early season crop forecasting with fasal ......•boro – rice sown in winter and harvested in...
TRANSCRIPT
Presented at International Seminar on “Approaches and Methodologies for Crop Monitoring and Production Forecasting” Under AMIS Global
inititaive, 25-26 May, 2016, Dhaka Bangladesh
FASAL Institute of Economic Growth (IEG)
University of Delhi Delhi 110007
EARLY SEASON CROP FORECASTING WITH FASAL ECONOMETRIC MODEL:
ITS USEFULNESS BEYOND FORECASTING
Nilabja Ghosh
OBJECTIVE
Consider Econometric modeling as a method to determine OUTLOOK of crop production before any crisis strikes- understand FASAL
Demonstration of estimates and forecasts
Select Cases to demonstrate usefulness: PULSES for price control
Policy options against crop promotion
Acknowledgement:
All Analysts who worked for FASAL at various times and contributed significantly to building up database, refining and applying model and method, setting up Software since 2005.
Contribution to this presentation by M. Rajeshwor, FASAL and Yogesh Bhatt for work on Biofuel acknowledged
ESTIMATION OF CROP OUTPUT IN INDIAN AGRICULTURE
A long time historical practice from colonial time
Associated with land revenue system, Crop cutting experiments
System improved over time- more scientific
Burden on administration
Errors, poor implementation of methods
Delays: Too late to be useful for Policy.
Gining significance for policy making in a world of volatile and integrated global markets, weather failures (climate change?),-
Need to avoid food insecurity, inflation, price crash, potential unrests and suffering Early information to plan stocking and procurement, timely
organization of logistics, credit, making trade strategy and market negotiations
EARLY ESTIMATES
Assessments of production and Acreages of crops early in the season- Hhistorical practice Generally always subjective- eye observation
State government responsibility – local officials in charge –maintained a register Excessive burden as they have multiple functions
Institution available only in some states, others have small sample surveys (EARAS, TRS)- time consuming
Early alerts and warnings useful even if imprecise
Need for rigorous -method based, transparent and data driven scientific forecasts with regular monitoring and updating in tune with changes in production climate and model performance for Timely Policy support
FASAL AN UMBRELLA PROGRAM -
Recommendation (1996-2000) for a ‘comprehensive project’ and ‘strong mechanism’ with ‘latest techniques’ to meet in-season forecast requirements
Led by ISRO-SAC, associated with India’s space program Visualized by SAC (Dr. Parihar): Since 2005 as a regular
Systematic mechanism Multi-discipline and innovative: Generating multiple in-season crop
forecasts at intervals in the year
Partnership of Institutions under coordination of MoA- IEG, SASA, State SAC, IMD ISRO NRSC etc Final most reliable estimate from FASAL -RS
Ministry puts on public domain 1AE-Sep 2AE-Jan 3AE-Apr 4AE-Jul and Final-Jan estimates of area and production of different crops. Inputs from State government, FASAL – validation, support and
comparison Under improvement, evolution
ECONOMETRIC MODEL BASED FORECAST
Earliest among all in-season forecasts (F0 F1) Least informed -Based on reasonable assumptions on driving
variables, known econ. Conditions, Prediction- Normal and alternate weather conditions scenarios. Policy assessment- potential of using estimated marginal
effects and simulations
Two stages: Acreage and Yield- Model estimation ,
Forecasts State level early estimates of area and yield (with ranges) for select crops in
major growing states
Forecast production- F-Area X F-Yield, Range based on SE of Area and yield predictions
Projected for All-India aggregate using state totals and proportion based on recent history.-or alternative methods. • Alternative scenarios of weather
Model in REDUCED FORM and ESTIMATION
Driving variables are predetermined or exogenous
Functional form-Linear allowing interactions and quadratic terms
Specification chosen on the basis of diagnostics- Sign of coefficients, t-stat>1
Robustness across specifications and sample sizes. Rbar Sq, DW, UR of error, AR corrected if indicated
Dynamic Area equation (Nerlovian partial adjustment, price expectations),
Yield allows for time trend, dummy variables for Policy (NFSM, BGREI, Bt. )
Estimation-Seemingly Unrelated Regression Equations (SURE) for competing crops in each state,
• Data: Official sources- MOA, IMD, M-Com&Ind Sample 1985-86 to 2013-14
Regular post sample validation, revision and up-dation of model
7
COVERAGE OF CROPS AND EXPLANATORY VARIABLES
Explaining dependent variable Crop area and Crop yield per hectare Kharif - Rice, Jowar, Bajra, Maize, Cotton, Jute, Groundnut, Soybean,
Sugarcane, Arhar, Moong and Urad. Rabi- Wheat, R&M, Groundnut, Jowar, Maize, Gram, - Major growing states, new states (JH, CHH, UKH, BH, MP, UP) with
limited data. Onion, Potato- experimentally
Explanatory variables • Economic: Expected prices of crops and substitute crops (using state
crop calendar), MSP (rice, wheat in procuring states), fertilizer price (cost)
• Irrigation: Source wise -Total available area as the variable (farmer allocates among crops)
• Rainfall and Temperature (States): monthly averages Sowing and growing seasons identified
• State level crop calendars, IMD monthly data • Rainfall effects: Distribution matters interactions with soil moisture,
reservoir, ground water and adverse effect of excess rainfall
• Temperature specification: Dummy for higher than average by 2%.
RAINFALL EFFECT IN MODEL
Distribution matters: allow pre-season-(also pre-sowing, pre-monsoon, last monsoon) RF for soil moisture effect-monthly data
Monsoon- June-Sept-Ocober,
Timely rainfall or adequate soil moisture can influence crop choice and allocation of non-water input among crops
Quadratic (squared) Rainfall: Excess rainfall (compared to optimum) may harm
Interactions: Interactions of irrigation (source-wise) with rainfall- temporal distribution
Complements: Pre-season (s) and current Rainfall can influence productivity of irrigation from specific sources-(enhancing reservoir level, ground water, tank water, help drainage of rainwater etc.)
Substitute: Current or recent rainfall can influence productivity of irrigation (good rainfall can reduce need for irrigation, create drainage and w-management problem etc.)
AREA EQUATION
Where = expected price (previous harvest month prices (Kharif/Rabi) and MSP Sub = Substitute crop in that season in the Region SRS = Source wise m = sowing/pre-sowing months T = Temperature
= Price of fertilizer =growing months/pre-sowing months Rainfall DumT = Dummy As necessary for Technology programme R F and Temp effect + or - Others as in Area slide
YIELD EQUATION
EXAMPLES
12
SEASONS CROP CALENDAR
West Bengal: Rice Kharif: Aus (Autumn, minor) : Feb-Apr (Sowing) – July-Aug (Harvesting) and
Aman (major): July-Aug (Sowing)- Nov-Dec (Harvesting)
Rabi (Boro): Nov-Dec (Sowing) – Mar – June (Harvesting)
Largest harvest is Aman, occurring in November and December second harvest is Aus, involving traditional strains but more often including high-yielding, dwarf varieties. Rice for the Aus harvest is sown in March or April, benefits from April and May rains, matures during in the summer rain, and is harvested during the summer. Another rice-growing season extending during the dry season from October to March. The production of this Boro rice
Madhya Pradesh: Wheat Rabi: Oct-Nov (Sowing) – Feb-Mar (Harvesting)
Karnataka-: Maize Kharif: May-June (Sowing) – Sep-Oct (Harvesting)
Uttar Pradesh: Potato Rabi: Oct- Nov (Sowing) – Feb-March (Harvesting)
13
RICE SEASONS WEST BENGAL
• Sowing Season: Rice is sown mainly thrice in a year:
• Aman – Rice sown in the rainy season (July-August) and harvested in winter. India produces Aman Rice mainly.
• Aus –Rice sown in summer along with the pre-monsoonal showers and harvested in autumn is called Aus Rice. The quality of this rice is rather rough.
• Boro – Rice sown in winter and harvested in summer is called Boro Rice or spring Rice.
14
AREA EQUATION
15
Back to back droughts in 2014-15 and 2015-16:
Challenges for FASAL
State-wise and Total (Select growing states) Rainfall Departure (%) in 2014
16
J U N - S E P T
O C T – D E C
State-wise and Total (Select growing states) Rainfall Departure (%) in 2015
17
And
hra
Prad
esh
Ass
am
Bih
ar
Guj
arat
Har
yana
Karn
atak
a
Mad
hya
Prad
esh
Mah
aras
htra
Ori
ssa
Punj
ab
Raj
asth
an
Tam
ilnad
u
Utt
ar P
rade
sh
Wes
t B
enga
l
All
Indi
a
-8.1
-4.9
-24.
7
-11.
1
-36.
6
-19.
9
-13.
9
-26.
5 -10.
0
-31.
6
9.7
-9.9
-43.
2
5.4
-17.
6
And
hra
Prad
esh
Ass
am
Bih
ar
Guj
arat
Har
yana
Karn
atak
a
Mad
hya
Prad
esh
Mah
aras
htra
Ori
ssa
Punj
ab
Raj
asth
an
Tam
ilnad
u
Utt
ar P
rade
sh
Wes
t B
enga
l
All
Indi
a
-4.8
-53.
1
-82.
6
-87.
6
-78.
9
-19.
8
-58.
1
-61.
8
-68.
1
-74.
6
-80.
3
51.8
-76.
0
-72.
7
-27.
0
J U N - S E P T
O C T – D E C
Validation of all India Production (recent 2 years)
IEG-2014-15 (MOA)
2014-15
Error
%
IEG-
2015-16
3AE-
2015-16
Error
%
Crops Million Tonnes Million Tonnes
Rice Kharif 84.8 91.4 -7.2 92.6 90.6 2.2
Rice Rabi 12.3 14.1 -12.8 12.3 12.8 -3.7
Rice Total 97.1 105.5 -8.0 104.8 103.4 1.4
Wheat 88.4 86.5 2.2 88.7 94.0 -5.7
Maize Kharif 17.5 17.1 2.3 16.0 15.5 3.2
Arhar 2.9 2.8 3.6 2.7 2.6 3.8
Moong Kharif 0.9 0.9 3.4 1.0 1.0 -2.0
Gram 7.3 7.3 0.2 8.5 7.5 13.6
Potato 44.6 48.0 -7.1 - 48.1 -
Usefulness -forecasting for controlling prices of Pulses
• Pulses - dominant items in Indian diet - sources of nutrition (protein)
• India is deficit in Pulses production
• Promotion by policy – TMOP-1990 , TMO-1980, ISOPOM-1995 etc.
• Need for timely imports • Pulses: 4.6 Ml. T in 2014-15 and 2.24 Ml. T imported in 2015 (April-Sept)
• Erratic Imports- options limited
• Managing scarcity of supply and price rise of Pulses a Challenge for the economy
• FASAL provided outlook in 2nd consecutive drought year 2015-16
• Presented at Krishi Bhavan, Ministry of Agriculture, New Delhi on 19th August, 2015, Forecast of Pulses (also Oilseeds) using actual rainfall data up to 18th August, 2015
8000
9000
10000
11000
12000
13000
14000
15000
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
Total Pulses Area
Actual Fitted
5000
6000
7000
8000
9000
10000
11000
12000
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
Total Pulses Production
Actual Fitted20
Policy of promoting Biofuel for Energy security and GHG emission control (taken from Yogesh Bhatt)
Agro-based Biofuels – considered possible solution to the depleting sources of fossil fuels and GHG emissions Limited land resources, Diversion of crop land to biofuel can
compromise food production, Yield improvement may offset decrease in food crop acreage
Possible feedstock in India- Maize, Sugarcane, Soybean
The National Biofuel Policy by the Ministry of New and Renewable Energy (MNRE) released in December, 2009 To accelerate promotion of use of biofuels to increasingly
substitute petrol and diesel for transport
India has a target of 5 % blending by 2012, 10 % by 2017 and 20 % after 2017 as per policy 2009.
FASAL Model- useful to identify crops whose acreage likely to be hit in different states, to Simulate yield improvement necessary to offset production loss due to land diversion
Yield improvement required at aggregate level (%): Simulation 2013-14
Crops Maize Sugarcane Soyabean
Fo
od
gra
ins
&
cere
als
Rice 0.7 2.1 0.03
wheat 0.5 0.8 -
Bajra 3.2 6.6 0.3
Jowar 5.0 0.7 0.8
Ragi 2.4 17.7 0.5
Pu
lses
Arhar 1.9 5.8 0.3
gram 1.0 1.7 -
Moong 2.2 4.8 0.7
Urad 1.8 4.7 0.5
Oil
seed
s Groundnut 1.1 7.7 0.4
rpmst 1.3 2.9 -
Fib
re
Cotton 1.1 2.9 0.1
Jute 3.9 8.2 -
Yield improvement (%) needed in Indian states for promoting biofuels (Rs. 1000/tones)
Food crop: Jowar, Substitute: Maize Food crop: Arhar, Substitute: Sugarcane
(7.75,8.96](4.4,7.75][2.85,4.4]No data
(3.88,19.73](1.06,3.88](.6,1.06][.2,.6]No data
TOWARDS A REFORMED POLICY PARADIGM
• Need for coordinated, well deliberated policy making on production, stocks, import, export, distribution, credit etc.
• Inter-Ministerial consultation required on outlook formed by multiple alternate agencies and rational methodologies
• Overcome errors and delays of forecasts/estimates by validation of field observation with FASAL
• Satellite RS information becoming more important • Early stages RS has limitations,
• Econometric model can generate method based prediction for early planning- also macro-planning by aiding GDP estimation-
• Can prevent serious hardship, crisis and policy emergency
• In the long run- can be extended over multiple-countries, linked by geography, trade and information flows for integrated good results of prevention of hunger and disaster
AREA EQUATION
Rice (K) West Bengal
Maize (K) Karnataka
Rice (R) West Bengal
Wheat (R) Madhya Pradesh
Potato (R) Uttar
Pradesh
Constant 6.2 0.03 -2.0 -0.2 -2.5
Price 3.1*** (Sub- Urad)
2.4** (Sub-Urad)
4.8*** (Sub- Urad)
3.8*** (Sub- Gram,
Moong)
1.5 (Sub-Wheat,
Gram, Moong)
Rainfall 2.72** (Pre-Monsoon+
Monsoon)
-2.3** (Monsoon)
2.6** (Aug) 5.1*** (Nov)
3.4*** (Monsoon) 2.7** (October)
Rainfall2 -2.8*** (Aug)
Irrigation -3.04*** (Well)
1.7 (Total)
3.2*** (Canal+Well)
-2.7** (Well)
Interaction 3.1*** Canal*Monsoon(-1)
1.9* All*Monsoon
-2.9*** Canal*Monsoon(-1)
4.2*** (Well*Monsoon)
2.3** (Tank*Aug)
3.2*** (Well*Feb)
Temperature -2.2** (April_Min)
1.9* (Sep_Max)
Area (-1) -1.6 1.8 11.4*** 4.4*** 7.3***
Adjusted R-squared
0.80 0.97 0.96 0.92 0.88
26
YIELD EQUATION Rice (K)
West Bengal Maize (K) Karnataka
Rice (R) West Bengal
Wheat (R) Madhya Pradesh
Potato (R) Uttar Pradesh
Constant 13.0 0.86 17.6 -0.10 9.5
Price 6.7*** (defl-Fert)
2.5** (defl-Fert)
2.0* (defl-Fert)
2.7** (defl-Fert)
3.9*** (defl -Wheat,
Moong)
Substitute Price
-3.0*** (Urad/
defl-Fert)
- - - -
Rainfall 2.0* (Jan) -2.5** (Feb) -4.1*** (Sep)
2.6** (Monsoon)
4.1*** Monsoon(-1)
2.4** (Apr) -3.3*** (June)
-1.6 (Oct) 8.1*** (Monsoon)
2.6** Jan(+1) -5.8*** Mar(+1)
-2.6** (Dec) 2.6**
Jan(+1)
Irrigation 2.7** (Canal+Well+Oth)
-1.1 (Well)
2.9*** (Well+Canal)
Interaction 1.9* (Well*Monsoo
n) -3.1***
Total*Dec(-1)
2.7** (Canal*Sep)
2.3** (Well*Dec(-1))
5.8*** (Canal+Well)*Jan
2.2** (Well*Oct)
Temperature Aug_Min Oct_Min
Adjusted R-squared
0.95 0.72 0.50 0.93 0.69
27