changes in the agro-climate effects on cereal crop yields: panel evidence from india (1972-2002)...

7/31/2019 Changes in the Agro-Climate Effects on Cereal Crop Yields: Panel Evidence from India (1972-2002) with Implications for Sub-Saharan Africa

1/53

Changes in the Agro-Climate Effects on Cereal Crop Yields:Panel Evidence from India (1972-2002)

with Implications for Sub-Saharan Africa

SSD Seminar

e . ,

Takuji W. Tsusaka

Keijiro Otsuka

Copyright 2012 by Takuji W. Tsusaka and Keitjiro Otsuka. Readers may make verbatim copies of this documentfor non-commercial purposes by any means, provided this copyright notice appears on all such copies.


2/53

Introduction

Modern Varieties (MVs): High-yielding crop varieties suitable for the

regional agro-climate; Especially wheat

Index: 1961=100

The Green Revolution (GR) in Asia

Key Factors for the GR

Growth in agricultural production consistently outpaced population growth, owing to the

Green Revolution (e.g., Otsuka and Kalirajan, 2006).

Changes in Per-cap* Cereal Crop Production (Value-added)

World

Sub Saharan Africa

1

and rice varieties.

Irrigation: Stable and sufficientsupply of water.

Fertilizer: Intensive use of chemicalfertilizer.

Other: Markets (inputs/outputs),infrastructure (e.g. road), credit,

education.

The technological innovation and other complementary factors spurred the agricultural

productivity in Asia, which led to rural poverty reduction (as well as non-farm sector growth).(e.g., Otsuka et al., 2009; Lipton, 2007; Otsuka and Yamano, 2005; Fan et al., 2000)

Source: FAOSTAT* National

South Asia

Southeast Asia


3/53

5.0

6.0

Introduction

(Index: 1961=100)

Agricultural Stagnation in Sub-Saharan Africa (SSA)

Staple food production has been increasing in SSA, but the rate of increase is not high enough

and has been exceeded by its population growth.

(tons/ha)

Average Cereal Yields, 3-Year Moving Averages

World

SSA

Changes in Per-cap* Cereal Crop Production (Value-added)

0.0

1.0

2.0

3.0

4.0

1963

1966

1969

1972

1975

1978

1981

1984

1987

1990

1993

1996

1999

2002

2005

2008

2

SSA sees a decline in per-capita agricultural production.

Source: Calculation with FAOSTAT Data

* National

or mer ca

Asia

South Asia

North Africa

SSA

100

South

Asia

SoutheastAsia


4/53

Why has SSA missed the GR? (1)

Policies

(&Governance)

AgriculturalProductivity

Irrigation (Water

Management)

Markets/Credit/

Infrastructure/Education

High-YieldingVarietiesR&D

Climate

Endowments

Fertilizer

Introduction

3

Critics have long argued that there is limited potential to attain a GR in SSAdue to its adverse climate endowments.

Dry Climate: A number of studies show significant effects of climate on crop yields,

particularly positive effects of rainfall (Seo and Mendelsohn, 2007; Auffhammer et al., 2006; Olesen

and Bindi, 2002; Sanghi et al., 1998; Bruce et al., 1996; Reilly et al., 1996; Adams et al., 1995). Diverse Climate: It results in producing a broad range of staple crops, leading to

limited scale benefit of investing in standard technical packages as in the case of Asia

(Omano, 2003; Mwabu and Thorbecke, 2004).

One of the major constraints in SSA is its unfavorable (i.e.: dry) and diverse climate,

since climate is a direct input for agricultural production (Omano, 2003; Mwabu and Thorbecke,

2004).


5/53

Policies

(&Governance)


Irrigation (Water

Management)

Markets/Credit/



Climate

Endowments

Fertilizer

IntroductionWhy has SSA missed the GR? (2)

4

Under-developed Irrigation: irrigation and other water management systems have not

been widely introduced in SSA (e.g. Hayami and Godo, 2005; Spencer, 1994).

Insufficient Fertilizer Use:partly a consequence of high fertilizer prices due to poor

infrastructure (e.g., road), and lack of credit and education.

Other Constraints

The Asian GR technology has been

recognized as dependent on intensive and

controlled supply of water and fertilizers.

The adoption of improved

technologies in SSA has

been confined to limited

regions under favorableconditions.


6/53

IntroductionAny potential for African agriculture?

Country-specific case studies on African agriculture point out that the rice yields willsignificantly increase once the constraints are properly addressed along with the

adoption of modern technologies (Kajisa and Payongyon, 2008; Sakurai, 2006; Kijima et al., 2006;Diagne, 2006; Goufo, 2008).

The Asian GR has been technology-led, and thus investments in agricultural research

Some recent studies show that there is some potential for new technology adoption and

crop yield improvement in SSA, which has just yet to be effectively exploited.

5

Since all these studies are based on descriptive statistical analysis, a more formal

econometric testing on the subject would confirm this argument.

and extension would lead to growth in African agriculture(Otsuka and Kijima, 2010)

. In India, MVs of cereal crops were introduced in favorable areas at the initial stage of

the GR. But, the MV adoption rate in unfavorable areas started to pick up at the later

stage as technology continued to advance. (Byerlee, 1996; Fan and Hazell, 1999; Janiah et al.,

2005; Gollin, 2004).

Furthermore, In SSA, only


7/53

Objective

Conventionally, the Asian GR technology has been recognized as a resource-

demanding technology which relies on the intensive use of water as well as

fertilizers.

The GR technology generally results in aggravating the adverse effect of harsh

agro-climate on crop yields.

-

6

dependence of crop yields, that would make a positive case for the possibility of anAfrican GR.

Itis interesting and important to empirically explore whether and to what extent theinfluence of agro-climatic conditions on crop productivity has augmented or

mitigated by the GR in Asia.


8/53

Contribution of The Study

The over-time changes in the impacts of agro-climate on crop yields, which have yet to be

unveiled, are examined.

Few studies on the subject have ever employed a panel data set that covers sufficient

observations along cross-sectional, temporal, and crop-wise axes, due to the data constraints.This study uses a crop-by-crop district-level panel over a long period, which has at least three

advantages over the existing studies:

The use of fixed (or random) effect can

7

- control for the unobservable time-invariant district-specific effects, which can mitigate omittedvariable problems

- alleviate estimation biases which may arise, for example, from endogeneity of explanatory variables

and sample selection.

The long-term data set enables the assessment of the over-time changes in the impacts of climatic

conditions.

Yield functions are estimated for each individual crop and are compared with each other, which leadsto finding the comparative advantage of one crop over others in certain production environments.


9/53

42

8

18

23

12

1

24

9

1

India vs. SSAProportion of Area Harvested to Cereal Crops (%)

21

47

2~3

SSA India Other Asia

2003-2007 Avg.

Sorghum

Millet

70-73 Avg.

India

03-07 Avg.

35

19

8

61

6

3

27

32

10

44

34

Indias diverse cropping patterns reflect its diverse agro-climate.

The agricultural production environments (in some parts of India, if not all)

are similar to those in SSA, which implies a technology transferability.


* Cassava, Teff, Potatos, Ragi, Oats, Barley and other

Other*

Maize

Wheat

Rice

Source: Calculation with CMIE Data


10/53

Bajra (Pearl Millet) Field in India

9


11/53

1.5

2.0

2.5

3.0

3.5

4.0

India

(tons/ha)

Southeast Asia

Combined Cereal Yields (3-year MA): India vs. SSA

India vs. SSA

10

0.0

0.5

1.0

1963

1965

1967

1969

1971

1973

1975

1977

1979

1981

1983

1985

1987

1989

1991

1993

1995

1997

1999

2001

2003

2005

2007


Sub-Saharan Africa

Despite the less favorable production environments, cereal crop yield in SSA was not significantly

inferior to that in India until the early 80s.

Today there is a gap of two-fold in cereal yield.


12/53

2.5

3.0

3.5

2.5

3.0

3.5

Wheat

Rice

Maize

Yield (tons/ha) Yield (tons/ha)

Wheat

India Sub-Saharan Africa

Cereal Yields (3-year MA) by Crop: India vs. SSA

India vs. SSA

11

0.0

0.5

1.0

1.5

.

1963

1966

1969

1972

1975

1978

1981

1984

1987

1990

1993

1996

1999

2002

2005

0.0

0.5

1.0

1.5

.

1963

1966

1969

1972

1975

1978

1981

1984

1987

1990

1993

1996

1999

2002

2005

Despite the much more favorable economic and climatic conditions in India, the yields for

sorghum and millet are almost the same in both regions, indicating a limited

transferability of the technology from India to SSA.

SorghumMillet

RiceMaize

Sorghum

Millet



13/53

Wheat

India vs. SSACereal Yields by crop and their growth: India vs. SSA

Yields (tons/ha) Growth

(times)

India

Yields (tons/ha) Growth

(times)

SSA

61-63 Avg. 05-07 Avg. 61-63 Avg. 05-07 Avg.

0.7 2.1 3.10.8 2.5 3.0 1.2

India Yield

SSA Yield

05-07 Avg.

12

The difference in current rice yield is huge, followed by maize. Room for the transfer of rice and maizetechnology?

When it comes to sorghum and millet, there would be limited transferability of technology from Asia to SSA.

In SSA, as far as the yield growth rate is concerned, a GR seems to be occurring in wheat, but not as much inthe other crops.Possible to expand the wheat area?

Maize

Millet

Sorghum

1.3 1.8

1.41.0 1.6 1.5

0.8 1.0 1.4

0.6 0.9 1.5

1.5 3.32.1

1.1 2.3 2.1

0.5 0.8 1.7

0.4 0.9 2.1

1.91.4

0.8

1.0Source: Authors calculation with FAOSTAT Data


14/53

Limitation of Wheat Expansion in SSAWheat production map of the world

(Average percentage of land used for wheat productiontimes average yield in each grid cell)

Temperate Zone and SSA

13

Source: Compiled by the University of Minnesota Institute on the Environment with data from: Monfreda, C., N.Ramankutty, and J.A. Foley. 2008. Farming the planet: 2. Geographic distribution of crop areas, yields,physiological types, and net primary production in the year 2000. Global Biogeochemical Cycles 22: GB1022

Wheat can be grown well only under acool climate, which is associated with thetemperate climate zone. In the Africancontinent, the temperate climate zone isfound only in limited part.

Wheat is thus grown only in the Republicof South Africa, the highlands in Ethiopia,and a few other regions, which is matchedwith the mere 3 percent of the total croparea planted to wheat.


15/53

India vs. SSA

Evolution of Cropping PatternsArea Harvested by Cereal Crop

Other

Maize

Sorghum

Millet

Million haAAGR

(%)

-4.00.8-1.7

-1.3

India

Other

Wheat Rice

AAGR

(%)

0.5

1.2

2.6

-0.5

SSAMillion ha

14


In India, the three GR crops seem to be

replacing Millet and Sorghum.

Wheat

Rice

1.6

0.5

Millet

Sorghum

Maize

1.2

1.5

In SSA, all crops except wheat are

spreading. In particular, rice recently.


16/53

India in Focus:

Diffusion of Irrigation and Modern Varieties

0.6

0.7

0.8

0.9

1.0

Proportion of Irrigated Area by Crop

Wheat

Rice

Wheat

Rice

Maize

Millet

Sorghum

Proportion of Area Sown to MVs by Crop

0.6

0.7

0.8

0.9

1.0

15

-

0.1

0.2

0.3

0.4

0.5

1970 1975 1980 1985 1990 1995 2000 2005

The irrigation coverage varies largely by crop, and

it has not been increasing considerably overtime.

Source: CMIE Database

Maize

MilletSorghum0.0

0.1

0.2

0.3

0.4

0.5

1970 1975 1980 1985 1990 1995 2000

There has been a rapid increase in area planted to

MVs, even for sorghum and millet in recent years,though their yields are not growing.


17/53

India in Focus:

Agro-Climate and Crop Choice

Temperature

()

Rainfall

(mm)

1998-2002 Five-Year Average

Millet

26.2

794

Sorghum

26.3

848

Maize

25.6

863

Rice

25.5

1,007

Wheat

22-23

852

16

Millet and Sorghum are grown in drier and slightly warmer environments.

Sources: India Water Portal; CMIE Database

Irrigation(%)

Yield

(kg/ha)

# Districts

18

1,001

269

14

821

258

34

1,825

327

58

2,007

412

79

2,153

356


18/53

Data Source: IndiaDistrict-Level Panel Data Construction Covering ~600 Districts

Variable Raw Data

Agricultural Output (by crop)Yield (by crop)

Area Sown (by crop) CMIE

CMIE

Source

ClimateTemperature

Rainfall CMIE

India Water Portal of the MD

17

The database is composed of five different sources,

including private research corporations: CMIE and Datanet India.

Commonly Available Years :1972-2002Notes: CMIE = Center for Monitoring Indian Economy Pvt. Ltd.,

MD = India Meteorological Department, X.Zhang@IFPRI , K. Kumar@WB

Irrigated Area (by crop)

MVs Adoption Rate (by crop)Technology (by crop)

CMIE

(Not Available on District Level)

Literacy RateControls

Population Density

X. Zhang, Datanet India

K. Kumar, Datanet India


19/53

Econometric Approach

Eliminate Sample Selection Bias by 2-Step Estimation (Heckman, 1979)

18

The inverse Mills ratio is calculated using the result of the probit estimation.


20/53

Econometric Approach

19

Since consistent data on technology (MV adoption rate and other) are unavailable on districtlevel, it is assumed that theyear dummies and the time trend variables capture the impacts of

technology.

The interaction terms between explanatory variables and time trend variables (e.g., Xt , Xt2)

are meant to examine whether there have been over-time changes in the impacts of climate and

other explanatory variables due to any technological change.


21/53

Theoretical Framework

Yield The marginal effect is the slope of

tangent on the yield curve.

It may differ from place to place.

It changes when agricultural

technology changes

Yield Function

1) Climate Effects on Crop Yield

20

ma e ar a e e.g. ra n a , empera ureflooddrought

Climate Variable (e.g. rainfall, temperature)

Yield

TV

Early MV

Newer MV

Researchers claim that early generations

of MVs are resource-demanding and

sensitive to harsh agro-climate. How about newer MVs?

Interesting to examine the changing

impacts

2) Changes in Climate Effects

?


22/53

Regression Results for India, 1972 to 2002

Rice Dependent Variable: Rice Yield (Ln) Estimated Coefficients on Selected Explanatory Variables:

Temperature

Coefficients

Temp

Temp t Temp t2

TempIrri

Explanatory Variable

Rainfall

Rain all t

***

***

***

0.1122

-0.00440.0001

-0.0684 ***

0.5828

-0.0234

***

**

Variable

Yield

Yield Function(Initial)

21

The positive impacts of temperature, rainfall, irrigation are found, which indicates the upward slopingpart of the yield function of each variable.

The result for population density is supportive of the induced innovation hypothesis of Hayami andRuttan (1985) which states that as population increases, increasing scarcity of land induces thedevelopment and di usion o land-saving and yield-enhancing technologies.

Statistical significance: *10%, **5%, ***1%

Irrigation

Coverage

PopulationDensity

Rainfall t2

RainfallIrri

Irri

Irri t Irri t2

PopDen

PopDen t PopDen t2

***

*

0.0001-0.1563

3.1820

0.0132

-0.0005

0.0914-0.0014

0.0001

***

**

***

1 C 11 %

1% pt. 3.2 %

1 % 0.09 % (elasticity=0.09)

1 % 0.58 % (elasticity=0.58)


23/53



Temperature

Coefficients

Temp

Temp t Temp t2

TempIrri


Rainfall

Rain all t

***

***

***

0.1122

-0.00440.0001

-0.0684 ***

0.5828

-0.0234

***

**

Time

Marginal Effect

(=Dependence)

Avg. over

the period

22

The impact of climatic variables decreases over time (at a diminishing rate):

The predicted irrigation effect (%/% pt.) increases over time but slows down: 3.2 (72) 3.5 (86) 3.6 (02)


Irrigation

Coverage

PopulationDensity

Rainfall t2

RainfallIrri

Irri

Irri t Irri t2

PopDen

PopDen t PopDen t2

***

*

0.0002-0.1563

3.1820

0.0132

-0.0005

0.0914-0.0014

0.0001

***

**

***

Time

Marginal Effect

(=Dependence)

Avg. over

the period

*


24/53



Temperature

Coefficients

Temp

Temp t Temp t2

TempIrri


Rainfall

Rain all t

***

***

***

0.1122

-0.00440.0001

-0.0684 ***

0.5828

-0.0234

***

**

Irrigation

Marginal Effect of Climate

(=Dependence on Climate)

23

It is indicated that irrigation can reduce the dependence of rice yield on climatic factors, to some extent.

The over-time changes in the impacts of climate are distinct from the influence of irrigation diffusion, since thatinfluence is controlled for by the climate-irrigation interaction terms.

Therefore, the critically important finding is thatthe dependence of rice yield on climate mitigated over timeregardless of the availability of irrigation, which cannot be understood without considering the impact of theadoption of MVs with shorter maturity and drought-tolerance traits.


Rainfall t2

RainfallIrri 0.0001-0.1563 ***


25/53


Wheat, Maize, Sorghum, and Millet Dependent Variable: Ln Yield Estimated Coefficients on Selected Explanatory Variables:

Temperature

Wheat Maize

**

***

***

Temp

Temp t Temp t2

TempIrri


Rainfall

Rain all t

-0.0162

-0.00310.0000

0.0456 ***

0.3209

-0.0187

***

*

-0.0662

0.0085-0.0002

0.0403 *

0.0709

-0.0015

***

***

Sorghum Millet

*

***

***

0.0401

-0.00210.0001

-0.0305

0.5066

-0.0333

***

0.0561

0.0030-0.0001

-0.1305 ***

0.2750

-0.0027

*

24

The impacts of climatic variables on crop yields decreased over time at a diminishing rate in severalcases. At least, in no single case, the impact of climate augmented.

Irrigation leads to a reduced climate dependence of crop yields. Induced innovation hypothesis is supported in all crops in recent years at least.


***

Irrigation

Coverage

***PopulationDensity

Rainfall t2

RainfallIrri

Irri

Irri t Irri t2

PopDen

PopDen t PopDen t2

0.0004

-0.1291

-0.0698

-0.0018

-0.0003

0.11560.0009

0.0000

***

***

0.0000

0.0011

-0.9403

0.0051

-0.0003

-0.21120.0127

-0.0001

***

***

0.0007

0.1191

0.4155

-0.0619

0.0017

0.1473-0.0009

0.0000

**

**

*

-0.0002

0.0782

3.2329

-0.0207

0.0004

0.08820.0090

-0.0003

**

***

***

***

***

**


26/53

Concluding Remarks

(1) Summary of the Findings

From the Descriptive Statistics

The gap in aggregate cereal yield between Asia and SSA was so minor until the early 1980s

despite the much more favorable climatic, economic, and political conditions in Asia. The yield diversion occurred due to the adoption of improved technology in Asia and the

failure of that in SSA. In other words, the Asian GR is likely to be a technology-led revolution.

The Asian GR technolo ies were develo ed rinci all for wheat and rice followed b maize. In

25

fact, Indian farmers have been steadily replacing the lower-yielding crops (sorghum and millet)

by the higher-yielding crops, which is one of the reasons why the compound cereal yield has

been growing in India.

Given the absence of the yield difference for sorghum and millet between Asia and SSA eventoday, the technology transferability from Asia to SSA for these two crops seems to be absolutely

limited.


27/53

Concluding Remarks

(1) Summary of the Findings From the Regression Results

The impacts of climatic conditions (temperature and rainfall) on crop yields have reduced overtime, due to the adoption of MVs and associated technologies.

i. The impact of temperature, whether the average is positive or negative, declined for rice

and maize.

ii. The rainfall effect mitigated for wheat, rice, and sorghum.

The traits of MVs have contributed to alleviating, not aggravating, the influence of climatic

conditions, which is in contrast with the conventional notion that MVs are typically

resource-demanding and are higher-yielding only under favorable environments.*

26

*A possible reason is that the short maturity varieties can grow up in a shortened period during which rainfall is assured. It is also likely that improved droughttolerance of MVs reduces the downward yield risk, which leads to a decrease in the marginal effect in the low range of rainfall.

Role of Irrigationi. Rice MVs require more irrigation water than do TVs. Interestingly enough, the rate of

increase in irrigation effect is relatively large in the initial phase of the GR, but slows

down in the later phase.

ii. Irrigation works to mitigate the influence of climate endowments on crop yields.

The induced innovation hypothesis proposed by Hayami and Ruttan (1985) is broadly supported. Continued population pressure is likely to have increased the relative profitability of land-saving

and yield-enhancing technologies along the lines of the hypothesis.


28/53

Concluding Remarks

(2) Policy Implications It is highly desirable to reverse the declining trend in the investment in international agricultural

research activities, to enhance agricultural productivity in regions with unfavorable climates

including SSA.

Facing a tight budget for international agricultural research, crop-wise foci would be necessaryto clarify policy priorities.

i. Rice: A critically important implication of this study should be a focus on rice as astrategic crop in SSA, because of the abundant evidences of improved resistance to harsh

climate, and the large gap in current yield between Asia and SSA, indicating an

27

opportunity to transfer the Asian technology.

ii. Maize: Since maize is the most widely cultivated crop in SSA, the productivity of maize

farming must be enhanced. The advantage of maize crop is that the yield is not adversely

affected by the unavailability of irrigation, meaning that maize has comparative advantage

in rain-fed farming systems. Therefore, maize can be the second strategic crop after rice.

Yet, it must be recognized that unlike rice, the maize technology developed in Asia is not

conducive to weakening the impact of drought on maize yield.iii. Wheat: The limitation of wheat area expansion in SSA requires due attention in spite of its

outstanding yield growth in the region.

iv. Sorghum and Millet: There exists no difference in yield at all between Asia and SSA.

Technologies for sorghum and millet in SSA are unlikely to be developed from the

experiences in Asia.


29/53

Concluding Remarks (Contd)

(2) Policy Implications Try to switch from low-performing crops (sorghum and millet) to high-performing crops (wheat,

rice, and maize), as it has been a driver for achieving growth in overall crop productivity in India.

Also in SSA, by the mid-20th century, maize had immigrated and replaced much of sorghum and

millet fields in Eastern and Southern Africa, partly because maize yielded more grain (Anthony1988).

Whether TVs or MVs, crop shift from sorghum and millet to rice and maize, wherever

applicable, is strongly suggested for fostering the agricultural productivity growth in SSA.

28

nvestment n rr gat on can e an e ect ve measure or tac ng ars agro-c mate en owments

in SSA, as well as the looming threat of climate change.


30/53

Thank you very much for listening.

29


31/53

Appendix

30


32/53

Introduction

Importance of Agricultural Productivity in Developing Countries

PovertyReduction

Economic

Growth

Agricultural

Productivity

Quotes: Food Security

31

Agricultural productivity plays a critical role in

economic growth, poverty reduction, food security in developing countries.

3/4 of the poor in sub-Saharan Africa live in rural areas where agriculture is a dominant sector.(WDR, 2008)1 % decrease in agricultural GDP leads to a decrease in consumption of the three poorest decile

groups by 4-6 %. (Ligon and Sadoulet, 2007)

33 % of the economic growth in sub-Saharan Africa from 1990 to 2005 comes from the

agricultural sector. (WDR, 2007)

1 % increase in agricultural GDP leads to an increase in expenditure of the poorest deciles by

>2.5 percent. The effect is superior to that of non-farm income. (Christiansen and Demery, 2007)


33/53

Policies(&Governance)


Irrigation (Water

Management)

Markets/Credit/



Climate

Endowments

Fertilizer

Introduction

Why has SSA missed the GR? (3)

32No hope for agriculture in SSA?

Public spending on African agriculture, including R&D, has fallen to the record low of


34/53

1) Impacts of Climate Endowments on Crop Productivity

Contribution of this Study

The changes in the impacts of climate endowments have yet to be well known since the

dynamic evidence has been scanty.

1) The (static) impacts of climate on agricultural crop yields. Somewhat known

2) The over-time changes in the impacts of climate endowments. Not well known

Agronomic Yield Function Approach (a.k.a. Crop Modeling)

33

Method: Specific crops experience differing climate in laboratories. Then, Yields Datavs. Climate Data (temperature, precipitation, etc.) are collected.

Shortcoming: Bias (i.e., unlikely to reflect possible adjustments by farmers)

Result: Sensitive to Climate (e.g., Mendelsohn et al., 1994)

Cross Sectional Regression (a.k.a. Ricardian Approach)

Method: Empirically regress crop productivities (e.g., land prices) on climate

variables, plus other controls.


35/53


Cross Sectional Regression (Contd)

Result: Not as sensitive as in crop modeling approach.

Mostly on developed countries due to the data availability:

U.S. (Adams et al., 1995; Mendelsohn et al., 1994)

Other developed countries (Olesen and Bindi, 2002; Bruce et al.,

1996; Reilly et al., 1996)

34

Developing countries:

India and Brazil (Seo and Mendelsohn, 2007; Sanghi et al., 1998).

Negative impact of temperature and rainfall

Shortcoming: Omitted variable problems (e.g., unobservable skills of farmers, soil

quality) which could generate a bias of unexplained sign or magnitude.


36/53


Panel Data Approach

Method: District Fixed effect/Random effect is controlled for.

Shortcoming: Often unfeasible due to the data constraints, especially for developing countries.

Result: ambiguous or negative impact of temperature

US county-level analysis: Deschnes and Greenstone (2007), Schlenker and Roberts (2006)

India state-level analysis: Auffhammer et al. (2006)

35


37/53


Another Unique Aspect: Asia-Africa Comparison.

Although the quality and availability of data are inferior for the African study, the direct

comparison can assess the difference in the progress of technology adoption in the two regions.

The choice of India Similarities in agriculture between India and SSA

Diversity in agro-climate, resulting in similar cropping patterns.

Differing poverty incidence.

36

Relatively small average farm size.

Although there are signs of hope documented in some case studies on agricultural technological

situations in African countries (Diagne, 2006; Sakurai, 2006; Goufo, 2008; Kajisa and Payongayong, 2008;

Kijima et al. ,2006), the real challenge is to translate individual successes into sustainable and

systematic improvements in agricultural performance, which facilitate the identification of policy

priorities.

In order to achieve this goal, it is important to accumulate hard evidence to design appropriate

development strategies (Otsuka and Kijima, 2010).

This study, therefore, is expected to provide positive evidence through solid econometric analyses.

India in Focus:


38/53

India in Focus:

Irrigation and Crop Yields

Temperature

()

Millet

26.1 25.7

Irrigation Coverage

Low High

Sorghum

26.2 25.4

Irrigation Coverage

Low High

Maize

25.1 26.1

Irrigation Coverage

Low High

Rice

24.6 25.9

Irrigation Coverage

Low High

Wheat

23.4 25.7

Irrigation Coverage

Low High

1998-2002 Picture

37

Clearly, irrigation coverage is higher in rain scarce districts. Even under dry climates, irrigation

boosts the yields largely for wheat and rice, but not as much for the other crops.

Source: India Water Portal; CMIE Database

Rainfall

(mm)

Yield

(kg/ha)

# Obs for

crop yield

Irrigation Coverage = % of Sown Area for each crop

High > 50 %; Low < 50 %

811

964

228

701

1,207

41

877

836

224

658

720

34

895

1,630

215

802

2,198

112

1,127

1,418

178

920

2,455

234

1,045

1,203

65

809

2,365

291

Note: There are observations for which irrigation coverage is unknown

India in Focus:


39/53

India in Focus:

Irrigation and Crop Yields

Temperature

()

1971-1975 Picture

25.7 26.6

Low High

Millet

Irrigation Coverage

25.8 24.9

Low High

Sorghum

Irrigation Coverage

25.4 25.3

Low High

Maize

Irrigation Coverage

25.2 25.6

Low High

Rice

Irrigation Coverage

25.2 25.3

Low High

Wheat

Irrigation Coverage

38

Source: India Water Portal; CMIE Database

Rainfall

(mm)

Yield

(kg/ha)

# Obs for

crop yield

Looking back at the early 70s, the role of irrigation did not seem as crucial as in the late 90s.

Irrigation Coverage = % of Sown Area for each crop

High > 50 %; Low < 50 %

Note: There are observations for which irrigation coverage is unknown

916

522

145

801

752

7

961

568

159

735

601

10

1,032

1,041

146

898

1,276

41

1,138

858

130

963

1,368

75

1,134

1,036

100

868

1,430

111

D S SSA


40/53

Data Source: SSACountry-Level Panel Data Construction

Variable Raw Data

Agricultural Output (by crop)Yield (by crop)

Area Sown (by crop) FAOSTAT

FAOSTAT

Source

ClimateTemperature (TBR)

Rainfall (TBR) GOSIC

GOSIC

39

The database combines data from four public sources: Technology variables are unavailable over the

long period. Moreover, the database has many missing observations across countries.

The database covers :1967-2004Notes: GOSIC = The Global Observing Systems Information Center of the U.S.

TBR = To be replaced by new data.

Prices (by crop) Nominal Price of OutputDeflator

FAOSTATWDI

Literacy RateControls

Population Density (TBR)

UNESCO

WDI


41/53

Approach: SSA

Step 1 is dispensed with

40

For SSA, the sample selection model does not work out, probably because the number ofcross-sectional observations is not large(~30) and each crop is grown in many of those

countries. Thus, I directly perform the outcome estimations assuming that the biases are

negligible.

Otherwise, the methodology is largely the same as in the case of India, except that there

is no data for irrigation, and price is available.


42/53

Approach (Contd)

Consideration on Endogeneity

Iijt= Irrigated land area for crop i divided by total area sown to crop i (India)

Again, instruments are absent. However, in the early stage of the GR, most of the irrigation wasgravity irrigation which was installed by the public sector. Therefore, irrigation can be consideredfairly exogenous especially in the early stage.

District-s ecific effect model ma mitigate, if not eliminate, the endogeneit bias because irrigation

41

investment can be determined based on time-invariant factors such as district-specific geography

and environment.

It is assumed that the endogeneity of these variables is not serious in this analysis.

h ( d)


43/53

Approach (Contd)

For SSA: Two Specifications for Yield Functions

[Model 1] With Year Dummies (from 1968 to 2004, 1967 as the base year) Without Time Trend Variables

Two-way Fixed Effect Model

To absorb the average yearly change in yield that is not explained by the explanatory variables

42

Aggregate macroeconom c an c mat c s oc s

Overall technological improvements

[Model 2] Without Year Dummies

With Time Trend Variables; tand t

2

(t= 0 for 1967)

To capture the trend in general technological improvement and its acceleration (or deceleration) whichis not picked up by the interaction terms (Xtand Xt2).

Four specifications for SSA: M1 w/o P; M2 w/o P; M1 w/P; M2 w/P

Fixed Effect Regression Results for SSA, 1967 to 2004


44/53


Wheat

Dependent Variable: Wheat Yield (Ln) Estimated Coefficients on Selected Explanatory Variables:

Temperature

Model 1

Temp

Temp t Temp t2


0.2699

-0.0229

0.0004

*0.0670

-0.0077

0.0001

0.4518

-0.0277

0.0004

0.1445

-0.0091

-0.0001

Model 2 Model 1 Model 2

Without Price With Price

***

***

***

***

***

***

*

**

**

Model 1: Year Dummies

Model 2: Time Trend

43

The impact of temperature is positive but decreases over time (at a diminishing rate).

The impact of rainfall is almost insignificant.

The effect of population density is initially very significantly positive, which is supportive of the inducedinnovation hypothesis. But the effect weakens over time. Exhaustion of technology?


Rainfall

Population

Density

Rainfall

Rainfall t Rainfall t2

PopDen

PopDen t

PopDen t2

-0.2229

0.0149

-0.0001

1.6184

-0.1279

0.0017

*

***

-0.1300

0.0054

0.0001

1.3718

-0.0913

0.0010

*

*

***

-0.4205

0.0426

-0.0007

0.1473

-0.0009

0.0000

-0.1841

0.0199

-0.0003

0.0882

0.0090

-0.0003***

**

***

***

** ***

***

***

***

***



45/53

g ,

Rice

Dependent Variable: Rice Yield (Ln) Estimated Coefficients on Selected Explanatory Variables:

Temperature

Model 1

Temp

Temp t Temp t2


-0.1485

0.0136

-0.0002

-0.1339

0.0111

-0.0002

-0.1425

0.0136

-0.0003

-0.1154

0.0103

-0.0002



**

***

***

*

***

***

*

***

***


Model 2: Time Trend

**

***

***

44

The declining impacts of climate are found for rainfall as well as temperature.

The effect of population density is very significantly positive, which is supportive of the induced

innovation hypothesis. Unlike wheat, the effect increases over time (at a diminishing rate).


Rainfall

Population

Density

Rainfall


PopDen

PopDen t

PopDen t2

0.6256

-0.0413

0.0006

1.4199

0.0531

-0.0012

***

***

0.5375

-0.0310

0.0005

1.6446

0.0413

-0.0010

**

0.3229

-0.0200

0.0003

1.9960

0.0460

-0.0011

0.1102

0.0005

0.0000

2.0197

0.0334

-0.0009***

**

***

***

** **

***

**

***

***

***

**

*

*



46/53

g ,

Maize

Dependent Variable: Maize Yield (Ln) Estimated Coefficients on Selected Explanatory Variables:

Temperature

Model 1

Temp

Temp t Temp t2


-0.0237

-0.0015

0.0000

*-0.0001

-0.0051

0.0001

0.0815

-0.0093

0.0001

0.0594

-0.0095

0.0001



***

*

***

**


Model 2: Time Trend

45

The declining impacts of rainfall is found.

Unlike wheat and rice, the effect of population density is mostly insignificant.


Rainfall

Population

Density

Rainfall


PopDen

PopDen t

PopDen t2

0.6656

-0.0560

0.0011

0.1530

0.0225

-0.0004

0.5151

-0.0371

0.0007

0.3365

0.0101

-0.0001

0.1644

-0.0136

0.0004

-0.3977

-0.0076

0.0002

0.1574

-0.0073

0.0002

-0.7004

-0.0076

0.0002

**

*

***

***

******

***

***



47/53

g ,

Sorghum

Dependent Variable: Sorghum Yield (Ln) Estimated Coefficients on Selected Explanatory Variables:

Temperature

Model 1

Temp

Temp t Temp t2


-0.1902

0.0060

-0.0001

-0.1392

0.0017

0.0000

-0.1138

0.0012

0.0000

-0.1041

-0.0002

0.0000



***

*

* *


Model 2: Time Trend

***

46

The declining impact of temperature is found in Model 1 without price.

The impact of rainfall is totally insignificant.

The induced innovation hypothesis is not supported for sorghum in SSA.


Rainfall

Population

Density

Rainfall


PopDen

PopDen t

PopDen t2

0.1685

-0.0081

0.0000

-1.7672

0.0023

0.0001

0.0415

0.0064

-0.0002

-1.5738

-0.0122

0.0004

***

0.2404

-0.0065

0.0000

0.5120

-0.0486

0.0009

0.1634

0.0077

-0.0003

0.4091

-0.0529

0.0010**

*** ***

***

***

***



48/53

g

Millet

Dependent Variable: Millet Yield (Ln) Estimated Coefficients on Selected Explanatory Variables:

Temperature

Model 1

Temp

Temp t Temp t2


-0.1822

0.0057

-0.0001

-0.0986

0.0024

0.0000

0.0405

-0.0108

0.0001

0.0818

-0.0102

0.0001



**

**

*

**


Model 2: Time Trend

47

The declining impacts of climate is not found for millet in SSA.

The impact of rainfall is totally insignificant.

The induced innovation hypothesis is not supported for millet in SSA.


Rainfall

Population

Density

Rainfall


PopDen

PopDen t

PopDen t2

0.2547

-0.0217

0.0005

-0.1933

-0.0092

0.0002

*

0.1610

-0.0132

0.0004

-0.0358

-0.0172

0.0004

*

0.1788

-0.0266

0.0007

-0.4396

0.0084

-0.0001

0.0360

-0.0101

0.0004

-0.2293

-0.0032

0.0002*

**

I I A ll T h l Th M d?


49/53

Is It Actually Technology That Mattered?

The results strongly indicate that the impact of climatic factors on crop yields have declined over time,

for wheat, rice, maize, and sorghum, after the irrigation effects are controlled for (in India).

Although it seems reasonable to assume that technological progress represented by the adoption of high-

yielding MVs and other improved production practices has contributed to these over-time changes, it is

not directly proven by the regression analyses since the time trend variables can reflect the effects of a

variety of factors including infrastructure, among other things.

The difficulty is that technology variables, such as MV adoption rate, are unavailable at the district level

48

n t e case o n a. oreover, even t ose var a es were ava a e, t e r use wou enta a pro em o

endogeneity bias, which would not be easy to correct for.

One attempt to obtain a more direct evidence of the impact of technology is to use irrigation as a proxy

for the compound effects of irrigation and MVs if the correlation between irrigation rate and MV adoption

rate is high.

Thus, I propose to investigate the relationship between MV adoption rate and irrigation rate using the

state-level data.

I It A t ll T h l Th t M tt d? (C td)


50/53

Coefficient of Correlation between Modern Variety Adoption Rate and Irrigation Rate,

State-wise, by Crop, Three-Year Moving Averages

Is It Actually Technology That Mattered? (Contd)

49

Source: Authors calculation with data from Indiastat and Center for

Monitoring Indian Economy.

Period Wheat Rice Maize Sorghum Millet

1974-1988 0.76 0.79 0.62 -0.27 0.30

1989-2002 0.38 0.35 0.46 -0.08 0.09

This trend is supported by preceding studies by Janaiah et al.(2006), Gollin (2006), and Byerlee (1996),

stating that MVs were adopted primarily in irrigated areas in the early phase of the GR.

Use district-level irrigation rate for wheat, rice, and maize in the early phase of the GR, as a proxy

for district-level MV adoption rate.

Regression with the Proxy Variable: India


51/53

Regression with the Proxy Variable: India Dependent Variable: Ln Yield

Time Period: 1974 to 1988

Estimated Coefficients on Climate Variables:

Temperature

Wheat Rice

***

***

Temp Temp t TempIrri (Tech)


-0.0211-0.0057

0.0440

***

***

0.1135-0.0048

-0.0635

***

Maize

*0.0707

-0.0017

0.0182

Irrigation = Technology Indicator

50

It is confirmed that the rainfall elasticity of rice yield decreases by 0.0037 when the MV adoption rateincreases by 1 percentage points.

Difficulty: Early generations of MVs may be more resource-demanding.


Rainfall ***

Irrigation

Coverage

Rainfall

Rainfall t RainfallIrri (Tech)

Irri (Tech)

Irri (Tech) t

.

-0.0141

0.0635

-0.3754

-0.0083

.

-0.0046

-0.3741

4.3988

0.0059

**

.

0.0083

-0.0071

-0.4557

0.0171 **

***

***



52/53

Concluding Remarks (Cont d)

(3) Remaining Issues

Variables that directly represent technology adoption are missing in the analyses.

i. Only the state level MV adoption rate is available in Indias descriptive statistics.

ii. The MV adoption rate, even if it is available, does not express the quality of the MVs, and

thus, does not reflect the continuous improvement in the traits of MVs. The MV

adoption rate may understate the actual effect of available technology on crop yields.

51

. e mpact o s o sorg um an m et on y e s s unc ear s nce t e recent y surg ng

MV adoption rates for these two crops do not lead to the yield growth apparently.

TheMV adoption rate, in this sense, may overstate the actual effect of available technology.

Finding a much more refined indicator of technology would produce more reliable results.

The regressions employed in the analyses are not weighted regressions: i.e., all the districts in

India, larger ones and smaller ones, are treated with equal importance. So are all the countries inSSA. Since there are major and minor districts and countries, it may be preferable to contrive a

measure to take some weighting factor into account, especially for SSA where countries of a

range of economic sizes are included.



53/53

Concluding Remarks (Cont d)

(3) Remaining Issues

The quality and availability of the data for SSA have to be improved if possible.

i. The number of cross-sectional observations is limited and many small countries are left

out of the regressions.

ii. The price variable employed in the SSA analyses should be refined, in terms of both

quality and availability.

52

iii. Another suggestion may be using rural population density in place of national population

density in the country as a whole, since most of agriculture is undertaken by rural farmers.

Throughout this study, the agricultural productivity is expressed in terms of the physical cropyield. Although it would be a daunting task, expressing the productivity in monetary term or

using total factor productivity, instead of physical crop yield, may be an intellectually

stimulating challenge.

changes in the agro-climate effects on cereal crop yields: panel evidence from india (1972-2002)...

Documents