm household modelSargent (1987) and Hansen and Sargent (1980)Applications in agriculture: Eckstein (1984, 1985), Tegene, Huffman, Miranowski (1988), and SeyoumTaffesse (2003)
We consider a representative infinitely-lived dynastic farm household which maximizes its discounted expected intertemporal utility via its consumption, production, and saving choices. The household is assumed to have a linear one-period utility function,The impact of the randomness of some variables on the choices farmers make considered without modeling their behavior towards risk. In each period yield and price risks are realized before consumption decisions are taken. Under these circumstances, the farm household’s production and consumption decisions are separable. The farm household maximizes its discounted expected utility by first maximizing its discounted expected profits, and subsequently choosing the level of consumption and/or savings subject to the corresponding budget constraint. Exogenously given land allocated for the production of 2 cropsCrop production islinear in crop acreage, stochastic, and involves a one-period lag between cultivation and harvest (harvest at t+j is a function of acreage at t+j−1).
With the assumption of RE the EE of the optimization problem is:Households’ current acreage demand for Crop 1 is a function of past allocations of acreage to Crop 1, expected output prices, and realized and expected actual and opportunity costs. Equation (2.9) does not constitute a decision rule as it includes the processes
We use zonally aggregated CSA annual AgSS household dataPeriod covering the 2003/4-2010/11Geographic coverage: zones in Tigray, Amhara, Oromiya, and SNNP – The 4 regions on average accounted for over 90 percent of agricultural output and cultivated area during the 2003/4-2010/11 period.Moreover, the remaining regions miss most of the data essential for the analysis.Crops included: grains (21 types of crops)-Grown widely in all 4 regionsGrains accounted for 86% total output and over 90% area (and the 5 important cereals accounted for 69% and 71%)Data on other, mostly permanent, crops intermittent – and notorious for their measurement errorZonal aggregation needed: Analyses requires the use of panel dataHHS resampled each year and zones are representative of regionsDifferent number of woredas surveyed across the years
During the 2003/4-2010/11 period grains/5 cereals accounted for an average of 86/69 percent of the total agricultural output and over 90/71 percent the area.
Systems (Arellano and Bover (1995) and Blundell and Bond (1998)) and
Teff acreage demand LR elasticity WRT price range 1.4-2.7 and SR elasticity range 0.01-0.34LR elasticities WRT opportunity cost of teff acreage range -0.36- -0.54 and SR elasticity range -0.13-0.38
Teff acreage demand LR elasticity WRT price range 1.4-2.7 and SR elasticity range 0.01-0.34LR elasticities WRT opportunity cost of teff acreage range -0.36- -0.54 and SR elasticity range -0.13-0.38
However, that implies y1=338 and The estimates of and ensure the stationarity conditions and and provide and as Moreover, the last two results together with the short run elasticity imply that d is negative, which verifies using the SR elasticity calculated from estimated coefficientsThe latter satisfies the requirement that d and have opposite signs and implies decline in land preparation costs outweigh the decline in marginal productivity resulting from successive cultivation of the same plot That d is negative, in turn, implies decline in land preparation costs outweigh the decline in marginal productivity
Among works that studied acreage demand we put results of thosusing similar methodology and/or variables in this table
We use zonally aggregated CSA annual AgSS household dataPeriod covering the 2003/4-2010/11Geographic coverage: zones in Tigray, Amhara, Oromiya, and SNNP – The 4 regions on average accounted for over 90 percent of agricultural output and cultivated area during the 2003/4-2010/11 period. Moreover, the remaining regions miss most of the data essential for the analysis.Crops included: grains (21 types of crops)- During the 2003/4-2010/11 period grains/5 cereals accounted for an average of 86/69 percent of the total agricultural output and 96/71 percent the area. Grown widely in all 4 regionsData on other, mostly permanent, crops intermittent – and notorious for their measurement error